JP2013004021A - Collision risk determination device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To accurately determine collision risk.SOLUTION: In a collision risk determination device 100, an environment detection part 5b detects positional distribution and movement state distribution of moving articles; a map generation part 5c generates an existence possibility map; a moving article generation part 5d, a position update part 5e and a distribution change part 5f predict a future positional distribution of the moving articles on the existence possibility map; a course candidate computation part 5a computes plural course candidates of the subject vehicle 10; a risk candidate computation part 5h computes respective collision risk candidates of the plural course candidates on the existence possibility map; and risk determination part 5i selects one course candidate from among the plural course candidates on the basis of the computed plural collision risk candidates, thereby determining the collision risk candidate of the selected one course candidate as a collision risk. Here, the risk candidate computation part 5h computes the collision risk candidates on the basis of a movement cost, a motion cost and collision probability.

Description

  The present invention relates to a collision risk determination device.

  2. Description of the Related Art Conventionally, a collision risk determination device has been developed that determines the collision risk of a host vehicle against a moving object (see, for example, Patent Document 1). In such a risk determination apparatus, a change in position that a moving object can take as time elapses is generated as a trajectory in time and space, and the course (position) of the moving object is probabilistically predicted. And the collision probability of the own vehicle with respect to a moving object is determined as a collision risk from the predicted future position of the moving object and the course candidate of the own vehicle.

JP 2007-233646 A

  Here, in recent collision risk determination devices, there is an increasing demand for collision safety of vehicles, and it is desired to improve the determination accuracy of the collision risk. In this regard, in the risk determination device as described above, there is room for improvement in, for example, the position prediction of a moving object. Further, for example, when the calculated course candidate of the own vehicle varies due to an error of a sensor or the like, there is a possibility that the collision risk degree to be determined varies accordingly.

  Then, this invention makes it a subject to provide the collision risk determination apparatus which can determine a collision risk accurately.

  In order to solve the above problems, a collision risk determination device according to the present invention is a collision risk determination device that determines a collision risk of a host vehicle against a moving object, and detects a position distribution and a movement state distribution of the moving object. A moving object detecting unit, a map generating unit for generating an existence possibility map that gives the existence possibility indicating the ease of existence or difficulty of existence of the moving object on a map indicating the surrounding area of the host vehicle, and movement A future position distribution prediction unit that predicts the future position distribution of the moving object on the existence possibility map based on the position distribution of the object, the movement state distribution, and the existence possibility, and the own position based on the movement state of the own vehicle A route candidate calculation unit that calculates a plurality of future route candidates of the vehicle, and a risk candidate calculation that calculates a collision risk candidate for each of the plurality of route candidates on the existence possibility map in which the future position distribution of the moving object is predicted And operations A risk determination unit that selects one of the plurality of route candidates based on the plurality of collision risk candidates, and determines the collision risk candidate of the selected one of the route candidates as a collision risk; The risk candidate calculation unit includes a movement cost obtained according to a route through which the route candidate passes and an exercise cost obtained according to a change in movement of the own vehicle when the vehicle moves along the route candidate. And calculating the collision risk candidate based on the future position distribution and the collision probability obtained according to the course candidate.

  In the collision risk determination device of the present invention, the position prediction accuracy of the moving object can be increased by predicting the future position distribution of the moving object on the existence possibility map. Also, for each of the plurality of course candidates, a collision risk candidate is calculated from the movement cost, the motion cost, and the collision probability, a course candidate is selected based on the collision risk candidate, and the collision risk candidate of the selected course candidate is collided. Judged as the risk level. As described above, the collision risk is not determined only from the collision probability of one course candidate, but the collision risk candidates of a plurality of course candidates are calculated, and the collision risk is determined from the collision risk candidates. Even if the course candidate and thus the collision probability varies, it is possible to suppress the adverse influence of the variation on the determination of the collision risk. Therefore, it is possible to accurately determine the collision risk.

  Moreover, it is preferable that the vehicle further includes a travel support unit that supports the travel of the host vehicle based on the collision risk determined by the risk determination unit. In this case, it is possible to accurately support the traveling of the host vehicle according to the determination of the collision risk with high accuracy.

  Moreover, it is preferable that the risk determination unit selects a route candidate having the lowest collision risk candidate among a plurality of route candidates. In this case, it is possible to favorably reflect the driver tendency of selecting a course with a low risk of collision.

