JP6665610B2 - Motor control device - Google Patents

Description

  The present invention relates to a motor control device that controls the position of a predetermined member (control target) such as a table or an arm using an electric motor in, for example, a machine tool or a robot.

  For example, Patent Document 1 discloses a motor-based position control device that causes a motor position detection value to follow a position command value. The position control device includes: a speed reference model that outputs a speed reference command value; a speed controller (speed feedback unit) that obtains a control amount (current command value) according to a deviation between the speed reference command value and the actual speed; And a speed feedforward section for outputting a speed feedforward output signal. Further, the position control device includes a position reference model that outputs a position reference command value, and a position controller (position feedback unit) that obtains a control amount (speed command value) according to a deviation between the position reference command value and the actual position. ), And a position feedforward section for outputting a position feedforward output signal. The input to the position feedforward unit is a position command value input to the position reference model, and the input to the speed feedforward unit is a speed command value input to the speed reference model.

  Further, for example, Patent Literature 2 discloses an embedded intelligent controller that performs position control on a control target by a position command and performs force control by a force command. It also includes a fuzzy neural network for a position control system, a fuzzy neural network for a force control system, and a switching unit for continuously switching between position control and force control by hybrid position / force control.

JP 2009-303432 A JP 2014-6566 A

  The position control device described in Patent Document 1 requires a position reference model and a speed reference model, which are models to be controlled, and is very difficult to apply to a system in which modeling of static friction or the like is difficult. . In addition, the input to the speed feedforward section is a speed command value, and there is a possibility that a position deviation or a speed deviation that greatly affects the accuracy of a machine tool, a robot, or the like cannot be effectively reduced.

  Further, the built-in intelligent controller described in Patent Literature 2 is a fuzzy neural network for smoothly switching between position control and force control in an industrial robot assembling work that requires not only position control but also force control. I use. However, the control is complicated, and it is difficult to apply to a control system that controls the position of a predetermined member such as a table of a machine tool or an arm of a robot with higher accuracy.

  The present invention has been made in view of such a point, and is a motor control device that controls the position of a predetermined member (control object) using an electric motor, and further reduces a position deviation and a speed deviation. It is an object of the present invention to provide a motor control device that can perform the control.

  In order to solve the above problems, a motor control device according to the present invention employs the following means. First, a first invention of the present invention is a motor control for controlling the position of the control object using an electric motor for moving the position of the control object and position detecting means for detecting a position related to the electric motor. A position deviation calculation unit that calculates a position deviation that is a deviation between a command position for the electric motor and an actual position that is an actual position based on a detection signal from the position detection unit; A position feedback control unit that performs feedback control according to the deviation to output a first provisional command speed; a lower command speed that is a lower command speed for the electric motor and includes the first provisional command speed; A speed deviation calculation unit that calculates a speed deviation that is a difference between the actual speed that is the actual speed based on the detection signal from the control unit and feedback control according to the speed deviation. A speed feedback control unit that outputs a first provisional command current, a speed feedforward control unit that performs feedforward control according to an upper command speed that is different from the lower command speed and outputs a second provisional command current, A current addition operation unit that outputs the command current by adding the first temporary command current and the second temporary command current; and a current output unit that outputs a drive current to the electric motor based on the command current. Have. The speed feedforward control unit receives a higher-order command input, and outputs a speed-side acceleration input unit that outputs the input higher-order command acceleration as a speed-side acceleration output; and the higher-order command speed is input and input. A speed-side speed input unit that outputs the higher-order command speed as a speed-side speed output, and a limited speed range limited to a speed range of the higher-order command speed is divided into a plurality of preset adjacent speed ranges, The upper command speed is provided corresponding to a plurality of boundary speeds that are the boundaries of the divided speed ranges, and the higher command speed is input, and a speed-side boundary speed output is output from a portion corresponding to the higher command speed. A plurality of speed-side boundary speed inputs, and a speed-side first output that is the speed-side acceleration output, the speed-side speed output, and the speed-side boundary speed outputs; A plurality of speed-side first learning weights corresponding to the speed-side first weight learning unit, which changes the plurality of speed-side first learning weights according to the speed deviation, each of the speed-side first outputs, and each of the speed-side first outputs. A speed-side output unit that adds a plurality of speed-side first multiplied values multiplied by the corresponding speed-side first learning weight and outputs the sum as the second provisional command current. It is.

  Next, a second invention of the present invention provides a motor for controlling the position of the control object using an electric motor for moving the position of the control object and position detecting means for detecting a position related to the electric motor. A control device, a position deviation calculation unit that calculates a position deviation that is a deviation between a command position for the electric motor and an actual position that is an actual position based on a detection signal from the position detection unit; A position feedback control unit that performs a feedback control according to a position deviation to output a first provisional command speed; a lower command speed that is a lower command speed for the electric motor and includes the first provisional command speed; A speed deviation calculating unit that calculates a speed deviation that is a difference between an actual speed that is an actual speed based on a detection signal from the means, and feedback control according to the speed deviation. A speed feedback control unit that outputs a first provisional command current, a speed feedforward control unit that performs feedforward control according to an upper command speed that is different from the lower command speed and outputs a second provisional command current, A current addition operation unit that outputs the command current by adding the first temporary command current and the second temporary command current; and a current output unit that outputs a drive current to the electric motor based on the command current. Have. The speed feedforward control unit receives the upper command speed, outputs a speed-side positive speed output value when the input higher command speed is positive, and the input higher command speed is negative. In the case, the speed-side positive / negative speed firing unit that outputs the speed-side negative speed output value, and a limited speed range limited to the speed range of the upper command speed, is divided into a plurality of adjacent speed ranges set in advance, It has a plurality of boundary speeds that are boundaries between the divided speed ranges, the upper command speed is input, and a speed difference between the upper command speed and the plurality of boundary speeds is within a predetermined speed difference. A speed-side boundary speed firing unit that outputs a speed-side boundary speed output value based on the speed difference from the boundary speed, and a higher-order command acceleration is input, and the input higher-order command acceleration is output as a speed-side acceleration output. A speed-side positive speed input unit that receives and outputs the speed-side normal speed output value as a speed-side normal speed output; and A speed-side negative speed output value is input, and a speed-side negative speed input unit that outputs the input speed-side negative speed output value as a speed-side negative speed output, and is provided corresponding to a plurality of the boundary speeds, A plurality of speed-side boundary speed input units for receiving a speed-side boundary speed output value and outputting the input speed-side boundary speed output value as a speed-side boundary speed output; the speed-side acceleration output; and the speed-side positive speed A plurality of speed-side first learning weights corresponding to each of the speed-side first outputs, which are output, the speed-side negative speed output, and the plurality of speed-side boundary speed outputs, according to the speed deviation; A first weight learning unit; A plurality of speed-side first multiplied values obtained by multiplying each of the speed-side first outputs and the speed-side first learning weight corresponding to each of the speed-side first outputs are added, and the second provisional command is obtained. A speed-side output unit that outputs a current.

  Further, a third invention of the present invention is the motor control device according to the second invention, wherein the position feedforward control section for performing feedforward control according to the upper command speed and outputting a second provisional command speed. A speed addition operation unit that adds the first provisional command speed and the second provisional command speed to output the lower command speed. The position feed forward control unit outputs the position-side positive speed output value when the upper command speed is input, and the input upper command speed is positive, and the input higher command speed is negative. A position-side positive / negative speed firing unit that outputs a position-side negative speed output value, and a plurality of the boundary speeds, the higher-order command speed is input, and a speed with the upper-order command speed among the plurality of the boundary speeds A position-side boundary speed firing unit that outputs a position-side boundary speed output value based on the speed difference from the boundary speed whose difference is within a predetermined speed difference, and the higher-order command acceleration is input; A position-side acceleration input unit that outputs the position-side normal speed output value, and a position-side normal speed input that outputs the position-side normal speed output value as a position-side normal speed output. Unit, the position-side negative speed output value is input, the position-side negative speed input unit that outputs the input position-side negative speed output value as a position-side negative speed output, and a plurality of the boundary speeds. Prepared, the position-side boundary speed output value is input, a plurality of position-side boundary speed input units that output the input position-side boundary speed output value as a position-side boundary speed output, and the position-side acceleration output, A plurality of position-side first learning weights corresponding to each of the position-side first speed outputs, the position-side negative speed outputs, the position-side boundary speed outputs, and the position-side first outputs, according to the position deviation. A position-side first weight learning unit to be changed; and a plurality of position-side first weights obtained by multiplying each of the position-side first outputs and the position-side first learning weight corresponding to each of the position-side first outputs. The multiplied value is added and the second temporary It has a position side output section that outputs a decree speed, a motor control device.

  According to a fourth aspect of the present invention, there is provided a motor control for controlling the position of the control target using an electric motor for moving the position of the control target, and position detecting means for detecting a position related to the electric motor. A position deviation calculation unit that calculates a position deviation that is a deviation between a command position for the electric motor and an actual position that is an actual position based on a detection signal from the position detection unit; A position feedback control unit that performs feedback control according to the deviation to output a first provisional command speed; a lower command speed that is a lower command speed for the electric motor and includes the first provisional command speed; A speed deviation calculation unit that calculates a speed deviation that is a difference between the actual speed that is the actual speed based on the detection signal from the control unit and feedback control according to the speed deviation. A speed feedback control unit that outputs a first provisional command current, a speed feedforward control unit that performs feedforward control according to an upper command speed that is different from the lower command speed and outputs a second provisional command current, A current addition operation unit that outputs the command current by adding the first temporary command current and the second temporary command current; and a current output unit that outputs a drive current to the electric motor based on the command current. Have. The speed feedforward control unit receives the upper command speed, outputs a speed-side positive speed output value when the input higher command speed is positive, and the input higher command speed is negative. In the case, the speed-side positive / negative speed firing unit that outputs the speed-side negative speed output value, and a limited speed range limited to the speed range of the upper command speed, is divided into a plurality of adjacent speed ranges set in advance, It has a plurality of boundary speeds that are boundaries between the divided speed ranges, the upper command speed is input, and a speed difference between the upper command speed and the plurality of boundary speeds is within a predetermined speed difference. From the boundary speed, a speed-side boundary speed firing unit that outputs a speed-side boundary speed output value based on the speed difference, and a higher-order command acceleration is input, and the input higher-order command acceleration is prepared in advance. A speed-side acceleration input unit that outputs a speed-side acceleration output toward each of the number of speed-side calculation units, and the speed-side normal speed output value is input, and the input speed-side normal speed output value is divided into a plurality of A speed-side positive speed input unit that outputs a speed-side positive speed output toward each of the speed-side calculation units, and the speed-side negative speed output value is input, and the input speed-side negative speed output value is a plurality of values. A speed-side negative speed input unit that outputs as a speed-side negative speed output toward each of the speed-side calculation units, and is prepared in correspondence with a plurality of the boundary speeds, and the speed-side boundary speed output value is input; The input speed-side boundary speed output value, a plurality of speed-side boundary speed input units that output as a speed-side boundary speed output toward each of the plurality of speed-side calculation units, and a plurality of the speed-side acceleration outputs, More than one A plurality of speed-side first learning weights corresponding to each of the speed-side first outputs, which are a degree-side positive speed output, a plurality of the speed-side negative speed outputs, a plurality of the speed-side boundary speed outputs, to the speed deviation. A speed-side first weight learning unit that changes in accordance with each of the speed-side first outputs; and a plurality of speed-side weights obtained by multiplying each of the speed-side first outputs by the speed-side first learning weight corresponding to each of the speed-side first outputs. A plurality of speed-side arithmetic units for adding the first multiplied value and outputting as a speed-side second output; and a plurality of speed-side second learning weights corresponding to each of the speed-side second outputs, to the speed deviation. A speed-side second weight learning unit that changes in accordance with each of the speed-side second outputs, and a plurality of speed-side weights obtained by multiplying each of the speed-side second outputs by the speed-side second learning weight corresponding to each of the speed-side second outputs. The second multiplied value is added as the second provisional command current. And a speed-side output unit for outputting.

  According to a fifth aspect of the present invention, there is provided the motor control device according to the fourth aspect, wherein the position feed-forward control unit outputs a second provisional command speed by performing feed-forward control according to the upper command speed. A speed addition operation unit that adds the first provisional command speed and the second provisional command speed to output the lower command speed. The position feed forward control unit outputs the position-side positive speed output value when the upper command speed is input, and the input upper command speed is positive, and the input higher command speed is negative. A position-side positive / negative speed firing unit that outputs a position-side negative speed output value, and a plurality of the boundary speeds, the higher-order command speed is input, and a speed with the upper-order command speed among the plurality of the boundary speeds A position-side boundary speed firing unit that outputs a position-side boundary speed output value based on the speed difference from the boundary speed whose difference is within a predetermined speed difference, and the higher-order command acceleration is input; A position-side acceleration input unit that outputs a position-side acceleration output value to each of a plurality of position-side operation units prepared in advance, and the position-side normal speed output value to which the position-side normal speed output value is input. A position-side positive speed input unit that outputs a force value as a position-side positive speed output toward each of the plurality of position-side operation units, and the position-side negative speed output value that is input, and the input position-side negative speed output value is input. A position-side negative speed input unit that outputs a speed output value as a position-side negative speed output toward each of the plurality of position-side calculation units; and a position-side negative speed input unit that is prepared in correspondence with the plurality of the boundary speeds. An output value is input, a plurality of position-side boundary speed input units that output the input position-side boundary speed output values as position-side boundary speed outputs toward each of the plurality of position-side operation units, A plurality of position-side second outputs corresponding to the position-side first outputs, which are the position-side acceleration output, the plurality of position-side positive speed outputs, the plurality of position-side negative speed outputs, and the plurality of position-side boundary speed outputs. 1 learning weight, A position-side first weight learning unit that changes in accordance with the deviation, a plurality of values obtained by multiplying each of the position-side first outputs and the position-side first learning weight corresponding to each of the position-side first outputs. A plurality of position-side arithmetic units for adding the position-side first multiplied values and outputting the same as a position-side second output; and a plurality of position-side second learning weights corresponding to each of the position-side second outputs, A position-side second weight learning unit that changes in accordance with the position deviation; a plurality of values obtained by multiplying each of the position-side second outputs by the position-side second learning weight corresponding to each of the position-side second outputs; And a position-side output unit that adds the position-side second multiplied value and outputs the sum as the second provisional command speed.

  According to a sixth aspect of the present invention, there is provided the motor control device according to any one of the first to fifth aspects, wherein the speed and the friction at the time of moving the position of the controlled object are changed. And a speed / physical phenomenon characteristic indicating a relationship between the limited speed range which is regarded as a non-linear characteristic, and divided into the respective speed ranges which can be partially regarded as a linear characteristic. And a speed that is a boundary between the divided speed ranges is set as the boundary speed.

  Further, a seventh invention of the present invention is the motor control device according to any one of the second invention to the fifth invention, wherein the speed-side boundary speed firing section has a predetermined speed. A distribution function having a predetermined distribution function having a spread width, and a distribution probability corresponding to a speed apart from the center of the distribution function by the speed difference between the input higher-order command speed and the boundary speed, Determined for each of the boundary velocities using a distribution function, the determined distribution probability is fired only for the non-zero boundary speed, and based on the determined distribution probability when the upper command speed is positive. A motor control device that outputs a positive value, a negative value based on the distribution probability obtained when the upper command speed is negative, as the speed-side boundary speed output value corresponding to the boundary speed at which the ignition occurred. It is.

  Further, an eighth invention of the present invention is the motor control device according to the third invention or the fifth invention, wherein the position-side boundary speed firing section increases the width of the preset speed to the expanded width. Having a predetermined distribution function, and a distribution probability corresponding to a velocity separated from the center of the distribution function by the velocity difference between the input higher-order command velocity and the boundary velocity, using the distribution function. For each of the boundary velocities, fire only for the boundary speed for which the obtained distribution probability is not zero, and when the upper command speed is positive, a positive value based on the obtained distribution probability is calculated. A motor control device that outputs a negative value based on the obtained distribution probability as the position-side boundary speed output value corresponding to the fired boundary speed when the upper command speed is negative.

  In the first invention, a speed-side boundary speed input unit and a speed-side first learning weight are prepared for each of a plurality of speed ranges, and a speed-side boundary speed output output from the speed-side boundary speed input unit corresponding to the upper command speed. Is multiplied by the speed-side first learning weight, the result is reflected on the command current of the speed feedforward control unit, and learning is performed in which the speed-side first learning weight is changed in a direction in which the speed deviation decreases. As described above, since the speed-side first learning weight is changed for each of the plurality of speed ranges in accordance with the upper command speed and the speed deviation, it is possible to provide a motor control device with a small speed deviation for all speeds.

  In the second invention, a speed-side boundary speed firing unit, a speed-side boundary speed input unit, and a speed-side first learning weight are prepared for each of a plurality of speed ranges, and a speed-side boundary speed firing unit corresponding to a higher-order command speed is provided. Are fired to output a speed-side boundary speed output from the plurality of speed-side boundary speed input units. Each speed-side boundary speed output is an output corresponding to the speed difference between the higher-order command speed and the boundary speed. A plurality of speed-side boundary speed outputs are multiplied by the speed-side first learning weights, respectively, and the sum of the outputs is multiplied by the speed feed. The learning is performed by changing the speed-side first learning weight in a direction in which the speed deviation is reduced by reflecting the command current of the forward control unit. As described above, the speed-side first learning weight is changed for each of the plurality of speed ranges in accordance with the upper command speed and the speed deviation, and a plurality of speed-side boundary speed outputs are output for one upper command speed. Each speed-side boundary speed output is an output corresponding to a speed difference between a higher-order command speed and a boundary speed, and is obtained by multiplying a plurality of speed-side boundary speed outputs by speed-side first learning weights, respectively, and obtaining a sum of these. The value has a continuity of speed as compared with the first invention, and it is possible to provide a motor control device having a smaller speed deviation with respect to all speeds.

  In the third invention, the same configuration as that of the speed feedforward control unit of the second invention is applied to the position feedforward control unit, and the position side first position of the position feedforward control unit is controlled according to the upper command speed and the position deviation. Since the learning weights are learned, it is possible to provide a motor control device with less speed deviation and position deviation for all speeds.