  Further, the route candidate selection unit further obtains an evaluation function based on the movement cost and the movement cost for each of the plurality of route candidates calculated by the route candidate calculation unit, and selects a part of the plurality of route candidates based on the evaluation function. It is preferable that the risk candidate calculation unit calculates a collision risk candidate for a part of the plurality of course candidates selected by the course candidate selection unit. In this case, it is possible to narrow down in advance the route candidates to be calculated by the risk candidate calculation unit, and it is possible to reduce the calculation time for the collision risk candidate.

  In addition, the future position distribution prediction unit records the position distribution on the existence possibility map, moves the position distribution based on the movement state distribution, and changes the position distribution based on the existence possibility. Therefore, it is preferable to set the future position distribution.

  The existence possibility map is a grid map, and each of the plurality of grids is set with a unit movement cost required when the host vehicle passes, and the movement cost is indicated on the existence possibility map. It is preferable that it is the sum of the unit movement costs of a plurality of grids through which the route candidates pass.

  The exercise cost is preferably a time integral value of the absolute value of the yaw rate change in the host vehicle, a time integral value of the absolute value of the steering change, or a time integral value of the absolute value of the acceleration.

Further, the future position distribution of the moving object is represented by a plurality of moving object particles, the existence possibility map is a grid map, and the collision probability is the total number of grids through which path candidates pass on the existence possibility map. Is M _total , the number of grids in which moving object particles are present is m _col , the total number of moving object particles is N _total, and the number of moving object particles in the grid through which the path candidate passes is n _col , n be represented by _{col / N total · m col /} M total, is preferred.

  According to the present invention, it is possible to accurately determine the collision risk.

It is a block diagram which shows the collision risk determination apparatus which concerns on 1st Embodiment. (A) is a figure for demonstrating the example of the calculation result of a course candidate calculation part, (b) is a figure for demonstrating the other example of the calculation result of a course candidate calculation part. It is a flowchart which shows operation | movement of the collision risk determination apparatus of FIG. It is a figure which shows the example of driving environment. It is a figure which shows the example of a presence possibility map corresponding to the driving | running | working environment of FIG. It is a figure which shows the example of the table used for the setting of a presence possibility. (A) The figure which shows the example of the existence possibility map which has arrange | positioned the moving object particle, (b) The figure which shows the example of the existence possibility map which changed the position of the moving object particle, (c) The position of the moving object particle is changed. It is a figure which shows the other example of the existence possibility map which was done. It is a figure which shows the example of the table used for the setting of the movement state distribution of a moving object. It is a figure for demonstrating the calculation of a collision risk candidate. It is a figure explaining the grid and moving object particle | grains of a presence possibility map. It is a block diagram which shows the collision risk determination apparatus which concerns on 2nd Embodiment of this invention. It is a flowchart which shows operation | movement of the collision risk determination apparatus of FIG.

  Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the following description, the same or equivalent elements will be denoted by the same reference numerals, and redundant description will be omitted.

[First Embodiment]
First, the first embodiment will be described. FIG. 1 is a block diagram showing a collision risk degree judging device according to the first embodiment. The collision risk determination device 100 according to the present embodiment is mounted on a host vehicle 10 such as an automobile, and determines the collision risk of the host vehicle 10 against a moving object 50 such as a pedestrian. The collision risk determination apparatus 100 includes a motion sensor 1, a laser radar 2, a camera 3, a GPS 4, and an ECU (Electronic Control Unit) 5.

  The motion sensor 1 includes, for example, a vehicle speed sensor that measures the speed of the host vehicle 10, a gyro sensor that measures a yaw rate, or an acceleration sensor that measures acceleration. The motion sensor 1 detects the motion state (running state) of the host vehicle 10. The motion sensor 1 is connected to the ECU 5 and outputs the detected motion state data of the host vehicle 10 to the ECU 5.

  The laser radar 2 irradiates the surroundings (in this case, the front) of the host vehicle 10 while scanning the laser in one dimension (horizontal direction), and detects the two-dimensional position of the object irradiated with the laser by reflection of the laser. To do. The laser radar 2 is connected to the ECU 5 and outputs the detected position data of the object to the ECU 5.

  The camera 3 is composed of a small CCD camera or a CMOS camera, and photographs the surroundings of the host vehicle 10 (here, the front). The camera 3 is attached to, for example, the upper part of the front window of the host vehicle 10. The camera 3 is connected to the ECU 5 and outputs imaged image data such as a road condition ahead of the photographed image to the ECU 5. The GPS 4 detects the position of the host vehicle 10. The GPS 4 is connected to the ECU 5 and outputs the detected position data of the host vehicle 10 to the ECU 5.

  The ECU 5 includes, for example, a CPU, a ROM, a RAM, and the like. The ECU 5 includes a course candidate calculation unit 5a, an environment detection unit 5b, a map generation unit 5c, a moving object generation unit 5d, a position update unit 5e, and a distribution change unit 5f as functional configurations. The course candidate calculation unit 5a calculates a plurality of course candidates, that is, a plurality of paths that are likely to be taken by the host vehicle 10 in the future, based on the motion state of the host vehicle 10 detected by the motion sensor 1.

  FIG. 2 is a diagram for explaining an example of the calculation result of the course candidate calculation unit. For example, the course candidate calculation unit 5a calculates a plurality of course candidates S as shown in FIG. 2A when it is determined that the host vehicle 10 travels a curve from the vehicle speed, yaw rate, and acceleration of the host vehicle 10. To do. For example, when it is determined that the host vehicle 10 is to change lanes, a plurality of course candidates S as shown in FIG.

  The environment detection unit 5b refers to an electronic map stored in the map database (map DB) 5g, and based on outputs from the laser radar 2, the camera 3, and the GPS 4, a traveling environment (for example, a moving object) around the host vehicle 10 50 position, the state of the moving object 50, and other traveling environment conditions). The map generation unit 5c is based on the traveling environment detected by the environment detection unit 5b, and the presence possibility map is assigned with the existence possibility indicating the ease of the moving object 50 on the map of the surrounding area of the host vehicle 10. 11 (see FIG. 5).

  The moving object generation unit 5 d generates a plurality of moving object particles as data representing the moving object 50, and places (records) these moving object particles on the existence possibility map 11 as a position distribution of the moving object 50. At the same time, the moving object generation unit 5d gives each moving object particle a moving state corresponding to the moving state distribution of the moving object 50. The position update unit 5e moves each moving object particle based on the moving state of the moving object 50. The distribution changing unit 5 f changes the arrangement by eliminating and duplicating each moving object particle based on the existence possibility of the existence possibility map 11.

  Moreover, ECU5 has the risk candidate calculation part 5h, the risk determination part 5i, and the driving assistance part 5j as the functional structure. The risk candidate calculation unit 5 h calculates a collision risk candidate that becomes a collision risk candidate for each of the plurality of course candidates S on the existence possibility map 11. The risk determination unit 5i selects one route candidate S from among the plurality of route candidates S based on the plurality of collision risk candidates calculated by the risk candidate calculation unit 5h, and selects the one selected route candidate The collision risk candidate of S is determined as the collision risk.

  The travel support unit 5j supports the travel of the host vehicle 10 based on the route candidate S selected by the risk determination unit 5i and the determined collision risk. Here, the driving support unit 5j is brake control by controlling the brake device 12, steering control by controlling the steering device 13, and attention by controlling an HMI (Human Machine Interface) 14 such as a monitor and a speaker. Implement arousal.

  Next, the operation of the above-described collision risk determination device 100 will be described in detail with reference to the flowchart shown in FIG.

  In the collision risk determination device 100 of the present embodiment, first, output values from the motion sensor 1, the laser radar 2, the camera 3, and the GPS 4 are acquired (S1). Subsequently, based on the motion state of the host vehicle 10 detected by the motion sensor 1, the route candidate calculation unit 5a calculates a plurality of route candidates S in the host vehicle 10 (S2).

  On the other hand, based on the two-dimensional position of the object detected by the laser radar 2, the forward image captured by the camera 3, and the position of the host vehicle 10 detected by the GPS 4, the environment detection unit 5b travels around the host vehicle 10. The environment is detected (S3). Specifically, first, local map information representing a coordinate system viewed from the own vehicle 10 in a certain area in front of the own vehicle 10 is generated. The local map information is a grid map divided by a grid having a certain size. In each grid (block), the probability that a stationary object exists at that position is recorded, and in the initial state, for example, 0.5 (intermediate value) is recorded as the existence probability.