  The speed feedforward control unit of the second invention is a perceptron having each input unit, a speed-side first weight learning unit, and a speed-side output unit, and the speed feedforward control unit of the fourth invention has This is a neural network having an input unit, a speed-side first weight learning unit, a speed-side operation unit, a speed-side second weight learning unit, and a speed-side output unit. The outputs from the plurality of input units are each multiplied by a weight, and the combination pattern of the sum is more varied in the neural network than in the perceptron. A motor control device with a small speed deviation can be provided.

  A position feedforward control unit according to a third invention is a perceptron having each input unit, a position-side first weight learning unit, and a position-side output unit, and the position feedforward control unit according to the fifth invention includes: This is a neural network having an input unit, a position-side first weight learning unit, a position-side operation unit, a position-side second weight learning unit, and a position-side output unit. The outputs from the plurality of input units are each multiplied by a weight, and the combination pattern of the sum is more versatile in the neural network than in the perceptron. It is possible to provide a motor control device with small speed deviation and position deviation.

  In the sixth aspect, the plurality of speed ranges are applied to a speed range in which a predetermined physical phenomenon including friction is non-linear with respect to the speed. And a motor control device with less noise.

  According to the seventh aspect, a distribution function that varies in position according to a higher-order command speed and spans a plurality of speed ranges is used, and each speed-side boundary speed output is obtained from a value of the distribution function on the boundary speed. Therefore, it is possible to provide a motor control device having more speed continuity and a smaller speed deviation for all speeds.

  According to the eighth aspect, the position changes according to the upper command speed and the distribution function spanning a plurality of speed ranges is used, and each position-side boundary speed output is obtained from the value of the distribution function on the boundary speed. As a result, it is possible to provide a motor control device that has more continuity of speed and has less speed deviation and position deviation for all speeds.

FIG. 2 is a diagram illustrating an example of an external appearance of a robot to which the motor control device according to the first embodiment is applied. FIG. 3 is a diagram illustrating an example of an overall configuration of a motor control device according to the first and third embodiments. FIG. 4 is a conceptual diagram illustrating an internal configuration of a speed feedforward control unit in FIG. 2 according to the first embodiment. FIG. 9 is a diagram illustrating an example of (synthetic) friction characteristics including a static friction characteristic and a viscous friction characteristic (non-linear characteristics as a whole) and a method of setting a boundary speed. FIG. 11 is a diagram illustrating an example of a firing state from a speed-side positive / negative speed firing portion and a speed-side boundary speed firing portion in an example where upper command speed Vv> 0 and boundary speed V11 <higher command speed Vv <boundary speed V12. . FIG. 11 is a diagram illustrating an example of a firing state from a speed-side positive / negative speed firing unit and a speed-side boundary speed firing unit in an example where upper command speed Vv> 0 and boundary speed V12 <upper command speed Vv <boundary speed V13. . FIG. 4 is a diagram for explaining an example of a firing state from a speed-side positive / negative speed firing portion and a speed-side boundary speed firing portion in an example where upper command speed Vv <0 and boundary speed −V02 <upper command speed Vv <boundary speed−V01. It is. FIG. 4 is a diagram illustrating an example of a firing state from a speed-side positive / negative speed firing unit and a speed-side boundary speed firing unit in an example where upper command speed Vv <0 and boundary speed −V03 <upper command speed Vv <boundary speed−V02. It is. FIG. 3 is a conceptual diagram illustrating an internal configuration of a position feedforward control unit in FIG. 2 according to the first embodiment. FIG. 9 is a diagram illustrating an example of a firing state from a position-side positive / negative speed firing unit and a position-side boundary speed firing unit in an example where upper command speed Vp> 0 and boundary speed V11 <upper command speed Vp <boundary speed V12. . FIG. 4 is a diagram for explaining an example of a firing state from a position-side positive / negative speed firing unit and a position-side boundary speed firing unit in an example where upper command speed Vp <0 and boundary speed −V02 <upper command speed Vp <boundary speed−V01. It is. FIG. 9 is a diagram illustrating an example of a position deviation generated by conventional control. FIG. 7 is a diagram illustrating an example of a position deviation generated by the control of the first embodiment (effect of the present application over the related art). FIG. 9 is a diagram illustrating an example of an overall configuration of a motor control device according to second and fourth embodiments. FIG. 13 is a conceptual diagram illustrating an internal configuration of a speed feedforward control unit in FIG. 2 according to a third embodiment. FIG. 13 is a conceptual diagram illustrating an internal configuration of a position feedforward control unit in FIG. 2 according to a third embodiment. FIG. 15 is a conceptual diagram illustrating an internal configuration of a speed feedforward control unit in FIG. 14 according to the fifth, sixth, and seventh embodiments. In the fifth embodiment, an example of the ignition state of the speed-side boundary speed input unit H [m] in the case where the upper command speed Vv> 0 and the boundary speed V12 <the higher command speed Vv <the boundary speed V13 will be described. FIG. In the fifth embodiment, in a case where the upper command speed Vv <0 and the boundary speed −V03 <the higher command speed Vv <the boundary speed −V02, an example of an ignition state of the speed-side boundary speed input unit H [m]. FIG. In the sixth embodiment, an example of the ignition state of the speed-side boundary speed input unit H [m] in the case where the upper command speed Vv> 0 and the boundary speed V12 <the higher command speed Vv <the boundary speed V13 will be described. FIG. In the sixth embodiment, in the case where the upper command speed Vv <0 and the boundary speed −V03 <the higher command speed Vv <the boundary speed −V02, an example of the firing state of the speed-side boundary speed input unit H [m] FIG. In the seventh embodiment, an example of the ignition state of the speed-side boundary speed input unit H [m] in the case of the upper command speed Vv> 0 and the boundary speed V12 <the higher command speed Vv <the boundary speed V13 will be described. FIG. In the seventh embodiment, in a case where the upper command speed Vv <0 and the boundary speed −V03 <the higher command speed Vv <the boundary speed −V02, an example of a firing state of the speed-side boundary speed input unit H [m]. FIG.

  Hereinafter, embodiments of the present invention will be sequentially described with reference to the drawings. In the figures in which the X, Y, and Z axes are described, the X, Y, and Z axes are orthogonal to each other.

●● [First Embodiment (FIGS. 1 to 11)]
A motor control device 92U according to the first embodiment will be described using a control device for controlling the robot shown in FIG. 1 as an example. The robot shown in FIG. 1 includes a base 91, a first turning section 92, a first swing section 93, a second swing section 94, a second swing section 95, and the like.

  The base 91 is provided with an electric motor 91M having an encoder 91E (which is an angle detecting means and can also be used as a position detecting means). The electric motor 91M turns the first turning portion 92 with respect to the base 91 based on the drive current from the motor control device 91U. Further, the motor control device 91U detects the turning angle θ1 of the first turning portion 92 based on the detection signal from the encoder 91E.

  The first turning section 92 is provided with an electric motor 92M having an encoder 92E (which is an angle detecting means and can also be used as a position detecting means). The electric motor 92M swings the first swing unit 93 with respect to the first turning unit 92 based on the drive current from the motor control device 92U. Further, the motor control device 92U detects the swing angle θ2 of the first swing unit 93 based on the detection signal from the encoder 92E.

  The first swing unit 93 is provided with an electric motor 93M having an encoder 93E (which is an angle detecting means and can also be used as a position detecting means). The electric motor 93 </ b> M causes the first swing unit 93 to swing the second swing unit 94 based on the drive current from the motor control device 93 </ b> U. Further, motor control device 93U detects a turning angle θ3 of second turning portion 94 based on a detection signal from encoder 93E.

  The second turning section 94 is provided with an electric motor 94M having an encoder 94E (which is an angle detecting means and can also be used as a position detecting means). The electric motor 94M swings the second swing unit 95 with respect to the second turning unit 94 based on the drive current from the motor control device 94U. Further, the motor control device 94U detects the swing angle θ4 of the second swing portion 95 based on the detection signal from the encoder 94E.

  Then, the command position around θ1 is transmitted from the robot control device 60 to the motor control device 91U, the command position around θ2 is transmitted to the motor control device 92U, and the command position around θ3 is transmitted to the motor control device 93U. The command position around θ4 is transmitted to the motor control device 94U.

  Hereinafter, a motor control device 92U that controls the electric motor 92M will be described using the electric motor 92M as an example with reference to FIGS. Note that the motor controllers 91U, 93U, and 94U that control the electric motors 91M, 93M, and 94M, respectively, are similar, and description thereof is omitted.

● [Configuration of motor control device 92U for controlling electric motor 92M (FIG. 2)]
As shown in FIG. 2, the motor control device 92U that controls the electric motor 92M includes a first control unit 41 that receives a command position 10in and an actual position 27out and outputs a lower command speed 15out, a lower command speed 15out, and a command. The second control unit 42 receives the position 10in and the actual position 27out and outputs a command current 25out, and has a current output unit 43 that receives the command current 25out and outputs a drive current 31out for driving the electric motor 92M. are doing. In this case, the command position is the same as the command angle, and the current position (angle) is detected based on a detection signal from the encoder 92E.

● [Configuration of First Control Unit 41 (FIG. 2)]
The first control unit 41 includes a position deviation calculation unit 10, a position feedback control unit 11, a position-side input speed calculation unit 12, a position-side input acceleration calculation unit 13, a position feedforward control unit 14, a speed addition calculation unit 15, and the like. are doing.

  The position deviation calculator 10 calculates the command position 10in (in this case, the command (rotation) angle) for the electric motor 92M and the actual position of the position (in this case, the rotation angle of the output shaft of the electric motor 92M) for the electric motor 92M. (The actual position based on the detection signal from the encoder 92E) is input, and a position deviation 10out which is a deviation between the command position 10in and the actual position 27out is calculated and output. Note that the actual position 27out is obtained from a detection signal from the encoder 92E. If a speed reduction mechanism with a reduction ratio a exists between the electric motor 92M and the first swinging portion 93, the value multiplied by a to the command position of the first swinging portion 93, which is the control target, is: It becomes the command position 10in of the electric motor 92M.

  The position feedback control unit 11 receives the position deviation 10out, performs feedback control according to the input position deviation 10out, and outputs the first provisional command speed 11out. The position feedback control unit 11 has at least one of a proportional term (P), an integral term (I), and a differential term (D), which is so-called PID control, and outputs a first provisional command speed 11out. Since the PID control is the same as the existing control, a detailed description is omitted.

  The position-side input speed calculator 12 receives the commanded position 10in (as the input position (in this case, the rotation angle)) and calculates the speed (in this case, the rotation angular speed) based on the time-dependent change of the commanded position 10in. An input speed of 12 out (corresponding to the position-side speed) is output.

  The position-side input acceleration calculation unit 13 receives the input speed 12out and calculates an input acceleration 13out (corresponding to the position-side acceleration) which is an acceleration (in this case, a rotational angular acceleration) based on a change of the input speed 12out with time. Output.

  The position feedforward control unit 14 receives the input speed 12out (position-side speed), the input acceleration 13out (position-side acceleration), and the position deviation 10out, and performs feedforward control according to these inputs to perform the second provisional command. Output speed 14out. The position feed forward control unit 14 is configured by a network with a weight learning function, and will be described later in detail.

  The speed addition calculation unit 15 receives the first provisional command speed 11out and the second provisional command speed 14out, and adds the first provisional command speed 11out and the second provisional command speed 14out to the lower command speed 15out (in this case, command turning). Angular velocity).

● [Configuration of Second Control Unit 42 (FIG. 2)]
The second control unit 42 includes a speed deviation calculation unit 20, a speed feedback control unit 21, a speed-side input speed calculation unit 22, a speed-side input acceleration calculation unit 23, a speed feedforward control unit 24, a current addition calculation unit 25, and an actual speed. It has an operation unit 28 and the like.

  The speed deviation calculator 20 calculates the lower order command speed 15out, the actual speed 28out (actual speed based on the detection signal from the encoder 92E), which is the actual speed of the output shaft of the electric motor 92M (in this case, the rotational angular speed). , And calculates and outputs a speed deviation 20out which is a deviation between the lower command speed 15out and the actual speed 28out. The actual speed 28out is obtained by the actual speed calculator 28 from the time-dependent change of the actual position 27out (in this case, the actual rotation angle) according to the detection signal from the encoder 92E. The actual speed calculator 28 receives the actual position 27out and outputs an actual speed 28out.

  The speed feedback control unit 21 receives the speed deviation 20out, performs feedback control according to the input speed deviation 20out, and outputs a first provisional command current 21out. The speed feedback control unit 21 has at least one of a proportional term (P), an integral term (I), and a differential term (D), which is so-called PID control, and outputs a first provisional command current 21out. Since the PID control is the same as the existing control, a detailed description is omitted.

  The speed-side input speed calculator 22 receives the command position 10in (as the input position (in this case, the rotation angle)) and calculates the speed (the rotation angular speed in this case) based on the time-dependent change of the command position 10in. Outputs the input speed 22out (corresponding to the speed side speed).

  The speed-side input acceleration calculator 23 receives the input speed 22out, and calculates an input acceleration 23out (corresponding to the speed-side acceleration) which is an acceleration (in this case, a rotational angular acceleration) based on a change of the input speed 22out with time. Output.

  The speed feedforward control unit 24 receives an input speed 22out (speed-side speed), an input acceleration 23out (speed-side acceleration), and a speed deviation 20out, and performs feedforward control according to these inputs to perform the second provisional command. Outputs current 24out. The speed feedforward control unit 24 is configured by a network with a weight learning function, and will be described later in detail.

  The current addition calculation unit 25 receives the first provisional command current 21out and the second provisional command current 24out, and outputs a command current 25out obtained by adding the first provisional command current 21out and the second provisional command current 24out.

● [Configuration of current output unit 43 (FIG. 2)]
The current output unit 43 includes a current deviation calculation unit 30, a current feedback control unit 31, and the like.

  The current deviation calculator 30 receives the command current 25out and the driving current 31out (actual current) that is the current that is actually output, and outputs a current that is a deviation between the command current 25out and the driving current 31out (actual current). The deviation 30out is calculated and output.

  The current feedback controller 31 receives the current deviation 30out, performs feedback control according to the input current deviation 30out, and outputs a driving current 31out for driving the electric motor 91M. The current feedback control unit 31 has at least one of a proportional term (P), an integral term (I), and a differential term (D), which is so-called PID control, and outputs a drive current 31out. Since the PID control is the same as the existing control, a detailed description is omitted.

● [Configuration of Network with Weight Learning Function of Speed Feedforward Control Unit 24 (FIG. 3)]
Next, a network with a weight learning function in the speed feedforward control unit 24 will be described with reference to FIG. Since the speed feedforward control unit 24 has a greater effect on reducing the position deviation than the position feedforward control unit 14, the speed feedforward control unit 24 will be described first. The speed feedforward control unit 24 has an input processing unit 241 and a simple perceptron 242. The input processing unit 241 includes a speed-side positive / negative speed firing unit 24E and a speed-side boundary speed firing unit 24F. The simple perceptron 242 has an input layer 24A, a speed-side first weight learning unit 24G, and an output layer 24C.

  The speed (input speed 22out) input to the speed feedforward control unit 24 shown in FIG. 3 is obtained by differentiating the command position 10in from the robot controller 60 as shown in FIG. Instead of calculating from the differentiation of the position 10in, information on various positions may be obtained by differentiating, or a command speed from the robot controller 60 may be used. For this reason, hereinafter, the speed input to the speed feedforward control unit 24 shown in FIG. 3 will be referred to as the upper command speed Vv (different from the lower command speed 15out) instead of the input speed 22out. Similarly, the acceleration (input acceleration 23out) input to the speed feedforward control unit 24 shown in FIG. 3 is obtained by differentiating the command position 10in from the robot control device 60 twice as shown in FIG. May be obtained by differentiating the information on various positions twice, or by differentiating the command speed from the robot control device 60, without obtaining it from the twice differentiation of the command position 10in. For this reason, hereinafter, the acceleration input to the speed feedforward control unit 24 shown in FIG. 3 is described as the higher-order command acceleration αv instead of the input acceleration 23out.

● [About the input processing unit 241 (FIGS. 3 to 8)]
The higher-order command speed Vv is input to the speed-side positive / negative speed firing unit 24E. The speed-side positive / negative speed firing unit 24E ignites the “Vv (> 0)” side and ignites the “Vv (<0)” side when the input upper command speed Vv is positive (> 0). Instead, it outputs the speed-side positive speed output value (the value is the value of the upper command speed Vv) to the speed-side positive speed input unit K [2]. The speed-side positive / negative speed firing unit 24E fires on the side of “Vv (<0)” and fires on the side of “Vv (> 0)” when the input upper command speed Vv is negative (<0). Instead, it outputs the speed-side negative speed output value (the value is the value of the higher-order command speed Vv) to the speed-side negative speed input unit K [3].

  The higher-order command speed Vv is input to the speed-side boundary speed firing unit 24F. The speed-side boundary speed firing unit 24F serves as a boundary for each speed range when the limited speed range limited to the speed range of the upper command speed Vv is divided into a plurality of adjacent speed ranges set in advance. It has a plurality of boundary velocities (-V0n to V1n) as velocities (see FIG. 4). Then, the speed-side boundary speed firing section 24F outputs a speed-side boundary speed output value based on the speed difference from a boundary speed whose speed difference from the input higher-order command speed Vv is within a predetermined speed difference. The limited speed range, speed range, boundary speed, and speed-side boundary speed output value will be described with reference to FIGS.

● [Limited speed range, speed range, and boundary speed (Fig. 4)]
Next, the limited speed range, the speed range, and the boundary speed (-V0n to V1n) will be described with reference to FIG. When the first rocking portion 93 (see FIG. 1) is rocked by using the electric motor 92M, the friction of each portion (the friction inside the electric motor 92M and the friction between the first rocking portion 93 and the first turning portion 92). Due to friction, etc.). The frictional force is mainly represented by a combined friction of the static friction and the viscous friction, and the frictional force changes according to the speed.

  FIG. 4 shows speed / physical phenomenon characteristics showing the relationship between the magnitude of the upper command speed Vv for the electric motor 92M and a predetermined physical phenomenon (in this case, friction) when moving the position of the control object. The horizontal axis represents the speed (upper command speed Vv), and the vertical axis represents the frictional force. In FIG. 4, the characteristics indicated by the dashed line indicate the static friction characteristics, the characteristics indicated by the two-dot chain line indicate the viscous friction characteristics, and the characteristics indicated by the thick solid line indicate the static friction characteristics and the viscous friction characteristics. Fig. 9 shows the synthesized friction characteristics that have been synthesized.

  In the region where the speed is positive, the static friction characteristic shows a characteristic that gradually decreases as the speed increases from near zero, and the viscous friction characteristic shows a characteristic that gradually increases as the speed increases from near zero. . In the region where the speed is positive, the synthetic friction characteristic is regarded as a non-linear characteristic in the region where the speed is equal to or less than V1n (corresponding to the predetermined positive speed) and the speed is equal to or lower than V1n (and speed> 0). In a large area, it can be regarded as a linear characteristic. Therefore, in the region where the speed is positive, the synthetic friction characteristic is regarded as a non-linear characteristic as a whole.

  In the negative speed region, the static friction characteristic shows a characteristic that gradually increases as the speed decreases from near zero, and the viscous friction characteristic shows a characteristic that gradually decreases as the speed decreases from near zero. . Then, in the region where the speed is negative, the synthetic friction characteristic is regarded as a non-linear characteristic in the region where the speed is equal to or higher than −V0n (and speed <0), with the boundary of speed = −V0n (corresponding to a predetermined negative speed). In a region less than -V0n, it can be regarded as a linear characteristic. Therefore, in the region where the speed is negative, the synthetic friction characteristic is regarded as a non-linear characteristic as a whole. In addition, the example of FIG. 4 shows an example in which the slope of the combined friction characteristic in the region where the speed is less than -V0n is slightly gentler than the slope of the combined friction characteristic in the region where the speed is V1n or more. I have. In the example of FIG. 4, the curvature of the combined friction characteristics in the region where the speed is −V05 to −V01 is smaller than the curvature of the combined friction characteristics in the region where the speed is V11 to V15 (the degree of bending is small). (Loose). Therefore, as shown in the example of FIG. 4, the speed / frictional force characteristics (synthetic friction characteristics) are not point-symmetric in a region where the speed is positive (> 0) and in a region where the speed is negative (<0). Depending on the situation, the speed / frictional force characteristics (synthetic friction characteristics) may be point-symmetric between a region where the speed is positive and a region where the speed is negative. In this case, the sign of the speed is not considered. , May be treated as positive.

  Since the synthetic friction characteristic is a nonlinear characteristic as a whole, it is difficult to model it, and the position deviation cannot be appropriately reduced by the speed feedforward control using the conventional neural network. Therefore, in the present application, the composite friction characteristic of the non-linear characteristic is divided for each speed range that can be regarded as a linear characteristic, so that it is regarded as a linear characteristic for each speed range, and the position deviation can be appropriately reduced. . In other words, the composite friction characteristic of the non-linear characteristic is divided into speed ranges that can be regarded as linear characteristics, and weight learning described later is performed for each of the divided speed ranges (for each linear characteristic), so that the position can be determined at any speed. It is possible to reduce the deviation.

  Therefore, in the synthetic friction characteristics shown in FIG. 4, in the region where the speed is positive (> 0), the limited speed range in which the speed is less than V1n and the speed is greater than 0, which is regarded as the non-linear characteristic, is partially changed to the linear characteristic. It is divided into regions for each speed range that can be considered. For example, in the example of FIG. 4, the region is divided into a region of speed (V11) to speed (V12), a region of speed (V12) to speed (V13), a region of speed (V13) to speed (V14), and the like. Within the speed range of each region, the resultant friction characteristics are linear and can be regarded as linear characteristics. The speeds at the boundaries of the speed ranges of the respective areas are set as boundary speeds V11 to V1n. The boundary speed V1n on the positive side, which is the boundary between the speed region that can be considered as a linear characteristic and the speed region that is considered as a non-linear characteristic, is defined as a predetermined positive speed V1n. Therefore, in the region where the speed is positive, the boundary speeds (V11 to V1n) on the positive side are set in the limited speed range which is a region equal to or lower than the predetermined positive speed which is regarded as a non-linear characteristic. For example, in a portion where the curvature of the synthetic friction characteristic is relatively large, the speed range is narrow (for example, the speed range of the boundary speed V13 and the boundary speed V14), and in a portion where the curvature of the synthetic friction forcing is relatively small, the speed range is wide (for example, (Velocity range of boundary speed V11 and boundary speed V12). That is, the intervals between the boundary velocities are unequal. As a result, within each speed range, the resultant friction characteristics become linear, and can be regarded as linear characteristics within each speed range (the same applies when the following speeds are negative). In the case where the speed is 0 (zero), the value of the synthetic friction characteristic cannot be specified. Therefore, the boundary speed V11 (> 0) is not 0 (zero) but is set to a value close to 0 (zero). ing.

  Similarly, in the region where the speed is negative (<0), the limited speed range in which the speed is −V1n or more and the speed is less than 0, which is regarded as the non-linear characteristic, is set for each speed range in which the linear characteristic can be partially considered. Divided into regions. For example, in the example of FIG. 4, a region of speed (−V05) to speed (V-04), a region of speed (−V04) to speed (−V03), a region of speed (−V03) to speed (−V02), and the like. Divided into Within the speed range of each region, the resultant friction characteristics are linear and can be regarded as linear characteristics. The speeds at the boundaries of the speed ranges of the respective regions are set as boundary speeds -V01 to -V0n. The boundary speed −V0n on the negative side, which is the boundary between the speed region that can be considered as a linear characteristic and the speed region that is considered as a non-linear characteristic, is defined as a predetermined negative speed −V0n. Therefore, in the region where the speed is negative, the boundary speeds (-V01 to -V0n) on the negative side are set in the limited speed range that is a region equal to or higher than the predetermined negative speed regarded as the non-linear characteristic. In the case of speed = 0 (zero), since the value of the synthetic friction characteristic cannot be specified, the boundary speed −V01 (<0) is not 0 (zero), but is set to a value close to 0 (zero). Have been. Next, the operation of the speed-side positive / negative speed firing unit 24E and the speed-side boundary speed firing unit 24F will be described with reference to FIGS.

● [Operation of the speed-side positive / negative speed firing unit 24E and the speed-side boundary speed firing unit 24F (FIGS. 5 to 8)]
FIG. 5 shows an example in the case where the upper command speed Vv> 0 and V11 <the higher command speed Vv <V12. Then, as shown in the upper example of FIG. 5, coordinates are prepared such that the horizontal axis is the speed (upper command speed, boundary speed) and the vertical axis is the speed-side boundary speed output value, and the width of the spread width Nw in the speed direction is determined. A predetermined distribution function N (for example, a normal distribution function) is set in advance. Then, the distribution function N is arranged such that the value in the horizontal axis direction of the top Nc of the distribution function N becomes the upper command speed Vv. The value of the distribution function N in the vertical axis direction of the top Nc is 1.0. 5 shows an example in which the value of the distribution function N for the boundary speed V11 is 0.8, and the value of the distribution function N for the boundary speed V12 is 0.6. Note that the boundary velocities other than the boundary velocities V11 and V12 are all 0 (zero) because they are out of the range of the distribution function N. In the upper part of FIG. 5, the boundary speed −V01 appears to be within the range of the distribution function N. However, since the example of FIG. 5 is a case where the upper command speed Vv> 0, the boundary speeds −V01 to −V0n when the upper command speed Vv <0 is set and the range of the distribution function when the upper command speed Vv> 0 is satisfied. It is regarded as outside and all are regarded as 0 (zero).

  In the case of the example of FIG. 5, since the upper command speed Vv> 0, the speed-side positive / negative speed firing unit 24E fires Vv (> 0) corresponding to the speed-side positive speed input unit K [2]. The upper command speed Vv is output. Accordingly, the upper command speed Vv is input as the speed-side positive speed input value to the speed-side positive speed input unit K [2], and the input speed-side normal speed input value is input to the speed-side positive speed input unit K [2]. The value is output as the speed-side positive speed output. In the example of FIG. 5, Vv (<0) corresponding to the speed-side negative speed input unit K [3] in the speed-side positive / negative speed firing unit 24E does not fire. Therefore, nothing is input to the speed-side negative speed input unit K [3], and nothing is output from the speed-side negative speed input unit K [3].

  In the case of the example in FIG. 5, the speed-side boundary speed firing unit 24F fires the boundary speeds V11 and V12 corresponding to the speed-side boundary speed input units K [j + 1] and K [j + 2]. That is, in the upper diagram of FIG. 5, each speed difference between each boundary speed and the upper command speed is equal to or less than the boundary speed within the predetermined speed difference (within the spread width Nw / 2). Then, a speed-side boundary speed output value based on the speed difference between the speed and the distribution function N is output. In the example of FIG. 5, the speed-side boundary speed firing section 24F outputs 0.8 from the fired boundary speed V11 and outputs 0.6 from the fired boundary speed V12. Therefore, 0.8 is input to the speed-side boundary speed input unit K [j + 1], and the speed-side boundary speed input unit K [j + 1] outputs the input speed-side boundary speed output value (0.8 in this case). Output as speed-side boundary speed output. Further, 0.6 is input to the speed-side boundary speed input unit K [j + 2], and the speed-side boundary speed input unit K [j + 2] converts the input speed-side boundary speed output value (0.6 in this case) to the speed. Output as side boundary speed output. In the example of FIG. 5, since the boundary speeds -V0n, -V01, V13, V14, V15, and V1n in the speed-side boundary speed firing unit 24F do not fire, the speed-side boundary speed input units K [4] and K [j]. , K [j + 3], K [j + 4], K [j + 5], K [j + n], nothing is input, and the speed-side boundary speed input sections K [4], K [j], K [j + 3], K [j + 4], K [j + 5] and K [j + n] do not output anything. Since the boundary speeds are arranged not at equal intervals but at irregular intervals, as shown in FIGS. 5 and 6, the number of boundary speeds to be fired changes according to the value of the upper command speed Vv.

  FIG. 6 shows an example in the case where the upper command speed Vv> 0 and V12 <the upper command speed Vv <V13. Then, as shown in the example in the upper part of FIG. 6, coordinates are prepared such that the horizontal axis is the speed (upper command speed, boundary speed) and the vertical axis is the speed-side boundary speed output value (similar to FIG. 5). A predetermined distribution function N (for example, a normal distribution function) having a width in the speed direction of Nw is set in advance. Then, the distribution function N is arranged such that the value in the horizontal axis direction of the top Nc of the distribution function N becomes the upper command speed Vv. The value of the distribution function N in the vertical axis direction of the top Nc is 1.0. 6, the value of the distribution function N for the boundary speed V12 is 0.1, the value of the distribution function N for the boundary speed V13 is 0.9, and the value of the distribution function N for the boundary speed V14 is An example of 0.3 is shown. The boundary velocities other than the boundary velocities V12, V13, and V14 are all 0 (zero) because they are out of the range of the distribution function N. Since the upper command speed Vv> 0, the boundary speeds -V01 to -V0n in the case of the upper command speed Vv <0 are all 0 (zero). As shown in FIGS. 5 and 6, when the speed difference between the upper command speed Vv and the boundary speed is large, the value of the distribution function N with respect to the boundary speed becomes small. Conversely, when the speed difference between the upper command speed Vv and the boundary speed is small, the value of the distribution function N with respect to the boundary speed increases. When the value of the distribution function N for each of the adjacent boundary speeds is the same, this means that the upper command speed Vv is a value that is the center of the adjacent boundary speeds.

  In the example of FIG. 6, since the upper command speed Vv> 0, the speed-side positive / negative speed firing unit 24E fires Vv (> 0) corresponding to the speed-side positive speed input unit K [2]. The upper command speed Vv is output. Accordingly, the upper command speed Vv is input as the speed-side positive speed input value to the speed-side positive speed input unit K [2], and the input speed-side normal speed input value is input to the speed-side positive speed input unit K [2]. The value is output as the speed-side positive speed output. In the example of FIG. 6, Vv (<0) corresponding to the speed-side negative speed input unit K [3] in the speed-side positive / negative speed firing unit 24E does not fire. Therefore, nothing is input to the speed-side negative speed input unit K [3], and nothing is output from the speed-side negative speed input unit K [3].

  In the example of FIG. 6, the speed-side boundary speed firing section 24F includes boundary speeds V12, V13, and V14 corresponding to the speed-side boundary speed input sections K [j + 2], K [j + 3], and K [j + 4]. set a fire. That is, in the upper diagram of FIG. 6, each speed difference between each boundary speed and the upper command speed is equal to or less than the boundary speed within the predetermined speed difference (within the spread width Nw / 2). , That is, a speed-side boundary speed output value based on the distribution function N is output. In the example of FIG. 6, in the speed-side boundary speed firing section 24F, 0.1 is output from the lit boundary speed V12, 0.9 is output from the lit boundary speed V13, and 0.3 is output from the lit boundary speed V14. Is output. Accordingly, 0.1 is input to the speed-side boundary speed input unit K [j + 2], and the speed-side boundary speed input unit K [j + 2] receives the input speed-side boundary speed output value (0.1 in this case). Output as speed-side boundary speed output. Further, 0.9 is input to the speed-side boundary speed input unit K [j + 3], and the speed-side boundary speed input unit K [j + 3] converts the input speed-side boundary speed output value (in this case, 0.9) to the speed. Output as side boundary speed output. Further, 0.3 is input to the speed-side boundary speed input unit K [j + 4], and the speed-side boundary speed input unit K [j + 4] outputs the input speed-side boundary speed output value (0.3 in this case) to the speed. Output as side boundary speed output. In the example of FIG. 6, since the boundary speeds -V0n, -V01, V11, V15, and V1n in the speed-side boundary speed firing unit 24F do not fire, the speed-side boundary speed input units K [4], K [j], K Nothing is input to [j + 1], K [j + 5] and K [j + n], and the speed-side boundary speed input sections K [4], K [j], K [j + 1], K [j + 5], K [ j + n] does not output anything.

  FIG. 7 shows an example in the case where the upper command speed Vv <0 and -V02 <the upper command speed Vv <-V01. As in the upper part of FIG. 5, as shown in the example of the upper part of FIG. 5, coordinates are prepared such that the horizontal axis is the speed (upper command speed) and the vertical axis is the speed-side boundary speed output value. A predetermined distribution function N (for example, a normal distribution function) having a width in the direction is set in advance. However, since the upper command speed Vv <0, the distribution function N is arranged so as to be convex downward. Then, the distribution function N is arranged such that the value in the horizontal axis direction of the top Nc of the distribution function N becomes the upper command speed Vv. The value of the distribution function N in the vertical axis direction of the top Nc is -1.0. 7 shows an example in which the value of the distribution function N for the boundary speed -V02 is -0.7, and the value of the distribution function N for the boundary speed -V01 is -0.5. The boundary velocities other than the boundary velocities -V02 and -V01 are all 0 (zero) because they are out of the range of the distribution function N. In the upper part of FIG. 7, the boundary velocity V11 appears to be within the range of the distribution function N. However, since the example of FIG. 7 is the case of the upper command speed Vv <0, the boundary speeds V11 to V1n in the case of the upper command speed Vv> 0 are set outside the range of the distribution function in the case of the upper command speed Vv <0. All, 0 (zero).

  In the case of the example of FIG. 7, since the upper command speed Vv <0, the speed-side positive / negative speed firing unit 24E fires Vv (<0) corresponding to the speed-side negative speed input unit K [3]. The upper command speed Vv is output. Therefore, the higher-order command speed Vv is input to the speed-side negative speed input unit K [3] as the speed-side negative speed input value, and the speed-side negative speed input unit K [3] receives the input speed-side negative speed input. The value is output as the speed side negative speed output. In the example of FIG. 7, Vv (> 0) corresponding to the speed-side positive speed input unit K [2] in the speed-side positive / negative speed firing unit 24E does not fire. Therefore, nothing is input to the speed-side positive speed input unit K [2], and nothing is output from the speed-side normal speed input unit K [2].

  In the case of the example in FIG. 7, the speed-side boundary speed firing unit 24F fires the boundary speeds -V02 and -V01 corresponding to the speed-side boundary speed input units K [j-1] and K [j]. That is, in the upper diagram of FIG. 7, each speed difference between each boundary speed and the upper command speed is equal to or less than the boundary speed within the predetermined speed difference (within the spread width Nw / 2). , That is, a speed-side boundary speed output value based on the distribution function N is output. In the example of FIG. 7, the speed-side boundary speed firing section 24F outputs -0.7 from the lit boundary speed -V02 and -0.5 from the lit boundary speed -V01. Accordingly, -0.7 is input to the speed-side boundary speed input unit K [j-1], and the speed-side boundary speed input unit K [j-1] receives the input speed-side boundary speed output value (in this case, -0.7) is output as the speed-side boundary speed output. The speed-side boundary speed input unit K [j] receives -0.5, and the speed-side boundary speed input unit K [j] receives the input speed-side boundary speed output value (-0.5 in this case). Is output as the speed-side boundary speed output. In the example of FIG. 7, since the boundary speeds -V0n, -V05, -V04, -V03, V11, and V1n in the speed-side boundary speed firing unit 24F do not fire, the speed-side boundary speed input units K [4] and K [ Nothing is input to j-4], K [j-3], K [j-2], K [j + 1], K [j + n], and the speed-side boundary speed input units K [4], K [ j-4], K [j-3], K [j-2], K [j + 1], K [j + n] do not output anything. Since the boundary speeds are arranged at irregular intervals instead of at equal intervals, as shown in FIGS. 7 and 8, the number of boundary speeds to be fired changes according to the value of the upper command speed Vv.