  Subsequently, based on the position of the object detected by the laser radar 2, the existence probability of the grid corresponding to the object position on the local map information is increased. At the same time, the existence probability of each grid existing on the straight line from the host vehicle 10 to the object position is reduced. In addition, an initial value of 0.5 is recorded as the existence probability in a grid in which information on the object position is not obtained due to reasons such as being hidden by another object (grid in the blind spot area).

  Thereby, even in a traveling environment in which many moving objects 50 exist, stationary objects such as roadside objects can be stably detected. Further, an invisible region that cannot be observed by the laser radar 2 can be detected as a grid with the existence probability remaining at the initial value (0.5). Further, a far region where the measurement points of the laser radar 2 are hardly obtained can be regarded as a blind spot region because it does not change from the initial value.

  Further, based on the position of the host vehicle 10 detected by the GPS 4, from the electronic map of the map database 5g, the area of the traveling section (lane, sidewalk, pedestrian crossing, etc.) and the section of the traveling path around the host vehicle 10, the sign display (signal , Pause, etc.), detect information about buildings. In addition, from the electronic map of the map database 5g, the surrounding area type (school zone, shopping street, residential area, etc.) of the vehicle 10 and road attributes (number of lanes, lane width, presence / absence of median) are detected.

  In addition, from the front image captured by the camera 3, using the learning type pattern recognition technology (for example, VM), the position and size of the moving object 50 existing in front, the type of the moving object 50 (for example, pedestrian, Two-wheeled vehicle, automobile, etc.), operating state (direction, gait, etc.) and moving state (speed, etc.) are detected. At the same time, the type (lane, sidewalk, pedestrian crossing, signal, temporary stop line, etc.) and area of various lane segments on the road ahead are detected from the front image.

  Subsequently, an area where a stationary object (guardrail, planting, building, parked vehicle, etc.) exists is detected from the local map information. Further, the height of each stationary object is detected from the front image photographed by the camera 3. In addition, by identifying the time difference of a plurality of local map information continuous in time series, the moving object that is moving and the moving object that is stationary are identified, and the area where the moving object 50 exists is specified and specified. The area where the moving object 50 is present may be associated with the type of the detected moving object 50.

  Subsequently, an existence possibility map is generated for each type of the moving object 50 based on the road segment area, stationary object area, road attribute, and the like detected in S3 (S4). For example, in a driving environment as shown in FIG. 4 (when the roadway is one lane on one side, the sidewalk 21 and the roadway 22 are separated by a curb 23, and the pedestrian crossing 24 is provided on the roadway 22), the sidewalk 21 is In a situation where there is a moving object M as a walking pedestrian, a presence possibility map 11 for the moving object M is generated as shown in FIG.

  In the existence possibility map 11, the existence possibility is given to the areas of the various road sections and the areas of the stationary objects. The existence possibility here represents the ease with which the moving object M exists with respect to the target region by 0.0 to 1.0. The existence possibility of the runway segment can be determined according to, for example, a combination of the runway segment and the road attribute, and is stored in advance as a table illustrated in FIG. 6, for example.

Further, the possibility of existence of the area of the stationary object is calculated according to the height h of the stationary object, and is calculated according to the following equation (1), for example.
Existence possibility in region of stationary object = 1.0−min (h, 1.0) (1)
However, min (a, b) is a function that returns a smaller one of a and b.

  In this existence possibility map 11, a unit movement cost is set for each grid. The unit movement cost is an index value required when the host vehicle 10 passes, and can be determined according to an area corresponding to the grid, and is stored in advance as a table. For example, the unit movement cost is set higher as the host vehicle 10 is less likely to move, and lower as the host vehicle 10 is easier to move.

  Next, based on the position distribution, moving state distribution, and existence possibility of the moving object 50, the moving object 50 on the existence possibility map 11 is obtained by the moving object generation unit 5d, the position update unit 5e, and the distribution change unit 5f. The future position distribution is arranged and predicted (S5). Specifically, first, the moving object generation unit 5d sets a region where the specified moving object 50 exists in the existence possibility map 11 as a particle generation candidate region and a blind spot region as a particle generation candidate region. . Then, for example, as shown in FIG. 7 (a), a plurality of moving object particles 31 are generated in each particle generation candidate region using a random number generator so that the total number of particles set in advance based on the processing capability of the ECU 5 is obtained. And place (record).