  FIG. 8 shows an example in the case where the upper command speed Vv <0 and -V03 <the upper command speed Vv <-V02. Then, as shown in the upper example of FIG. 8, coordinates are prepared such that the horizontal axis is the speed (upper command speed, boundary speed) and the vertical axis is the speed-side boundary speed output value (similar to FIG. 7). A predetermined distribution function N (for example, a normal distribution function) having a width in the speed direction of Nw is set in advance. However, since the upper command speed Vv <0, the distribution function N is arranged so as to be convex downward. Then, the distribution function N is arranged such that the value in the horizontal axis direction of the top Nc of the distribution function N becomes the upper command speed Vv. The value of the distribution function N in the vertical axis direction of the top Nc is -1.0. In the example shown in the upper part of FIG. 8, the value of the distribution function N for the boundary speed -V04 is -0.1, the value of the distribution function N for the boundary speed -V03 is -0.7, and the distribution for the boundary speed -V02. An example in which the value of the function N is -0.3 is shown. The boundary velocities other than the boundary velocities -V04, -V03, and -V02 are all 0 (zero) because they are out of the range of the distribution function N. Since the upper command speed Vv <0, the boundary speeds V11 to V1n when the upper command speed Vv> 0 are all 0 (zero).

  In the example of FIG. 8, since the upper command speed Vv <0, the speed-side positive / negative speed firing unit 24E fires Vv (<0) corresponding to the speed-side negative speed input unit K [3]. The upper command speed Vv is output. Therefore, the higher-order command speed Vv is input to the speed-side negative speed input unit K [3] as the speed-side negative speed input value, and the speed-side negative speed input unit K [3] receives the input speed-side negative speed input. The value is output as the speed side negative speed output. In the example of FIG. 8, Vv (> 0) corresponding to the speed-side positive speed input unit K [2] in the speed-side positive / negative speed firing unit 24E does not fire. Therefore, nothing is input to the speed-side positive speed input unit K [2], and nothing is output from the speed-side normal speed input unit K [2].

  In the case of the example in FIG. 8, the speed-side boundary speed firing unit 24F is a boundary speed corresponding to the speed-side boundary speed input units K [j-3], K [j-2], and K [j-1]. -V04, -V03, and -V02 fire. That is, in the upper diagram of FIG. 8, each speed difference between each boundary speed and the upper command speed is equal to or less than the boundary speed within the predetermined speed difference (within the spread width Nw / 2). , That is, a speed-side boundary speed output value based on the distribution function N is output. In the example of FIG. 8, in the speed-side boundary speed firing section 24F, −0.1 is output from the lit boundary speed −V04, −0.7 is output from the lit boundary speed −V03, and the lit boundary speed − -0.3 is output from V02. Therefore, -0.1 is input to the speed-side boundary speed input unit K [j-3], and the speed-side boundary speed input unit K [j-3] receives the input speed-side boundary speed output value (in this case, -0.1) is output as the speed-side boundary speed output. Further, -0.7 is input to the speed-side boundary speed input unit K [j-2], and the input speed-side boundary speed output value (in this case,- 0.7) is output as the speed-side boundary speed output. The speed-side boundary speed input unit K [j-1] receives -0.3, and the speed-side boundary speed input unit K [j-1] outputs the input speed-side boundary speed output value (in this case,- 0.3) is output as the speed-side boundary speed output. In the example of FIG. 8, since the boundary speeds -V0n, -V05, -V01, V11, and V1n in the speed-side boundary speed firing unit 24F do not fire, the speed-side boundary speed input units K [4] and K [j-4]. , K [j], K [j + 1], K [j + n], nothing is input, and the speed-side boundary speed input sections K [4], K [j-4], K [j], K [ j + 1] and K [j + n] do not output anything.

● [Input layer 24A in simple perceptron 242 (FIG. 3)]
As shown in FIG. 3, the input layer 24A in the simple perceptron 242 of the network with the weight learning function includes a speed-side acceleration input unit K [1], a speed-side positive speed input unit K [2], and a speed-side negative speed input unit K. [3], and a speed-side boundary speed input unit K [4] to K [j + n].

  The speed-side acceleration input unit K [1] receives the higher-order command acceleration αv, and outputs the input higher-order command acceleration αv as a speed-side acceleration output.

  As described above, when Vv (> 0) of the speed-side positive / negative speed firing unit 24E is fired, the speed-side positive speed input value Kv is input to the speed-side positive speed output unit K [2]. Then, the input speed-side normal speed output value is output as the speed-side normal speed output. Further, as described above, when Vv (<0) of the speed-side positive / negative speed firing unit 24E is fired, the speed-side negative speed output value that is the upper command speed Vv is output to the speed-side negative speed input unit K [3]. The input speed side negative speed output value is output as a speed side negative speed output.

  The speed-side boundary speed input units K [4] to K [j + n] are prepared corresponding to a plurality of boundary speeds -V0n to V1n. As described above, the speed-side boundary speed input units K [4] to K [j + n] receive the speed-side boundary speed output values from the boundary speeds fired in the speed-side boundary speed firing unit 24F, and input the speeds. The side boundary speed output value is output as the speed side boundary speed output. The speed-side acceleration output, the speed-side positive speed output, the speed-side negative speed output, and the speed-side boundary speed output output from each of the input units K [1] to K [j + n] are collectively output as the speed-side first output 24L1. And

● [About the speed-side first weight learning unit 24G in the simple perceptron 242 (FIG. 3)]
The speed-side first weight learning unit 24G is associated with the speed-side first learning weight U [1] associated with the speed-side acceleration input unit K [1] and the speed-side positive speed input unit K [2]. The speed-side first learning weight U [2], the speed-side first learning weight U [3] associated with the speed-side negative speed input unit K [3], and the speed-side boundary speed input units K [4] to K [ j + n], and has a storage function of storing the speed-side first learning weights U [4] to U [j + n] associated with each of them. The speed-side first weight learning unit 24G has a changing function of changing (learning) the speed-side first learning weights U [1] to U [j + n] based on the speed deviation 20out. The speed-side first weight learning unit 24G has a multiplication function of multiplying each of the input speed-side first outputs 24L1 by each of the speed-side first learning weights U [1] to U [j + n]. ing.

  The speed-side first weight learning unit 24G adjusts the speed-side first learning weights U [1] to U [j + n] in accordance with the speed deviation 20out so as to approach an optimum value for a predetermined evaluation function. Change (learn). For example, the speed-side first weight learning unit 24G uses the evaluation function whose horizontal axis is the speed-side first learning weight and whose vertical axis is the square of the speed deviation, so that the square of the speed deviation becomes small. The value of the first learning weight is changed (learned). An evaluation function is prepared for each weight. When the higher command speed Vv shown in the example of FIG. 5 is input, the speed-side first weight learning unit 24G outputs the speed-side first learning weight U [corresponding to the output speed-side first output 24L1. 1], U [2], U [j + 2], and U [j + 3], and the speed-side first learning weights U [3] to U [j + 1] corresponding to the speed-side first outputs 24L1 not output. ], U [j + 4] to U [j + n] are not learned.

  Then, the speed-side first weight learning unit 24G multiplies each of the input speed-side first outputs 24L1 by the corresponding speed-side first learning weights U [1] to U [j + n]. 24M1 is output. For example, when the upper command speed Vv shown in the example of FIG. 5 is input, the speed-side first weight learning unit 24G outputs αv * U [1], Vv * U [2], and 0.8 * U [ j + 1] and 0.6 * U [j + 2]. Further, for example, when the higher-order command speed Vv shown in the example of FIG. 7 is input, the speed-side first weight learning unit 24G outputs αv * U [1], Vv * U [3], and −0.7 * U. [J-1] and -0.5 * U [j] are output.

● [Output layer 24C in simple perceptron 242 (FIG. 3)]
The output layer 24C has a speed-side output section Q [1]. The speed-side output unit Q [1] converts a value obtained by adding each of the following speed-side first multiplication values of (11) to (14) by a predetermined function (for example, a sigmoid function), and converts the converted value to a fourth value. Output as 2 temporary command currents 24out. The speed-side output unit Q [1] can output the second provisional command current 24out without using the sigmoid function.
(11) A first speed-side multiplication value (αv * U [1) obtained by multiplying the speed-side acceleration output (αv) output from the speed-side acceleration input unit K [1] by the first speed-side learning weight U [1]. ]).
(12) When the speed-side positive speed output (Vv) is output from the speed-side positive speed input unit K [2], the speed obtained by multiplying the speed-side positive speed output (Vv) by the speed-side first learning weight U [2]. Side first multiplication value (Vv * U [2]).
(13) When the speed-side negative speed output is output from the speed-side negative speed input unit K [3], the speed obtained by multiplying the speed-side negative speed output (Vv) by the speed-side first learning weight U [3]. Side first multiplication value (Vv * U [3]).
(14) Speed-side first learning weights (U [4] to U [j + n]) corresponding to the speed-side boundary speed outputs output from the speed-side boundary speed input units K [4] to K [j + n]. Is the speed-side first multiplied value multiplied by.

● [Configuration of network with weight learning function of position feedforward control unit 14 (FIG. 9)]
Next, a network with a weight learning function in the position feedforward control unit 14 will be described with reference to FIG. Like the velocity feedforward control unit 24, the position feedforward control unit 14 includes an input processing unit 141 and a simple perceptron 142. The input processing unit 141 has a position-side positive / negative speed firing unit 14E and a position-side boundary speed firing unit 14F. The simple perceptron 142 has an input layer 14A, a position-side first weight learning unit 14G, and an output layer 14C.

  Note that the speed (input speed 12out) input to the position feedforward control unit 14 shown in FIG. 9 is obtained by differentiating the command position 10in from the robot controller 60 as shown in FIG. Instead of calculating from the differentiation of the position 10in, information on various positions may be obtained by differentiating, or a command speed from the robot controller 60 may be used. Therefore, hereinafter, the speed input to the position feedforward control unit 14 shown in FIG. 9 will be described as the upper command speed Vp (different from the lower command speed 15out) instead of the input speed 12out. Similarly, the acceleration (input acceleration 13out) input to the position feedforward control unit 14 shown in FIG. 9 is obtained by differentiating the command position 10in from the robot control device 60 twice as shown in FIG. May be obtained by differentiating the information on various positions twice, or by differentiating the command speed from the robot control device 60, without obtaining it from the twice differentiation of the command position 10in. For this reason, hereinafter, the acceleration input to the position feedforward control unit 14 shown in FIG. 9 is described as the upper command acceleration αp instead of the input acceleration 13out.

● [Input processing unit 141 (FIGS. 9 to 11)]
The upper command speed Vp is input to the position-side positive / negative speed firing unit 14E. The position-side positive / negative speed firing unit 14E fires on the side of “Vp (> 0)” and fires on the side of “Vp (<0)” when the input upper command speed Vp is positive (> 0). Instead, it outputs the position-side positive speed output value (the value is the value of the upper command speed Vp) to the position-side normal speed input unit J [2]. The position-side positive / negative speed firing unit 14E fires on the side of “Vp (<0)” and fires on the side of “Vp (> 0)” when the input upper command speed Vp is negative (<0). Instead, it outputs the position-side negative speed output value (the value is the value of the higher-order command speed Vp) to the position-side negative speed input unit J [3].

  The upper command speed Vp is input to the position-side boundary speed firing section 14F. The position-side boundary speed firing section 14F serves as a boundary for each speed range when the limited speed range limited to the speed range of the higher-order command speed Vp is divided into a plurality of preset adjacent speed ranges. It has a plurality of boundary velocities (-V0n to V1n) as velocities (see FIG. 4). The position-side boundary speed firing unit 14F outputs a position-side boundary speed output value based on the speed difference from the boundary speed at which the speed difference from the input higher-order command speed Vp is within a predetermined speed difference. The limited speed range, the speed range, the boundary speed, and the speed-side boundary speed output value are the same as those already described with reference to FIGS.

  In the example of FIG. 10, the upper command speed Vv of the example of FIG. 5 is replaced with the upper command speed Vp, the speed-side positive / negative speed firing unit 24E is replaced with the position-side positive / negative speed firing unit 14E, and the speed-side boundary speed firing unit 24F is positioned. It is replaced with the side boundary speed firing section 14F. In the example of FIG. 10, since the upper command speed Vp> 0, the position-side positive / negative speed firing unit 14E fires the Vp (> 0) corresponding to the position-side positive speed input unit J [2], and the higher position. The command speed Vp is output. Therefore, the upper command speed Vp is input to the position-side positive speed input unit J [2] as the position-side positive speed input value, and the position-side normal speed input unit J [2] receives the input position-side normal speed input. The value is output as the position-side positive speed output. In the example of FIG. 10, Vp (<0) corresponding to the position-side negative speed input unit J [3] in the position-side positive / negative speed firing unit 14E does not fire. Therefore, nothing is input to the position side negative speed input section J [3], and nothing is output from the position side negative speed input section J [3].

  In the example of FIG. 10, the position-side boundary speed firing unit 14F fires boundary speeds V12 and V13 corresponding to the position-side boundary speed input units J [j + 1] and J [j + 2]. That is, in the upper diagram of FIG. 10, each speed difference between each boundary speed and the upper command speed is equal to or less than the boundary speed within the predetermined speed difference (within the spread width Nw / 2). , That is, a position-side boundary speed output value based on the distribution function N is output. In the example of FIG. 10, the position-side boundary speed firing section 14F outputs 0.8 from the fired boundary speed V11 and outputs 0.6 from the fired boundary speed V12. Therefore, 0.8 is input to the position-side boundary speed input unit J [j + 1], and the position-side boundary speed input unit J [j + 1] outputs the input position-side boundary speed output value (0.8 in this case). Output as position-side boundary speed output. Further, 0.6 is input to the position-side boundary speed input unit J [j + 2], and the position-side boundary speed input unit J [j + 2] sets the input position-side boundary speed output value (0.6 in this case) to the position. Output as side boundary speed output. In the example of FIG. 10, since the boundary velocities -V0n, -V01, V13, V14, V15, and V1n in the position-side boundary speed firing unit 14F do not fire, the position-side boundary speed input units J [4] and J [j]. , J [j + 3], J [j + 4], J [j + 5], J [j + n], nothing is input, and the position-side boundary speed input units J [4], J [j], J [j + 3], J [j + 4], J [j + 5], and J [j + n] do not output anything. Since the boundary speeds are arranged at irregular intervals, not at equal intervals, the number of firing boundary speeds changes according to the value of the upper command speed Vp.

  In the example of FIG. 11, the upper command speed Vv of the example of FIG. 7 is replaced with the upper command speed Vp, the speed-side positive / negative speed firing unit 24E is replaced with the position-side positive / negative speed firing unit 14E, and the speed-side boundary speed firing unit 24F is positioned. It is replaced with the side boundary speed firing section 14F. In the example of FIG. 11, since the upper command speed Vp <0, the position-side positive / negative speed firing unit 14E fires the Vp (<0) corresponding to the position-side negative speed input unit J [3], and the upper position. The command speed Vp is output. Therefore, the upper command speed Vp is input to the position-side negative speed input unit J [3] as the position-side negative speed input value, and the position-side negative speed input unit J [3] receives the input position-side negative speed input. The value is output as the position side negative speed output. In the example of FIG. 11, Vp (> 0) corresponding to the position-side positive speed input unit J [2] in the position-side positive / negative speed firing unit 14E does not fire. Therefore, nothing is input to the position-side positive speed input unit J [2], and nothing is output from the position-side normal speed input unit J [2].

  In the example of FIG. 11, the position-side boundary speed firing unit 14F fires the boundary velocities -V02 and -V01 corresponding to the position-side boundary speed input units J [j-1] and J [j]. That is, in the upper diagram of FIG. 11, each speed difference between each boundary speed and the upper command speed is equal to or less than the boundary speed within the predetermined speed difference (within the spread width Nw / 2). , That is, a position-side boundary speed output value based on the distribution function N is output. In the example of FIG. 11, the position-side boundary speed firing section 14F outputs -0.7 from the fired boundary speed -V02 and -0.5 from the fired boundary speed -V01. Therefore, -0.7 is input to the position-side boundary speed input unit J [j-1], and the position-side boundary speed input unit J [j-1] outputs the input position-side boundary speed output value (in this case, -0.7) is output as the position-side boundary speed output. The position-side boundary speed input unit J [j] receives -0.5, and the position-side boundary speed input unit J [j] receives the input position-side boundary speed output value (-0.5 in this case). Is output as the position-side boundary speed output. In the example of FIG. 11, the boundary velocities -V0n, -V05, -V04, -V03, V11, and V1n in the position-side boundary speed firing unit 14F do not fire, so the position-side boundary speed input units J [4] and J [ Nothing is input to j-4], J [j-3], J [j-2], J [j + 1], J [j + n], and the position-side boundary speed input units J [4], J [ j-4], J [j-3], J [j-2], J [j + 1], and J [j + n] do not output anything. Since the boundary speeds are arranged at irregular intervals, not at equal intervals, the number of firing boundary speeds changes according to the value of the upper command speed Vp.

● [Input layer 14A in simple perceptron 142 (FIG. 9)]
As shown in FIG. 9, the input layer 14A in the simple perceptron 142 of the network with the weight learning function includes a position-side acceleration input unit J [1], a position-side positive speed input unit J [2], and a position-side negative speed input unit J. [3], and a position-side boundary speed input unit J [4] to J [j + n].

  The position-side acceleration input unit J [1] receives the higher-order command acceleration αp, and outputs the input higher-order command acceleration αp as a position-side acceleration output.

  As described above, when Vp (> 0) of the position-side positive / negative speed firing unit 14E fires, the position-side positive speed output value Jp, which is the upper command speed Vp, is input to the position-side positive speed input unit J [2]. Then, the input position-side normal speed output value is output as the position-side normal speed output. Further, as described above, when Vp (<0) of the position-side positive / negative speed firing unit 14E is fired, the position-side negative speed output value that is the higher-order command speed Vp is output to the position-side negative speed input unit J [3]. The input position-side negative speed output value is output as a position-side negative speed output.

  The position side boundary speed input units J [4] to J [j + n] are prepared corresponding to a plurality of boundary speeds -V0n to V1n. As described above, the position-side boundary speed input units J [4] to J [j + n] input the position-side boundary speed output values from the boundary speeds fired in the position-side boundary speed firing unit 14F, and input the input position. The side boundary speed output value is output as the position side boundary speed output. The position-side acceleration output, the position-side positive speed output, the position-side negative speed output, and the position-side boundary speed output output from each of the input units J [1] to J [j + n] are collectively obtained as the position-side first output 14L1. And

● [Position-side first weight learning unit 14G in simple perceptron 142 (FIG. 9)]
The position-side first weight learning unit 14G is associated with the position-side first learning weight W [1] associated with the position-side acceleration input unit J [1] and the position-side positive speed input unit J [2]. Position-side first learning weight W [2], position-side first learning weight W [3] associated with position-side negative speed input unit J [3], position-side boundary speed input units J [4] to J [ j + n], and has a storage function of storing the position-side first learning weights W [4] to W [j + n] associated with the respective ones of [j + n]. The position-side first weight learning unit 14G has a changing function of changing (learning) the position-side first learning weights W [1] to W [j + n] based on the position deviation 10out. The position-side first weight learning unit 14G has a multiplication function of multiplying each of the input position-side first outputs 14L1 by each of the position-side first learning weights W [1] to W [j + n]. ing.

  The position-side first weight learning unit 14G adjusts the position-side first learning weights W [1] to W [j + n] in accordance with the position deviation 10out so as to approach an optimum value for a predetermined evaluation function. Change (learn). For example, the position-side first weight learning unit 14G uses an evaluation function in which the horizontal axis represents the position-side first learning weight and the vertical axis represents the square of the position deviation, and reduces the position-side square to reduce the square of the position deviation. The value of the first learning weight is changed (learned). An evaluation function is prepared for each weight. When the upper command speed Vp shown in the example of FIG. 10 is input, the position-side first weight learning unit 14G outputs the position-side first learning weight W [corresponding to the output position-side first output 14L1. 1], W [2], W [j + 2], W [j + 3], and the position-side first learning weights W [3] to W [j + 1] corresponding to the position-side first outputs 14L1 that have not been output. ] And W [j + 4] to W [j + n] are not learned.

  Then, the position-side first weight learning unit 14G multiplies each of the input position-side first outputs 14L1 by a corresponding position-side first learning weight W [1] to W [j + n]. 14M1 is output. For example, when the upper command speed Vp shown in the example of FIG. 10 is input, the position-side first weight learning unit 14G outputs αp * W [1], Vp * W [2], and 0.8 * W [ j + 1] and 0.6 * W [j + 2]. Further, for example, when the higher-order command speed Vp shown in the example of FIG. 11 is input, the position-side first weight learning unit 14G outputs αp * W [1], Vp * W [3], and −0.7 * W. [J-1] and -0.5 * W [j] are output.

● [Output layer 14C in simple perceptron 242 (FIG. 9)]
The output layer 14C has a position-side output unit P [1]. The position-side output unit P [1] converts a value obtained by adding the following position-side first multiplication values of (11) to (14) by a predetermined function (for example, a sigmoid function), and converts the converted value to a first value. Output as 2 temporary command speeds 14out. The position-side output unit P [1] can output the second provisional command speed 14out without using the sigmoid function.
(11) A position-side first multiplication value (αp * W [1) obtained by multiplying the position-side acceleration output (αp) output from the position-side acceleration input unit J [1] by the position-side first learning weight W [1]. ]).
(12) The position obtained by multiplying the position-side positive speed output (Vp) by the position-side first learning weight W [2] when the position-side normal speed output is output from the position-side normal speed input unit J [2]. Side first multiplication value (Vp * W [2]).
(13) When the position-side negative speed output is output from the position-side negative speed input unit J [3], the position obtained by multiplying the position-side negative speed output (Vp) by the position-side first learning weight W [3]. Side first multiplication value (Vp * W [3]).
(14) Position-side first learning weights (W [4] to W [j + n]) corresponding to the position-side boundary speed outputs output from the position-side boundary speed input units J [4] to J [j + n]. Is the position-side first multiplied value.

  As described above, in the first embodiment, a characteristic having nonlinear characteristics as a whole is divided into a plurality of regions (in this case, regions for each speed range) which can be regarded as linear characteristics, and the input physical quantity (In this case, the position deviation can be further reduced by performing the calculation using the area corresponding to the upper command speed Vv and the upper command speed Vp). FIG. 12 shows a state of occurrence of a position deviation (error) in the conventional control in which the above-described region division is not performed and when the weights (the speed-side first learning weight and the position-side first learning weight) are not learned. An example is shown. FIG. 13 shows the occurrence of a position deviation when the control of the first embodiment (the above-described region division is performed and the speed-side first learning weight and the position-side first learning weight are performed) is performed. 9 shows an example of a state. In each case, the horizontal axis represents time, and the vertical axis represents the deviation (error) between the encoder and the command position. As is clear from a comparison between FIG. 12 and FIG. 13, the control of the first embodiment can greatly reduce the position deviation (error) as compared with the conventional control.

●● [Second embodiment (FIG. 14)]
Next, a motor control device 92U according to a second embodiment will be described with reference to FIG. The motor control device according to the second embodiment shown in FIG. 14 is different from the motor control device according to the first embodiment shown in FIG. 2 in that the position feedforward control unit 14, the position-side input speed calculation unit 12, the position The difference is that the side input acceleration calculation unit 13 and the speed addition calculation unit 15 are omitted, and the others are the same. Hereinafter, differences will be mainly described. Note that the first provisional command speed 11out output from the position feedback control unit 11 is directly input to the speed deviation calculation unit 20 as the lower command speed 15out.

  In the motor control device according to the second embodiment shown in FIG. 14, the position feedforward control unit 14 and the input (position) to the position feedforward control unit 14 in the motor control device according to the first embodiment shown in FIG. The side input speed calculation unit 12, the position side input acceleration calculation unit 13) and the output (speed addition calculation unit 15) are omitted. Therefore, the control by the motor control device of the second embodiment slightly increases the position deviation compared to the control by the motor control device of the first embodiment. However, as a result of the experiment by the inventor, it was found that the reduction of the position deviation by the velocity feedforward control unit was more dominant than the reduction of the position deviation by the position feedforward control unit. For example, the robot shown in the example of FIG. With the control described above, it was found that the position deviation was sufficiently reduced even if the position feedforward control unit was omitted. That is, for example, in the case of robot control, the position deviation can be reduced to an expected level even if the position feedforward control unit (and the input / output unit to the position feedforward control unit) is omitted. In this case, the processing load on the motor control device can be reduced.

●● [Third Embodiment (FIGS. 15 and 16)]
Next, a third embodiment of the motor control device 92U will be described. In the third embodiment, the overall configuration of the motor control device is the same as that of the first embodiment shown in FIG. 2, but the internal configuration of the speed feedforward control unit 24 is that shown in FIG. The difference is that the internal configuration of the position feedforward control unit 14 is as shown in FIG.

● [Configuration of Speed Feedforward Control Unit 24 (FIG. 15)]
Next, a configuration inside the speed feedforward control unit 24 according to the third embodiment will be described with reference to FIG. Since the speed feedforward control unit 24 has a greater effect on reducing the position deviation than the position feedforward control unit 14, the speed feedforward control unit 24 will be described first. The speed feedforward control unit 24 has an input processing unit 241 and a neural network 243. Note that the input processing unit 241 is the same as the input processing unit 241 of the first embodiment (see FIG. 3), and thus the description is omitted. Further, the upper command speed Vv and the upper command acceleration αv are the same as those in the first embodiment (see FIG. 3), and thus the description is omitted. The neural network 243 has an input layer 24A, a speed-side first weight learning unit 24G, an intermediate layer 24B, a speed-side second weight learning unit 24H, and an output layer 24C. In the third embodiment, an example in which the intermediate layer 24B includes three speed-side calculation units N [1] to N [3] will be described. However, the number of speed-side calculation units is limited to three. Not something.

● [Input layer 24A in neural network 243 (FIG. 15)]
As shown in FIG. 15, the input layer 24A in the neural network 243 includes a speed-side acceleration input unit K [1], a speed-side positive speed input unit K [2], a speed-side negative speed input unit K [3], and a speed-side The point having the boundary speed input sections K [4] to K [j + n] is the same as in the first embodiment (see FIG. 3). However, the output from each input unit is different from that of the first embodiment.

  The speed-side acceleration input unit K [1] receives the higher-order command acceleration αv and sends the input higher-order command acceleration αv to each of the plurality of speed-side arithmetic units N [1] to N [3] prepared in advance. And output it as the speed-side acceleration output.

  The speed-side positive speed input unit K [2] receives and inputs the speed-side positive speed output value, which is the upper command speed Vv, when Vv (> 0) of the speed-side positive / negative speed firing unit 24E fires. The speed-side positive speed output value is output as a speed-side normal speed output to each of the plurality of speed-side arithmetic units N [1] to N [3] prepared in advance. The speed-side negative speed input unit K [3] receives and inputs the speed-side negative speed output value, which is the upper command speed Vv, when Vv (<0) of the speed-side positive / negative speed firing unit 24E fires. The speed-side negative speed output value is output as a speed-side negative speed output to each of the plurality of speed-side arithmetic units N [1] to N [3] prepared in advance.

  The speed-side boundary speed input units K [4] to K [j + n] are prepared corresponding to a plurality of boundary speeds -V0n to V1n. The speed-side boundary speed input units K [4] to K [j + n] receive speed-side boundary speed output values from the boundary speeds fired in the speed-side boundary speed firing unit 24F, and the input speed-side boundary speed output values. Is output as a speed-side boundary speed output to each of a plurality of speed-side calculation units N [1] to N [3] prepared in advance. The speed-side acceleration output, the speed-side positive speed output, the speed-side negative speed output, and the speed-side boundary speed output output from each of the input units K [1] to K [j + n] are collectively output as the speed-side first output 24L1. And

● [About the speed-side first weight learning unit 24G in the neural network 243 (FIG. 15)]
A speed-side first learning weight U [1] [1] (U [s] [1] is added to the speed-side first output output from the speed-side acceleration input unit K [1] to the speed-side operation unit N [1]. ]) Are associated. The speed-side first output that is output from the speed-side positive speed input unit K [2] to the speed-side operation unit N [1] includes a speed-side first learning weight U [2] [1] (U [s] [ 1]). A speed-side first learning weight U [3] [1] (U [s] [) is output from the speed-side negative speed input unit K [3] to the speed-side operation unit N [1]. 1]). The speed-side first outputs output from the speed-side boundary speed input units K [4] to K [j + n] to the speed-side operation unit N [1] include speed-side first learning weights U [4] [1]-. U [j + n] [1] (U [s] [1]) is associated therewith. As described above, the speed-side first learning weight U [s] [1] is associated with each speed-side first output that is output to the speed-side operation unit N [1].

  Similarly, a speed-side first learning weight U [1] [2] (U [s]) is added to the speed-side first output output from the speed-side acceleration input unit K [1] to the speed-side operation unit N [2]. ] [2]). A speed-side first learning weight U [2] [2] (U [s] [is input to the speed-side first output output from the speed-side positive speed input unit K [2] to the speed-side operation unit N [2]. 2]). The speed-side first output output from the speed-side negative speed input unit K [3] to the speed-side operation unit N [2] includes a speed-side first learning weight U [3] [2] (U [s] [ 2]). The speed-side first outputs output from the speed-side boundary speed input units K [4] to K [j + n] to the speed-side operation unit N [2] include speed-side first learning weights U [4] [2]-. U [j + n] [2] (U [s] [2]) is associated therewith. Thus, the speed-side first learning weight U [s] [2] is associated with each speed-side first output that is output to the speed-side operation unit N [2].

  Similarly, the speed-side first learning weight U [1] [3] (U [s ] [3]). The speed-side first output outputted from the speed-side positive speed input unit K [2] to the speed-side operation unit N [3] includes speed-side first learning weights U [2] [3] (U [s] [ 3]). The speed-side first output outputted from the speed-side negative speed input unit K [3] to the speed-side operation unit N [3] includes a speed-side first learning weight U [3] [3] (U [s] [ 3]). The speed-side first outputs output from the speed-side boundary speed input units K [4] to K [j + n] to the speed-side operation unit N [3] include speed-side first learning weights U [4] [3]-. U [j + n] [3] (U [s] [3]) is associated therewith. Thus, the speed-side first learning weight U [s] [3] is associated with each speed-side first output that is output to the speed-side operation unit N [3].

  The speed-side first weight learning unit 24G has a storage function of storing these speed-side first learning weights U [s] [1], U [s] [2], and U [s] [3]. ing. The speed-side first weight learning unit 24G changes the speed-side first learning weights U [s] [1], U [s] [2], and U [s] [3] based on the speed deviation 20out ( Learning). The speed-side first weight learning unit 24G has a multiplication function of multiplying each of the input speed-side first outputs 24L1 by the corresponding speed-side first learning weight.

  The speed-side first weight learning unit 24G adjusts the values of the speed-side first learning weights U [s] [1], U [s] [2], and U [s] [3] according to the speed deviation 20out. Is changed (learned) so as to approach an optimum value for a predetermined evaluation function. For example, the speed-side first weight learning unit 24G uses the evaluation function whose horizontal axis is the speed-side first learning weight and whose vertical axis is the square of the speed deviation, so that the square of the speed deviation becomes small. The value of the first learning weight is changed (learned). An evaluation function is prepared for each weight. When the higher command speed Vv shown in the example of FIG. 5 is input, the speed-side first weight learning unit 24G outputs the speed-side first learning weight U [1] corresponding to the output speed-side first output. ] [1] to U [1] [3], U [2] [1] to U [2] [3], U [j + 2] [1] to U [j + 2] [3], U [j + 3] [ 1] to U [j + 3] [3], and does not learn the speed-side first learning weights corresponding to the other speed-side first outputs that have not been output.

  Then, the speed-side first weight learning unit 24G outputs a speed-side first multiplied value 24M1 obtained by multiplying each of the output speed-side first outputs by the corresponding speed-side first learning weight. For example, when the higher-order command speed Vv shown in the example of FIG. 5 is input, the speed-side first weight learning unit 24G outputs αv * U [1] [1] to αv * U [1] [3] and Vv * U [2] [1] to Vv * U [2] [3] and 0.8 * U [j + 1] [1] to 0.8 * U [j + 1] [3] and 0.6 * U [J + 2] [1] to 0.6 * U [j + 2] [3] are output. Also, for example, when the higher-order command speed Vv shown in the example of FIG. 7 is input, the speed-side first weight learning unit 24G outputs αv * U [1] [1] to αv * U [1] [3] and Vv * U [3] [1] to Vv * U [3] [3] and -0.7 * U [j-1] [1] to -0.7 * U [j-1] [3] , -0.5 * U [j] [1] to -0.5 * U [j] [3].

● [About the intermediate layer 24B in the neural network 243 (FIG. 15)]
The middle layer 24B has speed-side calculation units N [1] to N [3]. It should be noted that the number of speed-side arithmetic units is not limited to three.

  The speed-side operation unit N [1] outputs each of the speed-side first outputs output from each of the input units K [1] to K [j + n] to itself (the speed-side operation unit N [1]). A plurality of speed-side first multiplication values obtained by multiplying a speed-side first learning weight (U [s] [1]) corresponding to each of the speed-side first outputs are added, and the output layer is output as a speed-side second output. Output to 24C.

  The speed-side operation unit N [2] outputs each of the speed-side first outputs output from each of the input units K [1] to K [j + n] to itself (the speed-side operation unit N [2]). A plurality of speed-side first multiplied values obtained by multiplying a speed-side first learning weight (U [s] [2]) corresponding to each of the speed-side first outputs are added to form an output layer as a speed-side second output. Output to 24C.

  The speed-side operation unit N [3] outputs each of the speed-side first outputs output from each of the input units K [1] to K [j + n] to itself (the speed-side operation unit N [3]). A plurality of speed-side first multiplication values obtained by multiplying a speed-side first learning weight (U [s] [3]) corresponding to each of the speed-side first outputs are added to form an output layer as a speed-side second output. Output to 24C.

● [About the speed-side second weight learning unit 24H in the neural network 243 (FIG. 15)]
The speed-side second weight learning unit 24H includes a speed-side second learning weight Y [1] associated with the speed-side arithmetic unit N [1], and a speed-side second learning weight Y [1] associated with the speed-side arithmetic unit N [2]. It has a storage function of storing the 2 learning weights Y [2] and the speed-side second learning weights Y [3] associated with the speed-side operation unit N [3]. The speed-side second weight learning unit 24H has a changing function of changing (learning) the speed-side second learning weights Y [1] to Y [3] based on the speed deviation 20out. The speed-side second weight learning unit 24H has a multiplication function of multiplying each of the input speed-side second outputs 24L2 by each of the speed-side second learning weights Y [1] to Y [3]. ing.

  The speed-side second weight learning unit 24H adjusts the speed-side second learning weights Y [1] to Y [3] in accordance with the speed deviation 20out so as to approach an optimum value for a predetermined evaluation function. Change (learn). For example, the speed-side second weight learning unit 24H uses an evaluation function in which the horizontal axis represents the speed-side second learning weight and the vertical axis represents the square of the speed deviation, and reduces the speed-side square to reduce the square of the speed deviation. The value of the second learning weight is changed (learned). An evaluation function is prepared for each weight.

  The speed-side second weight learning section 24H multiplies each of the input speed-side second outputs 24L2 by a corresponding speed-side second learning weight Y [1] to Y [3]. 24M2 is output.

● [Output layer 24C in neural network 243 (FIG. 15)]
The output layer 24C has a speed-side output section Q [1]. The speed-side output unit Q [1] converts a value obtained by adding the following speed-side second multiplication values of (11) to (13) by a predetermined function (for example, a sigmoid function), and converts the converted value to a fourth value. Output as 2 temporary command currents 24out. The speed-side output unit Q [1] can output the second provisional command current 24out without using the sigmoid function.
(11) A speed-side second multiplied value obtained by multiplying the speed-side second output output from the speed-side operation unit N [1] by the speed-side second learning weight Y [1].
(12) A speed-side second multiplied value obtained by multiplying the speed-side second output output from the speed-side operation unit N [2] by the speed-side second learning weight Y [2].
(13) A speed-side second multiplied value obtained by multiplying the speed-side second output output from the speed-side operation unit N [3] by the speed-side second learning weight Y [3].