  Further, the moving object generation unit 5d sets the moving state distribution of the moving object 50 based on the detected operation state and moving state of the moving object 50, and sets the moving state of each moving object particle 31 based on the moving state distribution. decide. As the physical quantity used as the movement state, at least one of the direction of movement, speed, and acceleration is set.

  Here, since it is considered that the higher the speed of the moving object 50 is, the higher the straightness is and the lower the uncertainty of the movement is, so a dispersion / covariance table (see FIG. 8) corresponding to the average speed of the moving object M is used. Then, the distribution of the moving state of the moving object 50 is set. At this time, when the moving object 50 is a pedestrian and its motion state is staggering, the distribution of the moving state may be set so that the speed dispersion becomes large.

  It should be noted that the moving object particles 31 are also assigned with a label of the detection result of the type of the moving object 50 such as an automobile, a two-wheeled vehicle, and a pedestrian, and information on the moving state determined based on the set moving state distribution. be able to. Further, identification information for identifying the moving object 50 can also be assigned to the moving object particles 31. For example, the moving object particles 31 having the same identification information can be generated for the particle generation candidate region generated for one moving object 50.

  Subsequently, the position update unit 5 e moves the position of each moving object particle 31 based on the moving state given to the moving object particle 31. Thereby, the arrangement of each moving object particle 31 is updated. Subsequently, the distribution changing unit 5f performs the reselection of the moving object particle 31 by eliminating or duplicating the moving object particle 31 according to the existence possibility given to the area of the moving point, and the position distribution of the moving object 50 (The movement of the moving object 50 is restricted according to the ease of existence). At this time, the position distribution of the moving object 50 may be changed using the difference or ratio of the existence possibility given to the area before and after the movement. As described above, the future position distribution of the moving object 50 is arranged on the existence possibility map 11.

  Here, the details of the movement and change of the position of each moving object particle 31 will be exemplified.

  First, each moving object particle 31 transitions based on the moving state assigned to each moving object particle 31. And the moving object particle | grain 31 is annihilated with a high probability, so that the presence possibility of a transition destination is low. In addition, the moving object particle 31 is duplicated by the amount that has disappeared, and is arranged so as to overlap the position of the other moving object particle 31 that has not disappeared, or the position of the other moving object particle 31 that has not disappeared. It is placed at the peripheral position (position multiplied by random number). Thereby, the moving object particle 31 is newly produced | generated centering on the area | region with high possibility of existence. In addition, the above disappearance and replication are performed so that the total number of particles is constant.

  For example, when the pedestrian is walking in a stable manner and the speed dispersion is small, the moving object particle 31 disappears around a region where the possibility of existence of the transition destination is low, and the possibility of existence of the transition destination A moving object particle 31 is newly generated around a region where is high. As a result, as shown in FIG. 7B, the arrangement of the moving object particles 31 is changed, and it is indicated that the pedestrian crossing 24 is not crossed.

  On the other hand, if the pedestrian is swaying and scrambled and the velocity dispersion is large, the moving object particle 31 disappears around a region where the transition destination is less likely to exist, and the transition destination is more likely to exist. Moving object particles 31 are newly generated around the area. Thereby, as shown in FIG.7 (c), arrangement | positioning of the moving object particle | grains 31 is changed and it represents crossing the pedestrian crossing 24. FIG.

  Next, the travel cost and the motion cost are calculated by the risk candidate calculation unit 5h for each of the plurality of course candidates S obtained in S2 on the existence possibility map 11 in which the future position distribution is predicted (S6). . The movement cost is an index value obtained according to the route through which the route candidate S of the host vehicle 10 passes, and the value is larger as the route is more difficult to move (the value is smaller as the route is more likely to move). )

  Here, for each course candidate S, as shown in FIG. 9, a plurality of grids G on which the course candidate S passes on the existence possibility map 11, that is, a plurality of grids G on which the course candidates S overlap are selected. Then, the movement cost is calculated by summing the unit movement costs of each of the plurality of grids G.

  On the other hand, the exercise cost is an index value obtained in accordance with a change in the motion of the host vehicle 10 when the host vehicle 10 moves along the route candidate S, and the value is larger as the route is largely changed (the route remains unchanged). The smaller the value). Here, for example, the time integral value of the absolute value of the yaw rate change in the host vehicle 10, the time integral value of the absolute value of the steering change, or the time integral value of the absolute value of the acceleration in the front-rear direction and the left-right direction is calculated as the motion cost.