● [Configuration of position feed forward control unit 14 (FIG. 16)]
Next, the configuration inside the position feedforward control unit 14 in the third embodiment will be described with reference to FIG. The position feedforward control unit 14 has an input processing unit 141 and a neural network 143. Note that the input processing unit 141 is the same as the input processing unit 141 of the first embodiment (see FIG. 9), and thus the description is omitted. Further, the upper command speed Vp and the upper command acceleration αp are the same as those in the first embodiment (see FIG. 9), and thus the description is omitted. The neural network 143 has an input layer 14A, a position-side first weight learning unit 14G, an intermediate layer 14B, a position-side second weight learning unit 14H, and an output layer 14C. In the third embodiment, an example will be described in which the intermediate layer 14B includes three position-side operation units M [1] to M [3], but the number of position-side operation units is limited to three. Not something.

● [Input layer 14A in neural network 143 (FIG. 16)]
As shown in FIG. 16, the input layer 14A in the neural network 143 includes a position-side acceleration input unit J [1], a position-side positive speed input unit J [2], a position-side negative speed input unit J [3], and a position-side negative speed input unit J [3]. The point having the boundary speed input units J [4] to J [j + n] is the same as in the first embodiment (see FIG. 9). However, the output from each input unit is different from that of the first embodiment.

  The position-side acceleration input unit J [1] receives the higher-order command acceleration αp and sends the input higher-order command acceleration αp to each of the plurality of position-side arithmetic units M [1] to M [3] prepared in advance. And output it as the position-side acceleration output.

  When the Vp (> 0) of the position-side positive / negative speed firing unit 14E fires, the position-side positive speed input unit J [2] receives and inputs the position-side positive speed output value that is the higher-order command speed Vp. The position-side positive speed output value is output as a position-side normal speed output to each of the plurality of position-side arithmetic units M [1] to M [3] prepared in advance. When the position-side positive / negative speed firing unit 14E fires Vp (<0), the position-side negative speed input unit J [3] receives and inputs the position-side negative speed output value that is the upper command speed Vp. The position-side negative speed output value is output as a position-side negative speed output to each of the plurality of position-side arithmetic units M [1] to M [3] prepared in advance.

  The position side boundary speed input units J [4] to J [j + n] are prepared corresponding to a plurality of boundary speeds -V0n to V1n. The position-side boundary speed input units J [4] to J [j + n] receive position-side boundary speed output values from the boundary speeds fired in the position-side boundary speed firing unit 14F, and the input position-side boundary speed output values. Is output as a position-side boundary speed output to each of the plurality of position-side arithmetic units M [1] to M [3] prepared in advance. The position-side acceleration output, the position-side positive speed output, the position-side negative speed output, and the position-side boundary speed output output from each of the input units J [1] to J [j + n] are collectively obtained as the position-side first output 14L1. And

● [Position-side first weight learning unit 14G in neural network 143 (FIG. 16)]
The position-side first output that is output from the position-side acceleration input unit J [1] to the position-side operation unit M [1] includes a position-side first learning weight W [1] [1] (W [s] [1] ]) Are associated. The position-side first learning weight W [2] [1] (W [s] [) is output to the position-side first output output from the position-side positive speed input unit J [2] to the position-side operation unit M [1]. 1]). Position-side first learning weights W [3] [1] (W [s] [) are output from the position-side negative speed input unit J [3] to the position-side operation unit M [1]. 1]). Position side first learning weights W [4] [1] to position side first outputs that are output from the position side boundary velocity input units J [4] to J [j + n] to the position side calculation unit M [1]. W [j + n] [1] (W [s] [1]) is associated. Thus, the position-side first learning weight W [s] [1] is associated with each position-side first output that is output to the position-side operation unit M [1].

  Similarly, the position-side first learning weight W [1] [2] (W [s]) is output to the position-side first output output from the position-side acceleration input unit J [1] to the position-side operation unit M [2]. ] [2]). The position-side first output that is output from the position-side positive speed input unit J [2] to the position-side operation unit M [2] includes a position-side first learning weight W [2] [2] (W [s] [ 2]). The position-side first output that is output from the position-side negative speed input unit J [3] to the position-side operation unit M [2] includes a position-side first learning weight W [3] [2] (W [s] [ 2]). Position-side first learning weights W [4] [2] to position-side first outputs output from the position-side boundary speed input units J [4] to J [j + n] to the position-side operation unit M [2] are provided. W [j + n] [2] (W [s] [2]) is associated. As described above, the position-side first learning weight W [s] [2] is associated with each position-side first output that is output to the position-side operation unit M [2].

  Similarly, the position-side first learning weight W [1] [3] (W [s]) is output to the position-side first output output from the position-side acceleration input unit J [1] to the position-side operation unit M [3]. ] [3]). The position-side first output that is output from the position-side positive speed input unit J [2] to the position-side operation unit M [3] includes a position-side first learning weight W [2] [3] (W [s] [ 3]). A position-side first learning weight W [3] [3] (W [s] [) is output to the position-side first output output from the position-side negative speed input unit J [3] to the position-side operation unit M [3]. 3]). The position-side first outputs output from the position-side boundary speed input units J [4] to J [j + n] to the position-side operation unit M [3] include position-side first learning weights W [4] [3] to W [j + n] [3] (W [s] [3]) is associated therewith. As described above, the position-side first learning weight W [s] [3] is associated with each position-side first output that is output toward the position-side operation unit M [3].

  The position-side first weight learning unit 14G has a storage function of storing these position-side first learning weights W [s] [1], W [s] [2], and W [s] [3]. ing. The position-side first weight learning unit 14G changes the position-side first learning weights W [s] [1], W [s] [2], and W [s] [3] based on the position deviation 10out ( Learning). The position-side first weight learning unit 14G has a multiplication function of multiplying each of the input position-side first outputs 14L1 by the corresponding position-side first learning weight.

  The position-side first weight learning unit 14G sets the position-side first learning weights W [s] [1], W [s] [2], and W [s] [3] in accordance with the position deviation 10out. Is changed (learned) so as to approach an optimum value for a predetermined evaluation function. For example, the position-side first weight learning unit 14G uses an evaluation function in which the horizontal axis represents the position-side first learning weight and the vertical axis represents the square of the position deviation, and reduces the position-side square to reduce the square of the position deviation. The value of the first learning weight is changed (learned). An evaluation function is prepared for each weight. When the higher command speed Vp shown in the example of FIG. 10 is input, the position-side first weight learning unit 14G outputs the position-side first learning weight W [1] corresponding to the output position-side first output. ] [1] to W [1] [3], W [2] [1] to W [2] [3], W [j + 2] [1] to W [j + 2] [3], W [j + 3] [ 1] to W [j + 3] [3], and does not learn the position-side first learning weights corresponding to the other position-side first outputs that have not been output.

  Then, the position-side first weight learning unit 14G outputs a position-side first multiplied value 14M1 obtained by multiplying each of the output position-side first outputs by the corresponding position-side first learning weight. For example, when the higher-order command speed Vp shown in the example of FIG. 10 is input, the position-side first weight learning unit 14G outputs αp * W [1] [1] to αp * W [1] [3] and Vp * W [2] [1] to Vp * W [2] [3], 0.8 * W [j + 1] [1] to 0.8 * W [j + 1] [3], and 0.6 * W [J + 2] [1] to 0.6 * W [j + 2] [3] are output. Further, for example, when the higher-order command speed Vp shown in the example of FIG. 11 is input, the position-side first weight learning unit 14G determines αp * W [1] [1] to αp * W [1] [3] and Vp * W [3] [1] to Vp * W [3] [3] and -0.7 * W [j-1] [1] to -0.7 * W [j-1] [3] , -0.5 * W [j] [1] to -0.5 * W [j] [3].

● [About the intermediate layer 14B in the neural network 143 (FIG. 16)]
The intermediate layer 14B has position-side operation units M [1] to M [3]. Note that the number of position-side operation units is not limited to three.

  The position-side operation unit M [1] outputs each of the position-side first outputs output from each of the input units J [1] to J [j + n] toward itself (the position-side operation unit M [1]), and A plurality of position-side first multiplied values obtained by multiplying the position-side first learning weights (W [s] [1]) corresponding to the position-side first outputs are added to form an output layer as a position-side second output. Output to 14C.

  The position-side operation unit M [2] outputs each of the position-side first outputs output from each of the input units J [1] to J [j + n] to itself (the position-side operation unit M [2]). A plurality of position-side first multiplication values obtained by multiplying a position-side first learning weight (W [s] [2]) corresponding to each of the position-side first outputs are added to form an output layer as a position-side second output. Output to 14C.

  The position-side operation unit M [3] outputs each of the position-side first outputs output from each of the input units J [1] to J [j + n] toward itself (the position-side operation unit M [3]). A plurality of position-side first multiplication values obtained by multiplying the position-side first learning weights (W [s] [3]) respectively corresponding to the position-side first outputs are added to form an output layer as a position-side second output. Output to 14C.

● [Position-side second weight learning unit 14H in neural network 143 (FIG. 16)]
The position-side second weight learning unit 14H includes a position-side second learning weight X [1] associated with the position-side operation unit M [1] and a position-side second weighting unit X [1] associated with the position-side operation unit M [2]. It has a storage function of storing the second learning weight X [2] and the position-side second learning weight X [3] associated with the position-side operation unit M [3]. The position-side second weight learning unit 14H has a changing function of changing (learning) the position-side second learning weights X [1] to X [3] based on the position deviation 10out. The position-side second weight learning unit 14H has a multiplication function of multiplying each of the input position-side second outputs 14L2 by each of the position-side second learning weights X [1] to X [3]. ing.

  The position-side second weight learning unit 14H adjusts each of the position-side second learning weights X [1] to X [3] in accordance with the position deviation 10out so as to approach an optimum value for a predetermined evaluation function. Change (learn). For example, the position-side second weight learning unit 14H uses the evaluation function in which the horizontal axis indicates the position-side second learning weight and the vertical axis indicates the square of the position deviation, and reduces the position-side square to reduce the square of the position deviation. The value of the second learning weight is changed (learned). An evaluation function is prepared for each weight.

  The position-side second weight learning unit 14H multiplies each of the input position-side second outputs 14L2 by a corresponding position-side second learning weight X [1] to X [3]. 14M2 is output.

● [About output layer 14C in neural network 143 (FIG. 16)]
The output layer 14C has a position-side output unit P [1]. The position-side output unit P [1] converts a value obtained by adding each of the following position-side second multiplication values of (11) to (13) by a predetermined function (for example, a sigmoid function), and converts the converted value to a fourth value. Output as 2 temporary command speeds 14out. The position-side output unit P [1] can output the second provisional command speed 14out without using the sigmoid function.
(11) A position-side second multiplied value obtained by multiplying the position-side second output output from the position-side operation unit M [1] by the position-side second learning weight X [1].
(12) A position-side second multiplied value obtained by multiplying the position-side second output output from the position-side operation unit M [2] by the position-side second learning weight X [2].
(13) The position-side second multiplied value obtained by multiplying the position-side second output output from the position-side operation unit M [3] by the position-side second learning weight X [3].

  As described above, in the third embodiment, the intermediate layer 24B, the intermediate layer 14B, the speed-side second weight learning unit 24H, and the position-side second weight learning unit 14H are added to the first embodiment. At the same time, the number of weights of the speed-side first weight learning unit 24G and the position-side first weight learning unit 14G has been increased to change from a perceptron to a neural network, and outputs from a plurality of input units are multiplied by weights, respectively. Since the neural network has a wider variety of combination patterns of the sum than the perceptron, the speed deviation and the position deviation can be further reduced as compared with the first embodiment.

●● [Fourth Embodiment (FIG. 14)]
Next, a fourth embodiment of the motor control device 92U will be described. As shown in FIG. 14, the motor control device according to the fourth embodiment differs from the motor control device according to the third embodiment shown in FIG. 2 in that a position feedforward control unit 14, a position-side input speed calculation unit 12, the position-side input acceleration calculator 13 and the speed addition calculator 15 are omitted, and the other components are the same. Hereinafter, differences will be mainly described. Note that the first provisional command speed 11out output from the position feedback control unit 11 is directly input to the speed deviation calculation unit 20 as the lower command speed 15out.

  In the motor control device according to the fourth embodiment shown in FIG. 14, the position feedforward control unit 14 and the input (position) to the position feedforward control unit 14 in the motor control device according to the third embodiment shown in FIG. The side input speed calculation unit 12, the position side input acceleration calculation unit 13) and the output (speed addition calculation unit 15) are omitted. Therefore, the control by the motor control device of the fourth embodiment slightly increases the position deviation compared to the control by the motor control device of the third embodiment. However, as a result of the experiment by the inventor, it was found that the reduction of the position deviation by the velocity feedforward control unit was more dominant than the reduction of the position deviation by the position feedforward control unit. For example, the robot shown in the example of FIG. With the control described above, it was found that the position deviation was sufficiently reduced even if the position feedforward control unit was omitted. That is, for example, in the case of robot control, the position deviation can be reduced to an expected level even if the position feedforward control unit (and the input / output unit to the position feedforward control unit) is omitted. In this case, the processing load on the motor control device can be reduced.

●● [Fifth Embodiment (FIGS. 14, 17 to 19)]
Next, a fifth embodiment of the motor control device 92U will be described. The motor control device according to the fifth embodiment is different from the motor control device according to the second embodiment illustrated in FIGS. 14 and 3 in that the internal configuration of the speed feedforward control unit 24 is different from the configuration illustrated in FIG. The difference is that the configuration is changed to the configuration shown in FIG. In the fifth embodiment, the function of the speed-side boundary speed input unit in the internal configuration of the speed feedforward control unit 24 is different from that of the second embodiment. In particular, as shown in the examples of FIGS. 18 and 19, the firing state with respect to the input higher command speed Vv is different from that of the second embodiment shown in the examples of FIGS.

● [Configuration of Network with Weight Learning Function of Speed Feedforward Control Unit 24 (FIG. 17)]
As shown in FIG. 17, the speed feedforward control unit 24 according to the fifth embodiment has a simple perceptron 242A. The simple perceptron 242A has an input layer 24AA, a speed-side first weight learning unit 24G, and an output layer 24C. Since the functions of the speed-side first weight learning unit 24G and the output layer 24C are the same as those described in the first and second embodiments, the speed-side first weight learning unit 24G and the output layer 24C are used. The description of is omitted. The speed feedforward control unit according to the fifth embodiment shown in FIG. 17 does not include the input processing unit 241 in the speed feedforward control units according to the first and second embodiments shown in FIG. The difference is that the input unit K [m] is changed to a speed-side boundary speed input unit H [m]. Hereinafter, differences will be mainly described.

● [Input layer 24AA in simple perceptron 242A (FIG. 17)]
As shown in FIG. 17, the input layer 24AA of the simple perceptron 242A includes a speed-side acceleration input unit H [1], a speed-side speed input unit H [2], and a speed-side boundary speed input unit H [4] to H [j +]. (N-1)].

  The speed-side acceleration input unit H [1] receives the higher-order command acceleration αv, and outputs the input higher-order command acceleration αv as a speed-side acceleration output. The function of the speed-side acceleration input unit H [1] is the same as the function of the speed-side acceleration input unit K [1] in the first and second embodiments.

  The speed-side speed input unit H [2] receives the higher-order command speed Vv, and outputs the input higher-order command speed Vv as a speed-side speed output. The speed-side speed input unit H [2] has a positive speed compared to the speed-side positive speed input unit K [2] and the speed-side negative speed input unit K [3] of the first and second embodiments. The difference is that no distinction is made between the case and the negative case.

  The speed-side boundary speed input units H [4] to H [j + (n-1)] are prepared corresponding to a plurality of boundary speeds shown in FIG. In the example of FIG. 4, the speed range is a speed range of boundary speeds -V0n to -V0 (n-1), a speed range of boundary speeds -V02 to -V01, a speed range of boundary speeds V11 to V12, and a speed range of boundary speed V1. There are speed ranges (n-1) to V1n, and the speed-side boundary speed input units H [4] to H [j + (n-1)] are prepared corresponding to these plural boundary speeds. . Each of the speed-side boundary speed input units H [4] to H [j + (n-1)] receives a higher-order command speed Vv, and a speed-side boundary corresponding to the input higher-order command speed Vv. The speed-side boundary speed output is output from the speed input section. The speed-side acceleration output, the speed-side speed output, and the speed-side boundary speed output output from each of the input units H [1], H [2], H [4] to H [j + (n-1)] are calculated as follows: Collectively, the speed-side first output 24L1 is used. Note that the speed-side boundary speed input unit H [4] in FIG. 17 includes boundary speeds −V0n to −V0 (n−1) illustrated in FIG. 4 (−V0 (n−1) is omitted in FIG. 4). Speed range. The speed-side boundary speed input unit H [j-1] in FIG. 17 corresponds to the speed range of the boundary speeds -V03 to -V02 shown in FIG. 4, and the speed-side boundary speed input unit H [j + 2] corresponds to FIG. Corresponds to the speed range of the boundary speeds V12 to V13. In the speed-side boundary speed input unit H [j + (n-1)] in FIG. 17, boundary speeds V1 (n-1) to V1n (V1 (n-1) shown in FIG. 4 are omitted in FIG. ) Speed range.

[Example of firing state of velocity side boundary velocity input units H [4] to H [j + (n-1)] of input layer 24AA in simple perceptron 242A (FIGS. 18 and 19)]
FIG. 18 is an example in which the input upper command speed Vv is positive (> 0) and the boundary speed V12 <the higher command speed Vv <the boundary speed V13, and each speed range has a function A. An example is shown. In the upper graph in FIG. 18, when the horizontal axis is the X axis and the vertical axis is the Y axis, the function A is (X, Y) = (lower boundary speed, 0) within the speed range to which the function A belongs. Is a linear proportional function connecting the position of (X, Y) = (upper limit boundary speed, 1.0), not including (lower limit boundary speed, 0), and (upper limit boundary speed, 1.0). )including. For example, the function A in the speed range from the boundary speed V12 to the boundary speed V13 does not include (the lower limit boundary speed V12, 0) (displayed by a white circle), but includes (the upper limit boundary speed V13, 1.0) (the black circle). Displayed at).