  Next, on the existence possibility map 11 in which the future position distribution is predicted by the risk candidate calculation unit 5h, each of the plurality of course candidates S obtained in S2 is overlapped with the moving object particle 31 (overlap frequency). ) To calculate the collision probability (S7). The collision probability is obtained according to the future position distribution and the course candidate S.

Here, on the existence possibility map 11, the total number of grids G through which the path candidate S passes is M _total , the number of grids G in which the moving object particles 31 are present is m _col, and the total number of moving object particles 31 is N _{When the total} is set to n and the number of moving object particles 31 in the grid G through which the route candidate S passes is n _col , the collision probability is calculated according to the following equation (2).
Collision probability = n _col / N _total · m _col / M _total (2)

Note that the number n _col of the moving object particles 31 in the grid G through which the route candidate S passes assumes a collision with the host vehicle 10 on the basis of the moving object 50 side. Therefore, “n _col / N _total ” on the right side of the above formula (2) means the possibility of collision as viewed from the moving object 50 side. On the other hand, the number m _col of the grid G in which the moving object particles 31 are present assumes a collision with the moving object 50 on the basis of the own vehicle 10 side. Therefore, “m _col / M _total ” on the right side of the above equation (2) means the possibility of collision as viewed from the host vehicle 10 side.

Incidentally, a plurality of moving object particles 31 may exist in one grid G. For example, in the example shown in FIG. 10, the number n _col of moving object particles 31 in the grid G is “3”, and the number m _col of the grid G in which the moving object particles 31 are present is “1”.

Next, the risk candidate calculation unit 5h calculates a collision risk candidate for each of the plurality of course candidates S based on the movement cost, the exercise cost, and the collision probability according to the following expression (3) (S8). The collision risk level is an index value that represents the risk of collision between the host vehicle 10 and the moving object 50. Further, K1 to K3 in the following equation (3) are weighting factors that are corrected so as not to give the driver a sense of discomfort, and are obtained from, for example, experience or ergonomics.
Collision risk candidate = K1 × movement cost + K2 × movement cost +
K3 x collision probability (3)

  Next, the route candidate S having the lowest collision risk candidate is selected from the plurality of route candidates S (S9). Subsequently, the collision risk candidate of the selected route candidate S is determined as the collision risk (S10). Then, based on the collision risk determined to be the selected course candidate S by the travel support unit 5j, the travel support of the host vehicle 10 is performed by controlling the brake device 12, the steering device 13 and the HMI 14 (S11). . For example, while assisting driving with the selected course candidate S, when the determined collision risk exceeds the battle value, brake control, steering control, or alerting is performed to avoid a collision of the host vehicle 10. .

  As described above, in the present embodiment, “existence of moving object 50” determined by environmental factors and “uncertainty of movement of moving object 50 (moving state distribution)” determined by the state of moving object 50 are set. By moving the moving object 50 on the possibility map 11, it is possible to perform future position prediction in consideration of various combinations of traffic environments and states of the moving object 50. Therefore, the position prediction accuracy of the moving object 50 can be increased.

  In addition, for each of the plurality of course candidates S, a collision risk candidate that considers not only the path planning index such as movement cost and movement cost but also the collision probability is calculated, and the course candidate S is selected based on the collision risk candidate. Then, the collision risk candidate in the selected course candidate S is determined as the collision risk. In this way, the collision risk is not determined from only the collision probability of one course candidate S, but the collision risk candidates of a plurality of course candidates S are calculated and the collision risk is determined from the collision risk candidates. Even if the course candidate S and thus the collision probability vary due to an adverse effect such as sensor noise, it is possible to suppress the adverse influence of the variation on the determination of the collision risk.

  Therefore, according to the present embodiment, it is possible to accurately determine the collision risk. Further, as described above, since the course candidate S having the lowest collision risk candidate is selected from among the plurality of course candidates S (see S9 above), the driver has a tendency to select a course having a low collision risk. This can be suitably reflected, and driving support that matches the driver's needs is possible.

  In the present embodiment, as described above, traveling of the host vehicle 10 is supported based on the determined collision risk. Therefore, it is possible to accurately support the traveling of the host vehicle 10 according to the determination of the collision risk with high accuracy, and as a result, it is possible to suppress variations in driving support.

[Second Embodiment]
Next, a second embodiment will be described. In the description of the present embodiment, differences from the first embodiment will be mainly described.

  FIG. 11 is a block diagram showing a collision risk determination device according to the second embodiment. As shown in FIG. 11, the collision risk determination device 200 of the present embodiment is different from the collision risk determination device 100 in that the ECU 5 further includes a course candidate selection unit 5k.

  The course candidate selection unit 5k obtains an evaluation function based on the movement cost and the movement cost for each of the plurality of course candidates S calculated by the course candidate calculation unit 5a, and selects a part of the plurality of course candidates S based on the evaluation function. select. Further, the risk candidate calculation unit 5h according to the present embodiment calculates a collision risk candidate for a part of the route candidate S selected by the route candidate selection unit 5k.

In this embodiment, as shown in the flowchart of FIG. 12, after calculating the movement cost and the exercise cost of the plurality of course candidates S (after S6), the course candidate selection unit 5k uses the plurality of course candidates. N course candidates S are selected from S (S21). Specifically, for each of the plurality of course candidates S, the evaluation function is calculated according to the following expression (4) based on the calculated movement cost and movement cost. Then, among the plurality of course candidates S, N course candidates S are selected in ascending order of the evaluation function. Note that N is an integer of 2 or more, and P> N, where P is the number of route candidates S before selection.
Evaluation function = K1 × movement cost + K2 × movement cost (4)

  Subsequently, the risk candidate calculation unit 5h calculates a collision probability for each of the N course candidates S selected in S21 on the existence possibility map 11 in which the future position distribution is predicted (S22). Subsequently, the risk candidate calculation unit 5h calculates a collision risk candidate for each of the N course candidates S based on the movement cost, the exercise cost, and the collision probability (S23). Then, the process proceeds to the above 9 process.

  As described above, also in the present embodiment, the above-described effect of accurately determining the collision risk is achieved. Furthermore, in this embodiment, as described above, N course candidates are selected in advance from a plurality of course candidates S based on the evaluation function, and collision risk degree candidates are calculated for the selected N course candidates S. is doing. As a result, the course candidates S to be calculated for the risk candidate can be narrowed down in advance by the movement cost and the motion cost, and the calculation time for the collision risk candidate can be shortened. As a result, since it takes a relatively long time to calculate the collision risk candidate (particularly the collision probability), it is possible to reduce the calculation time of the entire apparatus.

  The preferred embodiments of the present invention have been described above. However, the present invention is not limited to the above-described embodiments. The present invention is modified without departing from the scope described in the claims or applied to others. It may be.

  For example, in the above-described embodiment, the collision risk determination device is mounted on the host vehicle 10, but may be installed on the road side (for example, a main intersection). Moreover, although the possibility of existence represents the ease with which the moving object exists, it may represent the difficulty with which the moving object does not exist.

  In the above embodiment, the laser radar 2 and the camera 3 are used, but instead of or in addition to this, a stereo camera, an apparatus that detects the position of an object by scanning an electromagnetic wave such as a millimeter wave forward, etc. May be used. Moreover, in the said embodiment, although the brake device 12, the steering device 13, and HMI14 are provided, at least 1 of these may be provided.

  Moreover, in the said embodiment, although the motion of the own vehicle 10 was estimated using the detection result of the motion sensor 1, it is not limited to this, For example, the own vehicle 10 is based on the time difference of the detection result of the laser radar 2. May be estimated. Further, for example, the motion of the host vehicle 10 may be estimated using the detection result of the position of the host vehicle 10 by the GPS 4. Moreover, you may estimate the motion of the own vehicle 10 combining the detection result of the motion sensor 1, the laser radar 2, and GPS4.

  In addition to obtaining the moving state distribution based on the detected operating state or moving state of the moving object 50, the moving state distribution of the moving object 50 may be directly detected, or predetermined for each type of moving object 50. Alternatively, a moving state distribution may be used. Furthermore, by using the observation result of the moving object 50 (operation state such as direction and gait), setting the dispersion and direction of the moving state, and using an average speed predetermined for each type of the moving object 50 The movement state distribution may be obtained.

  In the above embodiment, the total number of the moving object particles 31 is a constant value, but may be variable. When the total number of moving object particles 31 is constant, calculation efficiency independent of the complexity of the situation of the driving environment can be secured, while when variable, the accuracy of collision risk determination can be adjusted according to the ability of the ECU 5 to process. .