  In the example of FIG. 18, the upper command speed Vv is positive (> 0), the boundary speed V12 <the higher command speed Vv <the boundary speed V13, and the function A corresponding to the higher command speed Vv (the boundary speed V12 to V13 An example is shown in which the value of the speed range function A) is 0.6. In this case, as shown in FIG. 18, the speed-side boundary speed input unit H [j + 2] corresponding to the speed range of the boundary speeds V12 to V13, which is a portion corresponding to the upper command speed Vv, outputs the speed-side boundary speed output. = 0.6 is output as the speed-side boundary speed output. A speed-side boundary speed input unit H [j + 1] corresponding to the speed range of the boundary speeds V11 to V12, which is a portion not corresponding to the upper command speed Vv, and similarly, a speed not corresponding to the upper command speed Vv. The side boundary speed input units H [j + 3] to H [j + (n-1)] do not output anything. FIG. 18 shows an example in which the upper command speed Vv is positive (> 0). Therefore, the speed-side boundary speed input units H [4] to H [4] corresponding to the case where the upper command speed Vv is negative (<0). j] does not output anything. For example, when the upper command speed Vv = V13, the speed-side boundary speed input unit H [j + 2] corresponding to the speed range of the boundary speeds V12 to V13 outputs 1.0 as the speed-side boundary speed output, and When the command speed Vv = V12, the speed-side boundary speed input unit H [j + 1] corresponding to the speed range of the boundary speeds V11 to V12 outputs 1.0 as the speed-side boundary speed output. As described above, each of the speed-side boundary speed input units H [4] to H [j + (n-1)] has a function of determining whether or not to output a speed-side boundary speed output, and a function of determining whether to output the speed-side boundary speed output. And a function of calculating a value.

  FIG. 19 shows an example in which the input upper command speed Vv is negative (<0) and the boundary speed −V03 <upper command speed Vv <boundary speed−V02, and each speed range has a function A. An example is shown. In the upper graph in FIG. 19, when the horizontal axis is the X axis and the vertical axis is the Y axis, the function A is (X, Y) = (lower limit boundary speed, −1) within the speed range to which the function A belongs. .0) and a position of (X, Y) = (upper boundary speed, 0), which is a linear proportional function, including (lower boundary speed, −1.0), and (upper boundary speed, 0) is not included. For example, the function A in the speed range from the boundary speed -V03 to the boundary speed -V02 includes (lower boundary speed -V03, -1.0) (displayed by a black circle), and (upper boundary speed -V02, 0). Not included (indicated by white circles).

  In the example of FIG. 19, the upper command speed Vv is negative (<0), the boundary speed −V03 <the higher command speed Vv <the boundary speed −V02, and the function A (boundary speed −V03) corresponding to the higher command speed Vv is satisfied. An example is shown in which the value of the function A) in the speed range of -V02 is -0.3. In this case, as shown in FIG. 19, the speed-side boundary speed input unit H [j-1] corresponding to the speed range of the boundary speeds -V03 to -V02, which is a portion corresponding to the upper command speed Vv, The side boundary speed output = −0.3 is output as the speed side boundary speed output. Also, a speed-side boundary speed input unit corresponding to a speed range from the boundary speed -V0n to V0 (n-1) to a speed range from the boundary speed -V04 to -V03, which is a portion not corresponding to the upper command speed Vv. H [4] to H [j-2], similarly, the speed-side boundary speed input unit H [j] that does not correspond to the upper command speed Vv does not output anything. FIG. 19 shows an example in which the upper command speed Vv is negative (<0). Therefore, the speed-side boundary speed input units H [j + 1] to H [j] corresponding to the case where the upper command speed Vv is positive (> 0). j + (n-1)] also does not output anything. For example, when the upper command speed Vv = -V03, the speed-side boundary speed input unit H [j-1] corresponding to the speed range of the boundary speeds -V03 to -V02 outputs -1. 0, and when the upper command speed Vv = −V02, the speed-side boundary speed input unit H [j] corresponding to the speed range of the boundary speeds −V02 to −V01 outputs −1. Outputs 0. As described above, each of the speed-side boundary speed input units H [4] to H [j + (n-1)] has a function of determining whether or not to output a speed-side boundary speed output, and a function of determining whether to output the speed-side boundary speed output. And a function of calculating a value.

  In the simple perceptron 242A (see FIG. 17) of the motor control device according to the fifth embodiment, the function of the input processing unit 241 in the simple perceptron 242 (see FIG. 3) of the motor control device according to the second embodiment is defined by an input layer. 24AA, and the input processing unit 241 is omitted. Further, the speed-side positive speed input unit K [2] and the speed-side negative speed input unit K [3] of the second embodiment are integrated into the speed-side speed input unit H [2]. Therefore, the motor control device according to the fifth embodiment has a simplified configuration and a reduced processing load as compared with the motor control device according to the second embodiment. The second embodiment can output a plurality of boundary speed output values from a plurality of boundary speed input portions by igniting a plurality of boundary speed firing portions. On the other hand, the speed deviation is small. On the other hand, in the fifth embodiment, since only one boundary speed output value can be output from one boundary speed input unit, the discontinuity of the speed occurs and the configuration is simplified, but the speed deviation is reduced. It is easy to get a little worse.

●● [Sixth Embodiment (FIGS. 14, 17, 20, and 21)]
Next, a sixth embodiment of the motor control device 92U will be described. The motor control device according to the sixth embodiment differs from the motor control device according to the second embodiment shown in FIGS. 14 and 3 in that the speed feedforward control unit 24 is different from the motor control device according to the fifth embodiment. The difference is that the internal configuration is changed from the configuration shown in FIG. 3 to the configuration shown in FIG. Further, in the sixth embodiment, the function of the speed-side boundary speed input unit in the internal configuration of the speed feedforward control unit 24 is different, and the firing state for the input upper command speed Vv is different from that of FIGS. As shown in the example, the fifth embodiment is different from the fifth embodiment shown in the examples of FIGS. 18 and 19 (the difference is that the function A is replaced by the function B). Hereinafter, differences will be mainly described.

[Example of firing state of velocity side boundary velocity input units H [4] to H [j + (n-1)] of input layer 24AA in simple perceptron 242A (FIGS. 20 and 21)]
FIG. 20 is an example in which the input upper command speed Vv is positive (> 0) and the boundary speed V12 <the higher command speed Vv <the boundary speed V13, and each speed range has a function B. An example is shown. In the upper graph in FIG. 20, when the horizontal axis is the X axis and the vertical axis is the Y axis, the function B is (X, Y) = (lower limit boundary speed, 1...) Within the speed range to which the function B belongs. A linear proportional function connecting the position of (0) and the position of (X, Y) = (upper boundary speed, 0), including (lower boundary speed, 1.0), and (upper boundary speed, 0) Not included. For example, the function B in the speed range from the boundary speed V12 to the boundary speed V13 includes (lower boundary speed V12, 1.0) (displayed as a black circle) and does not include (upper boundary speed V13, 0) (open circle). Displayed at).

  In the example of FIG. 20, the upper command speed Vv is positive (> 0), the boundary speed V12 <the upper command speed Vv <the boundary speed V13, and the function B corresponding to the upper command speed Vv (the boundary speed V12 to V13 An example is shown in which the value of the speed range function B) is 0.4. In this case, as shown in FIG. 20, the speed-side boundary speed input unit H [j + 2] corresponding to the speed range of the boundary speeds V12 to V13, which is a portion corresponding to the upper command speed Vv, outputs the speed-side boundary speed output. = 0.4 is output as the speed-side boundary speed output. A speed-side boundary speed input unit H [j + 1] corresponding to the speed range of the boundary speeds V11 to V12, which is a portion not corresponding to the upper command speed Vv, and similarly, a speed not corresponding to the upper command speed Vv. The side boundary speed input units H [j + 3] to H [j + (n-1)] do not output anything. FIG. 20 shows an example in which the upper command speed Vv is positive (> 0). Therefore, the speed-side boundary speed input units H [4] to H [4] corresponding to the case where the upper command speed Vv is negative (<0). j] does not output anything. For example, when the upper command speed Vv = V12, the speed-side boundary speed input unit H [j + 2] corresponding to the speed range of the boundary speeds V12 to V13 outputs 1.0 as the speed-side boundary speed output, and When the command speed Vv = V13, the speed-side boundary speed input unit H [j + 3] corresponding to the speed range of the boundary speeds V13 to V14 outputs 1.0 as the speed-side boundary speed output. As described above, each of the speed-side boundary speed input units H [4] to H [j + (n-1)] has a function of determining whether or not to output a speed-side boundary speed output, and a function of determining whether to output the speed-side boundary speed output. And a function of calculating a value.

  FIG. 21 is an example in which the input upper command speed Vv is negative (<0) and the boundary speed −V03 <the higher command speed Vv <the boundary speed −V02, and each speed range has a function B. An example is shown. In the upper graph in FIG. 21, when the horizontal axis is the X axis and the vertical axis is the Y axis, the function B is (X, Y) = (lower limit boundary speed, 0) within the speed range to which the function B belongs. Is a linear proportional function connecting the position of (X, Y) = (upper limit boundary speed, −1.0), not including (lower limit boundary speed, 0), and (upper limit boundary speed, −1). .0). For example, the function B in the speed range from the boundary speed -V03 to the boundary speed -V02 does not include (lower limit boundary speed -V03, 0) (displayed by a white circle), and (upper limit boundary speed -V02, -1.0). ) (Indicated by black circles).

  In the example of FIG. 21, the upper command speed Vv is negative (<0), the boundary speed −V03 <the upper command speed Vv <the boundary speed −V02, and the function B (boundary speed −V03) corresponding to the upper command speed Vv is satisfied. An example is shown in which the value of the function B) in the speed range of -V02 is -0.7. In this case, as shown in FIG. 21, the speed-side boundary speed input unit H [j-1] corresponding to the speed range of the boundary speeds -V03 to -V02, which is a portion corresponding to the upper command speed Vv, The side boundary speed output = −0.7 is output as the speed side boundary speed output. Also, a speed-side boundary speed input unit corresponding to a speed range from the boundary speed -V0n to V0 (n-1) to a speed range from the boundary speed -V04 to -V03, which is a portion not corresponding to the upper command speed Vv. H [4] to H [j-2], similarly, the speed-side boundary speed input unit H [j] that does not correspond to the upper command speed Vv does not output anything. FIG. 21 shows an example in which the upper command speed Vv is negative (<0). Therefore, the speed-side boundary speed input units H [j + 1] to H [j] corresponding to the case where the upper command speed Vv is positive (> 0). j + (n-1)] also does not output anything. For example, when the upper command speed Vv = -V02, the speed-side boundary speed input unit H [j-1] corresponding to the speed range of the boundary speeds -V03 to -V02 outputs -1. 0, and when the upper command speed Vv = -V03, the speed-side boundary speed input unit H [j-2] corresponding to the speed range of the boundary speeds -V04 to -V03 is used as the speed-side boundary speed output. 1.0 is output. As described above, each of the speed-side boundary speed input units H [4] to H [j + (n-1)] has a function of determining whether or not to output a speed-side boundary speed output, and a function of determining whether to output the speed-side boundary speed output. And a function of calculating a value.

  In the simple perceptron 242A (see FIG. 17) of the motor control device according to the sixth embodiment, similarly to the fifth embodiment, the simple perceptron 242 (see FIG. 3) of the motor control device according to the second embodiment. Are integrated into the input layer 24AA, and the input processing unit 241 is omitted. Further, the speed-side positive speed input unit K [2] and the speed-side negative speed input unit K [3] of the second embodiment are integrated into the speed-side speed input unit H [2]. Therefore, the motor control device of the sixth embodiment has a simplified configuration and a reduced processing load as compared with the motor control device of the second embodiment, as in the fifth embodiment. Have been. Further, in the sixth embodiment, as in the fifth embodiment, only one boundary speed output value can be output from one boundary speed input unit. At the same time, the speed deviation is likely to be slightly worse.

●● [Seventh Embodiment (FIGS. 14, 17, 22, and 23)]
Next, a seventh embodiment of the motor control device 92U will be described. The motor control device according to the seventh embodiment differs from the motor control device according to the second embodiment shown in FIGS. The difference is that the internal configuration is changed from the configuration shown in FIG. 3 to the configuration shown in FIG. Further, in the seventh embodiment, the function of the speed-side boundary speed input unit in the internal configuration of the speed feedforward control unit 24 is different, and the firing state with respect to the input upper command speed Vv is shown in FIGS. As shown in the example, the fifth embodiment is different from the fifth embodiment shown in the examples of FIGS. 18 and 19 (the difference is that the function A is replaced by the function C). Hereinafter, differences will be mainly described.

[Example of firing state of velocity side boundary velocity input units H [4] to H [j + (n-1)] of input layer 24AA in simple perceptron 242A (FIGS. 22 and 23)]
FIG. 22 is an example in which the input upper command speed Vv is positive (> 0) and the boundary speed V12 <the higher command speed Vv <the boundary speed V13, and each speed range has a function C. An example is shown. In the upper graph in FIG. 22, when the horizontal axis is the X axis and the vertical axis is the Y axis, the function C is (X, Y) = (lower limit boundary speed, 0) within the speed range to which the function B belongs. , The position of (X, Y) = (the center of the lower boundary speed and the upper boundary speed, 1.0), and the position of (X, Y) = (upper boundary speed, 0) are sequentially connected. It is a function of an isosceles triangular shape (excluding the base), not including (lower boundary speed, 0) but including (upper boundary speed, 0). For example, the function C in the speed range from the boundary speed V12 to the boundary speed V13 does not include (lower limit boundary speed V12, 0) (displayed by a white circle), but includes (upper limit boundary speed V13, 0) (black circles). display).

  In the example of FIG. 22, the upper command speed Vv is positive (> 0), the boundary speed V12 <the upper command speed Vv <the boundary speed V13, and the function C corresponding to the higher command speed Vv (the boundary speed V12 to V13 An example in which the value of the speed range function C) is 0.8 is shown. In this case, as shown in FIG. 22, the speed-side boundary speed input unit H [j + 2] corresponding to the speed range of the boundary speeds V12 to V13, which is a location corresponding to the higher command speed Vv, outputs the speed-side boundary speed output. = 0.8 is output as the speed-side boundary speed output. A speed-side boundary speed input unit H [j + 1] corresponding to the speed range of the boundary speeds V11 to V12, which is a portion not corresponding to the upper command speed Vv, and similarly, a speed not corresponding to the upper command speed Vv. The side boundary speed input units H [j + 3] to H [j + (n-1)] do not output anything. FIG. 22 shows an example in which the upper command speed Vv is positive (> 0). Therefore, the speed-side boundary speed input units H [4] to H [4] corresponding to the case where the upper command speed Vv is negative (<0). j] does not output anything. For example, when the upper command speed Vv = V13, the speed-side boundary speed input unit H [j + 2] corresponding to the speed range of the boundary speeds V12 to V13 outputs 0 as the speed-side boundary speed output, and outputs the higher command speed. When Vv = V12, the speed-side boundary speed input unit H [j + 1] corresponding to the speed range of the boundary speeds V11 to V12 outputs 0 as the speed-side boundary speed output. As described above, each of the speed-side boundary speed input units H [4] to H [j + (n-1)] has a function of determining whether or not to output a speed-side boundary speed output, and a function of determining whether to output the speed-side boundary speed output. And a function of calculating a value.

  FIG. 23 shows an example in which the input upper command speed Vv is negative (<0) and the boundary speed −V03 <upper command speed Vv <boundary speed−V02, and each speed range has a function C. An example is shown. In the upper graph in FIG. 23, when the horizontal axis is the X axis and the vertical axis is the Y axis, the function C is (X, Y) = (lower boundary speed, 0) within the speed range to which the function C belongs. , The position of (X, Y) = (the center of the lower boundary speed and the upper boundary speed, −1.0), and the position of (X, Y) = (upper boundary speed, 0) in order. This is a function of a connected isosceles triangle (excluding the base), including (lower boundary speed, 0) and not (upper boundary speed, 0). For example, the function C in the speed range from the boundary speed −V03 to the boundary speed −V02 includes (lower boundary speed −V03, 0) (displayed by a black circle) and does not include (upper boundary speed −V02, 0). (Indicated by white circles).

  In the example of FIG. 23, the upper command speed Vv is negative (<0), the boundary speed −V03 <the upper command speed Vv <the boundary speed −V02, and the function C corresponding to the upper command speed Vv (the boundary speed −V03). An example is shown in which the value of the function C) in the speed range of -V02 is -0.6. In this case, as shown in FIG. 23, the speed-side boundary speed input unit H [j-1] corresponding to the speed range of the boundary speeds -V03 to -V02, which is a portion corresponding to the higher command speed Vv, The side boundary speed output = −0.6 is output as the speed side boundary speed output. Also, a speed-side boundary speed input unit corresponding to a speed range from the boundary speed -V0n to V0 (n-1) to a speed range from the boundary speed -V04 to -V03, which is a portion not corresponding to the upper command speed Vv. H [4] to H [j-2], similarly, the speed-side boundary speed input unit H [j] that does not correspond to the upper command speed Vv does not output anything. FIG. 23 shows an example in which the upper command speed Vv is negative (<0). Therefore, the speed-side boundary speed input units H [j + 1] to H [j] corresponding to the case where the upper command speed Vv is positive (> 0). j + (n-1)] also does not output anything. For example, when the upper command speed Vv = -V03, the speed-side boundary speed input unit H [j-1] corresponding to the speed range of the boundary speeds -V03 to -V02 outputs 0 as the speed-side boundary speed output. When the upper command speed Vv = -V02, the speed-side boundary speed input unit H [j] corresponding to the speed range of the boundary speeds -V02 to -V01 outputs 0 as the speed-side boundary speed output. As described above, each of the speed-side boundary speed input units H [4] to H [j + (n-1)] has a function of determining whether or not to output a speed-side boundary speed output, and a function of determining whether to output the speed-side boundary speed output. And a function of calculating a value.

  In the simple perceptron 242A of the motor control device of the seventh embodiment (see FIG. 17), similarly to the fifth embodiment, the simple perceptron 242 of the motor control device of the second embodiment (see FIG. 3). Are integrated into the input layer 24AA, and the input processing unit 241 is omitted. Further, the speed-side positive speed input unit K [2] and the speed-side negative speed input unit K [3] of the second embodiment are integrated into the speed-side speed input unit H [2]. Therefore, the motor control device of the seventh embodiment has a simplified configuration and a reduced processing load as compared with the motor control device of the second embodiment, as in the fifth embodiment. Have been. In the seventh embodiment, as in the fifth embodiment, only one boundary speed output value can be output from one boundary speed input unit. At the same time, the speed deviation is likely to be slightly worse.

  The motor control device 92U of the present invention is not limited to the configuration described in the present embodiment, the configuration of the speed feedforward control unit, the configuration of the position feedforward control unit, and the like. Various changes, additions, and deletions are possible within a range not changed.

  In the description of the present embodiment, the upper command speed Vv (upper command speed Vp) is output from the speed-side input speed calculator 22 (the position-side input speed calculator 12) based on the command position. May be directly output from the robot control device 60. Further, in the description of the present embodiment, the upper command acceleration αv (upper command acceleration αp) is obtained from the speed-side input acceleration calculator 23 (position-side input acceleration calculator 13) based on the upper command speed Vv (upper command speed Vp). However, as another embodiment, it may be directly output from the robot control device 60.

  In the description of the present embodiment, an example in which a normal distribution function is used as the predetermined distribution function in the speed-side boundary speed firing unit 24F and the position-side boundary speed firing unit 14F has been described. However, when using a distribution function, various distribution functions can be used without being limited to the normal distribution function. Similarly, the functions A, B, and C described in the fifth to seventh embodiments are not limited to the functions A, B, and C described in the fifth to seventh embodiments. Functions can be used.

  The motor control device described in the present embodiment is not limited to a robot, and can be applied to a motor control device that controls the position of a predetermined member in various devices.

  Above (≧), below (≦), greater than (>), less than (<) and the like may or may not include an equal sign.

Reference Signs List 10 position deviation calculation unit 10in command position 10out position deviation 11 position feedback control unit 11out first provisional command speed 12 position side input speed calculation unit 12out input speed (position side speed)
13 Position-side input acceleration calculator 13out Input acceleration (position-side acceleration)
14 Position feed forward control unit 14out Second provisional command speed 141 Input processing unit 142, 142A Simple perceptron 143 Neural network 14A, 14AA Input layer 14B Middle layer 14C Output layer 14E Position-side positive / negative speed firing unit 14F Position-side boundary speed firing unit 14G Position-side first weight learning unit 14H Position-side second weight learning unit 14L1 Position-side first output 14L2 Position-side second output 14M1 Position-side first multiplication value 14M2 Position-side second multiplication value 15 Speed addition calculation unit 15out Lower-order command speed Reference Signs List 20 speed deviation calculation unit 20out speed deviation 21 speed feedback control unit 21out first provisional command current 22 speed side input speed calculation unit 22out input speed (speed side speed)
23 Speed side input acceleration calculator 23out Input acceleration (speed side acceleration)
24 Speed feed forward control unit 24out Second provisional command current 241 Input processing unit 242, 242A Simple perceptron 243 Neural network 24A, 24AA Input layer 24B Middle layer 24C Output layer 24E Speed side positive / negative speed firing unit 24F Speed side boundary speed firing unit 24G Speed-side first weight learning unit 24H Speed-side second weight learning unit 24L1 Speed-side first output 24L2 Speed-side second output 24M1 Speed-side first multiplication value 24M2 Speed-side second multiplication value 25 Current addition operation unit 25out Command current 27out Actual position 28 Actual speed calculator 28out Actual speed 30 Current deviation calculator 30out Current deviation 31 Current feedback controller 31out Drive current 60 Robot controller 90 Robot 91U to 94U Motor controller 92M Electric motor 92E Encoder (position detecting means ( Degree detection means))
H [1] Speed side acceleration input unit H [2] Speed side speed input unit H [4] to H [j + (n-1)] Speed side boundary speed input unit J [1] Position side acceleration input unit J [2 ] Position-side positive speed input section J [3] Position-side negative speed input section J [4] to J [j + n] Position-side boundary speed input section K [1] Speed-side acceleration input section K [2] Speed-side positive speed input Section K [3] Speed side negative speed input section K [4] to K [j + n] Speed side boundary speed input section M [1] to M [3] Position side calculation section N [1] to N [3] Speed side Calculation unit N Distribution function U [1], U [2], U [3], U [4] to U [j + n] Speed-side first learning weight U [s] [1], U [s] [2] , U [s] [3] Speed-side first learning weight W [1], W [2], W [3], W [4] to W [j + n] Position-side first learning weight W [s] [1 ], W [s] [2] , W [s] [3] Position-side first learning weight X [1] to X [3] Position-side second learning weight Y [1] to Y [3] Speed-side second learning weight −V0n, −V05, -V04, -V03, -V02, -V01 Boundary speed V11, V12, V13, V14, V15, V1n Boundary speed Vv, Vp Upper command speed αv, αp Upper command acceleration

Claims (8)

An electric motor that moves the position of the control target, and a position detection unit that detects a position related to the electric motor, a motor control device that controls the position of the control target using a motor control device,
A position deviation calculation unit that calculates a position deviation that is a deviation between a command position for the electric motor and an actual position that is an actual position based on a detection signal from the position detection unit;
A position feedback control unit that performs a feedback control according to the position deviation to output a first provisional command speed;
A speed deviation that is a deviation between the lower command speed that is the lower command speed for the electric motor, including the first provisional command speed, and the actual speed that is the actual speed based on the detection signal from the position detection unit. A speed deviation calculation unit for calculating
A speed feedback control unit that performs a feedback control according to the speed deviation to output a first provisional command current;
A speed feedforward control unit that performs feedforward control according to an upper command speed that is different from the lower command speed and outputs a second provisional command current;
A current addition operation unit that outputs the command current by adding the first provisional command current and the second provisional command current;
A current output unit that outputs a drive current to the electric motor based on the command current,
The speed feedforward control unit includes:
A higher-order command acceleration is input, a speed-side acceleration input unit that outputs the input higher-order command acceleration as a speed-side acceleration output,
A speed-side speed input unit to which the higher-order command speed is input and that outputs the input higher-order command speed as a speed-side speed output;
The limited speed range limited to the speed range of the upper command speed is divided into a plurality of adjacent speed ranges that are set in advance, and a plurality of boundary speeds that are boundaries between the divided speed ranges are obtained. A plurality of speed-side boundary speed input units, which are prepared in correspondence with each other, wherein the higher-order command speed is input, and outputs a speed-side boundary speed output from a location corresponding to the higher-order command speed,
A plurality of speed-side first learning weights corresponding to each of the speed-side first outputs, which are the speed-side acceleration output, the speed-side speed output, and the plurality of speed-side boundary speed outputs, are changed according to the speed deviation. A speed-side first weight learning unit,
A plurality of speed-side first multiplied values obtained by multiplying each of the speed-side first outputs and the speed-side first learning weight corresponding to each of the speed-side first outputs are added to obtain the second provisional command. A speed-side output unit that outputs current.
Motor control device. An electric motor that moves the position of the control target, and a position detection unit that detects a position related to the electric motor, a motor control device that controls the position of the control target using a motor control device,
A position deviation calculation unit that calculates a position deviation that is a deviation between a command position for the electric motor and an actual position that is an actual position based on a detection signal from the position detection unit;
A position feedback control unit that performs a feedback control according to the position deviation to output a first provisional command speed;
A speed deviation that is a deviation between the lower command speed that is the lower command speed for the electric motor, including the first provisional command speed, and the actual speed that is the actual speed based on the detection signal from the position detection unit. A speed deviation calculation unit for calculating
A speed feedback control unit that performs a feedback control according to the speed deviation to output a first provisional command current;
A speed feedforward control unit that performs feedforward control according to an upper command speed that is different from the lower command speed and outputs a second provisional command current;
A current addition operation unit that outputs the command current by adding the first provisional command current and the second provisional command current;
A current output unit that outputs a drive current to the electric motor based on the command current,
The speed feedforward control unit includes:
The upper command speed is input, and outputs a speed-side positive speed output value when the input upper command speed is positive, and outputs a speed-side negative speed output value when the input higher command speed is negative. Speed side positive / negative speed firing part,
The limited speed range limited to the speed range of the higher-order command speed is divided into a plurality of adjacent speed ranges set in advance, and a plurality of boundary speeds that are boundaries between the divided speed ranges are obtained. The upper command speed is input, and a speed difference from the upper command speed among the plurality of boundary speeds is the boundary speed within a predetermined speed difference, and a speed-side boundary speed output value based on the speed difference is provided. An output speed-side boundary speed firing section,
A higher-order command acceleration is input, a speed-side acceleration input unit that outputs the input higher-order command acceleration as a speed-side acceleration output,
A speed-side positive speed output unit to which the speed-side normal speed output value is input and that outputs the input speed-side normal speed output value as a speed-side normal speed output;
A speed-side negative speed input unit that receives the speed-side negative speed output value and outputs the input speed-side negative speed output value as a speed-side negative speed output;
A plurality of speed-side boundary speed input units that are prepared corresponding to a plurality of the boundary speeds, receive the speed-side boundary speed output value, and output the input speed-side boundary speed output value as a speed-side boundary speed output. When,
A plurality of speed-side first learning weights corresponding to each of the speed-side first outputs, which are the speed-side acceleration output, the speed-side positive speed output, the speed-side negative speed output, and the plurality of speed-side boundary speed outputs. A speed-side first weight learning unit that changes in accordance with the speed deviation;
A plurality of speed-side first multiplied values obtained by multiplying each of the speed-side first outputs and the speed-side first learning weight corresponding to each of the speed-side first outputs are added to obtain the second provisional command. A speed-side output unit that outputs current.
Motor control device. The motor control device according to claim 2, wherein
A position feedforward control unit that performs feedforward control according to the higher-order command speed and outputs a second provisional command speed;
A speed addition operation unit that adds the first provisional command speed and the second provisional command speed to output the lower command speed,
The position feed forward control unit includes:
The upper command speed is input, and outputs a position-side positive speed output value when the input higher command speed is positive, and outputs a position-side negative speed output value when the input higher command speed is negative. Positive / negative velocity firing part
A plurality of the boundary speeds, wherein the upper command speed is input, and a speed difference between the plurality of the boundary speeds and the upper command speed is within a predetermined speed difference. A position-side boundary speed firing section that outputs a side boundary speed output value;
A position-side acceleration input unit that receives the higher-order command acceleration and outputs the input higher-order command acceleration as a position-side acceleration output;
The position-side positive speed output unit that receives the position-side normal speed output value and outputs the input position-side normal speed output value as a position-side normal speed output,
A position-side negative speed input unit that receives the position-side negative speed output value and outputs the input position-side negative speed output value as a position-side negative speed output;
A plurality of position-side boundary speed input units which are prepared in correspondence with a plurality of the boundary speeds, receive the position-side boundary speed output values, and output the input position-side boundary speed output values as position-side boundary speed outputs; When,
A plurality of position-side first learning weights corresponding to the position-side first outputs, which are the position-side acceleration output, the position-side positive speed output, the position-side negative speed output, and the plurality of the position-side boundary speed outputs. A position-side first weight learning unit that changes in accordance with the position deviation;
A plurality of position-side first multiplication values obtained by multiplying each of the position-side first outputs and the position-side first learning weights corresponding to each of the position-side first outputs are added, and the second provisional command is obtained. A position-side output unit that outputs a speed.
Motor control device. An electric motor that moves the position of the control target, and a position detection unit that detects a position related to the electric motor, a motor control device that controls the position of the control target using a motor control device,
A position deviation calculation unit that calculates a position deviation that is a deviation between a command position for the electric motor and an actual position that is an actual position based on a detection signal from the position detection unit;
A position feedback control unit that performs a feedback control according to the position deviation to output a first provisional command speed;
A speed deviation that is a deviation between the lower command speed that is the lower command speed for the electric motor, including the first provisional command speed, and the actual speed that is the actual speed based on the detection signal from the position detection unit. A speed deviation calculation unit for calculating
A speed feedback control unit that performs a feedback control according to the speed deviation to output a first provisional command current;
A speed feedforward control unit that performs feedforward control according to an upper command speed that is different from the lower command speed and outputs a second provisional command current;
A current addition operation unit that outputs the command current by adding the first provisional command current and the second provisional command current;
A current output unit that outputs a drive current to the electric motor based on the command current,
The speed feedforward control unit includes:
The upper command speed is input, and outputs a speed-side positive speed output value when the input upper command speed is positive, and outputs a speed-side negative speed output value when the input higher command speed is negative. Speed side positive / negative speed firing part,
The limited speed range limited to the speed range of the higher-order command speed is divided into a plurality of adjacent speed ranges set in advance, and a plurality of boundary speeds that are boundaries between the divided speed ranges are obtained. The upper command speed is input, and a speed difference from the upper command speed among the plurality of boundary speeds is the boundary speed within a predetermined speed difference, and a speed-side boundary speed output value based on the speed difference is provided. An output speed-side boundary speed firing section,
A higher-order command acceleration is input, the input higher-order command acceleration, a speed-side acceleration input unit that outputs as a speed-side acceleration output toward each of a plurality of speed-side calculation units prepared in advance,
A speed-side positive speed output unit to which the speed-side normal speed output value is input, and that the input speed-side normal speed output value is output as a speed-side normal speed output toward each of the plurality of speed-side arithmetic units; ,
The speed-side negative speed output value is input, and the input speed-side negative speed output value is output as a speed-side negative speed output toward each of the plurality of speed-side operation units. ,
The speed-side boundary speed output value is provided corresponding to a plurality of the boundary speeds, the speed-side boundary speed output value is input, and the input speed-side boundary speed output value is converted into a speed-side boundary A plurality of speed-side boundary speed input units for outputting as speed output;
A plurality of speeds corresponding to each of the plurality of speed-side acceleration outputs, the plurality of speed-side positive speed outputs, the plurality of speed-side negative speed outputs, and the plurality of speed-side boundary speed outputs; A speed-side first weight learning unit that changes the first side learning weight according to the speed deviation;
A plurality of speed-side first multiplied values obtained by multiplying each of the speed-side first outputs and the speed-side first learning weight corresponding to each of the speed-side first outputs are added to generate a speed-side second output. A plurality of speed-side arithmetic units that output as
A speed-side second weight learning unit that changes a plurality of speed-side second learning weights corresponding to each of the speed-side second outputs in accordance with the speed deviation;
A plurality of speed-side second multiplied values obtained by multiplying each of the speed-side second outputs and the speed-side second learning weight corresponding to each of the speed-side second outputs are added, and the second provisional command is obtained. A speed-side output unit that outputs current.
Motor control device. The motor control device according to claim 4, wherein
A position feedforward control unit that performs feedforward control according to the higher-order command speed and outputs a second provisional command speed;
A speed addition operation unit that adds the first provisional command speed and the second provisional command speed to output the lower command speed,
The position feed forward control unit includes:
The upper command speed is input, and outputs a position-side positive speed output value when the input higher command speed is positive, and outputs a position-side negative speed output value when the input higher command speed is negative. Positive / negative velocity firing part
A plurality of the boundary speeds, wherein the upper command speed is input, and a speed difference between the plurality of the boundary speeds and the upper command speed is within a predetermined speed difference. A position-side boundary speed firing section that outputs a side boundary speed output value;
The upper-order command acceleration is input, a position-side acceleration input unit that outputs the input higher-order command acceleration as a position-side acceleration output toward each of a plurality of position-side calculation units prepared in advance,
The position-side normal speed output value is input, and the input position-side normal speed output value is output as a position-side normal speed output toward each of the plurality of position-side calculation units. ,
The position-side negative speed output unit receives the position-side negative speed output value, and outputs the input position-side negative speed output value as a position-side negative speed output toward each of the plurality of position-side operation units. ,
The position-side boundary speed output value is provided corresponding to a plurality of the boundary speeds, and the input position-side boundary speed output value is input to each of the plurality of position-side operation units. A plurality of position-side boundary speed input units that output as a speed output;
A plurality of positions corresponding to each of the position-side first outputs, which are a plurality of the position-side acceleration outputs, a plurality of the position-side positive speed outputs, a plurality of the position-side negative speed outputs, and a plurality of the position-side boundary speed outputs. A position-side first weight learning unit that changes a side first learning weight according to the position deviation;
A plurality of position-side first multiplied values obtained by multiplying each of the position-side first outputs and the position-side first learning weight corresponding to each of the position-side first outputs are added to obtain a position-side second output. A plurality of the position-side calculation units that output as
A position-side second weight learning unit that changes a plurality of position-side second learning weights corresponding to each of the position-side second outputs according to the position deviation;
A plurality of position-side second multiplied values obtained by multiplying each of the position-side second outputs and the position-side second learning weights corresponding to each of the position-side second outputs are added to obtain the second provisional command. A position-side output unit that outputs a speed.
Motor control device. The motor control device according to any one of claims 1 to 5,
Speed, a predetermined physical phenomenon including friction at the time of moving the position of the control target, a speed-physical phenomenon characteristics showing the relationship between the limited speed range considered as non-linear characteristics, partially Divided into each of the speed ranges that can be regarded as linear characteristics, the speed that is the boundary of each of the divided speed ranges is set as the boundary speed,
Motor control device. A motor control device according to any one of claims 2 to 5,
The speed-side boundary speed firing section,
It has a predetermined distribution function that makes the width of the speed set in advance a spread width,
A distribution probability corresponding to a speed separated from the center of the distribution function by the speed difference between the input upper command speed and the boundary speed is obtained for each of the boundary speeds by using the distribution function, and is obtained. Further, the distribution probability is fired only for the boundary speed that is not zero, a positive value based on the distribution probability obtained when the upper command speed is positive, and a positive value based on the distribution probability obtained when the upper command speed is negative. A negative value based on the distribution probability is output as the speed-side boundary speed output value corresponding to the boundary speed that fired,
Motor control device. The motor control device according to claim 3 or 5,
The position-side boundary speed firing section,
It has a predetermined distribution function that makes the width of the speed set in advance a spread width,
A distribution probability corresponding to a speed separated from the center of the distribution function by the speed difference between the input upper command speed and the boundary speed is obtained for each of the boundary speeds by using the distribution function, and is obtained. Further, the distribution probability is fired only for the boundary speed that is not zero, a positive value based on the distribution probability obtained when the upper command speed is positive, and a positive value based on the distribution probability obtained when the upper command speed is negative. A negative value based on the distribution probability is output as the position-side boundary speed output value corresponding to the boundary speed that fired,
Motor control device.

Images