  In the above, the laser radar 2, the camera 3, the GPS 4, the environment detection unit 5b, and the map database 5g constitute a moving object detection unit. The moving object generation unit 5d, the position update unit 5e, and the distribution change unit 5f constitute a future position distribution prediction unit. The motion sensor 1 and the course candidate computation unit 5a constitute a course candidate computation unit.

  DESCRIPTION OF SYMBOLS 1 ... Motion sensor (course candidate calculation part), 2 ... Laser radar (moving object detection part), 3 ... Camera (moving object detection part), 4 ... GPS (moving object detection part), 5a ... Course candidate calculation part, 5b ... environment detection part (moving object detection part), 5c ... map generation part, 5d ... moving object generation part (future position distribution prediction part), 5e ... position update part (future position distribution prediction part), 5f ... distribution change part ( Future position distribution prediction unit), 5g ... map database (moving object detection unit), 5h ... risk level candidate calculation unit, 5i ... risk level determination unit, 5j ... travel support unit, 5k ... course candidate selection unit, 10 ... own vehicle , 11 ... existence possibility map, 31 ... moving object particles, 50 ... moving object, 100, 200 ... collision risk degree judging device, G ... grid.

Claims (8)

A collision risk determination device for determining a collision risk of a host vehicle against a moving object,
A moving object detection unit for detecting a position distribution and a moving state distribution of the moving object;
A map generation unit that generates an existence possibility map to which an existence possibility indicating the ease of existence or difficulty of existence of the moving object is given on a map indicating a surrounding area of the host vehicle;
A future position distribution prediction unit that predicts a future position distribution of the moving object on the existence possibility map based on the position distribution, the movement state distribution, and the existence possibility of the moving object;
A route candidate calculation unit that calculates a plurality of future route candidates of the host vehicle based on the motion state of the host vehicle;
On the existence possibility map where the future position distribution of the moving object is predicted, a risk candidate calculation unit for calculating a collision risk candidate for each of the plurality of course candidates;
Based on the calculated plurality of collision risk candidates, one of the plurality of route candidates is selected, and the collision risk candidate of the selected one of the route candidates is determined as the collision risk. A risk determination unit,
The risk candidate calculation unit includes:
The movement cost obtained according to the path through which the route candidate passes, the movement cost obtained according to the movement change of the own vehicle when the own vehicle moves along the route candidate, the future position distribution, and A collision risk determination apparatus, wherein the collision risk candidate is calculated based on a collision probability obtained according to the course candidate.   2. The collision risk determination device according to claim 1, further comprising a travel support unit that supports the travel of the host vehicle based on the collision risk determined by the risk determination unit.   The collision risk determination device according to claim 1, wherein the risk determination unit selects a route candidate having the lowest collision risk candidate from the plurality of route candidates. A path candidate that obtains an evaluation function based on the movement cost and the movement cost for each of the plurality of path candidates calculated by the path candidate calculation unit, and selects a part of the plurality of path candidates based on the evaluation function A selection unit,
The risk level candidate calculation unit calculates the collision risk level candidate for a part of the plurality of path candidates selected by the path candidate selection unit. The collision risk determination device described.   The future position distribution prediction unit records the position distribution on the existence possibility map, moves the position distribution based on the movement state distribution, and calculates the position distribution based on the existence possibility. The collision risk determination device according to claim 1, wherein the future position distribution is set by changing. The existence possibility map is a grid map, and each of the plurality of grids is set with a unit movement cost required when the host vehicle passes,
The collision risk according to any one of claims 1 to 5, wherein the movement cost is a sum of unit movement costs of a plurality of grids through which the route candidate passes on the existence possibility map. Degree determination device.   The movement cost is a time integral value of an absolute value of a yaw rate change in the host vehicle, a time integral value of an absolute value of a steering change, or a time integral value of an absolute value of acceleration. The collision risk determination device according to any one of claims 6 to 6. The future position distribution of the moving object is represented by a plurality of moving object particles,
The existence possibility map is a grid map;
In the collision probability map, the total number of grids through which the path candidate passes is M _total , the number of grids in which the moving object particles are present is m _col, and the total number of moving object particles is N _{When the total} number of moving object particles in the grid through which the path candidate passes is n _col ,
n _col / N _total · m _col / M _total
The collision risk determination device according to any one of claims 1 to 7, wherein

Images