KR101949609B1 - Method and system for updating occupancy map based on super ray - Google Patents

Abstract

A representative ray based occupancy map updating method and system are disclosed. A method of updating an occupancy map based on a grid and an octree, the method comprising: generating a mapping line based on point clouds data obtained from a sensor; Identifying a cell traversing the same cell for a plurality of cells constituting the occupancy map based on the mapping line to generate a super ray; and generating a plurality of Updating the occupancy map by updating the cell through which the representative ray corresponding to some of the rays associated with the cells passes.

Description

[0001] METHOD AND SYSTEM FOR UPDATING OCCUPANCY MAP BASED ON SUPER RAY [0002]

Embodiments of the present invention relate to techniques for efficiently updating point clouds with an occupancy map representation such as an octree and more particularly to techniques for updating point clouds with an edge or grid point Relates to a technique for updating an occupancy map using a mapping line based on a property of branching an access pattern of a map update.

Many robotic systems use a variety of sensor data to understand their environment. Point clouds are known to be effective sensor data representing the environment around a robot and are easily captured in recent consumer sensor depth sensors (eg, Kinect and Xtion). Point cloud data is represented by a large number of points representing geometric information in a high resolution environment, but with varying levels of sensor noise. In application programs such as path planning and SLAM, it is difficult to directly use the point cloud data because of the amount of generated data itself and the amount of noise. In order to solve the problems caused by the noise and the amount of data itself, a technique of representing an occupancy map based on an octree or a grid has been proposed, and the following non-patent documents [1] Kai M Wurm , Armin Hornung , Maren Bennewitz , Cyrill Stachniss , and Wolfram Burgard , " Octomap : A probabilistic, flexible, and compact 3d map representation for robotic systems ", in Proc . of the ICRA 2010 workshop on best practice in 3D perception and modeling for mobile manipulation, 2010, vol. In order to take into account the uncertainty of the point group, various occupations such as octree and grid have been proposed in [1] , [2] H Moravec, "Robot spatial perceptionby stereoscopic vision and 3d evidence grids" A method of expressing point cloud data with a map representation is described. In the case of a grid map, a huge amount of memory is required to handle a large external environment, and a high resolution is required for accurate map representation.

Recent applications use map representations to achieve higher performance. For example, instead of accessing large and uncertain point cloud data, the path planning algorithm may use an occupancy map representing free states or occupancy states in each cell to efficiently locate the non-collision path use.

However, building an occupancy map outside of the point cloud data takes considerable computational overhead, especially when using the map at high resolution to get high precision, it takes a lot of time to traverse and update the map. And, lowering the resolution of the map increases performance, but the accuracy is lowered, causing serious problems with various robotic tasks such as motion planning.

Therefore, there is a demand for an occupancy map updating technique for expressing point cloud data for achieving high performance without deteriorating the accuracy of the map. Korean Patent Laid-Open Publication No. 10-2011-0092592 relates to an apparatus and method for generating a three-dimensional map based on an octree and a grid, which detects a plurality of points having three-dimensional position information and detects a space including a plurality of detected points This paper presents a technique for generating a three-dimensional map by representing an octree and a grid-based occupancy map.

 [1] Kai M Wurm, Armin Hornung, Maren Bennewitz, Cyrill Stachniss, and Wolfram Burgard, "Octomap: A probabilistic, flexible and compact 3d map representation for robotic systems", in Proc. of the ICRA 2010 workshop on best practice in 3D perception and modeling for mobile manipulation, 2010, vol. 2. [2] H Moravec, "Robot spatial perceptionby stereoscopic vision and 3d evidence grids", Perception, (September), 1996. [3] Alberto Elfes, " Using occupancy grids for mobile robot perception and navigation ", Computer, vol. 22, no. 6, pp. 46-57, 1989. [4] Hans P Moravec and Alberto Elfes, "High resolution maps from wide angle sonar", in Robotics and Automation. Proceedings. 1985 IEEE International Conference on. IEEE, 1985, vol. 2, pp. 116-121. [5] Kai M Wurm, Armin Hornung, Maren Bennewitz, Cyrill Stachniss, and Wolfram Burgard, "Octomap: A Probabilistic, Flexible, and Compact 3d Map Representation for Robotic Systems", in Proc. of the ICRA 2010 workshop on best practice in 3D perception and modeling for mobile manipulation, 2010, vol. 2.

SUMMARY OF THE INVENTION The present invention is directed to a technique for updating an occupancy map based on a super ray for a set of point cloud data. That is, the present invention relates to a technique of updating together point cloud data that traverses the same cell set as the representative ray by updating the point cloud data corresponding to the representative ray of the map.

A method of updating an occupancy map based on a grid and an octree, the method comprising: generating a mapping line based on point clouds data obtained from a sensor; Identifying a cell traversing the same cell for a plurality of cells constituting the occupancy map based on the mapping line to generate a super ray; and generating a plurality of Updating the occupancy map by updating the cell through which the representative ray corresponding to some of the rays associated with the cells passes.

According to an aspect of the present invention, the step of generating the mapping line may include a step of converting the coordinate system of the point cloud data into a world coordinate system corresponding to the occupancy map, Generating a virtual cell box including all the point cloud data corresponding to any one of a plurality of cells in the occupied map, and generating a virtual cell box including overlapping data between the cell box and the slice associated with the cell box based on the sensor origin on the occupied map. and generating an overlapping line segment as the mapping line.

According to another aspect, the step of generating the overlapping line segment with the mapping line includes determining a plurality of intersection points connecting the both end points of the cell box from the sensor origin on a first slice including the sensor origin Determining a grid point located between the determined intersection points, and dividing the initial frustum into a plurality of seed frustums based on the lattice points.

According to another aspect, the step of generating the mapping line may include generating a mapping line based on a point on the mapping line connecting the mapping line through the grid point from the sensor origin on the occupied map, May be divided into a plurality of segment segments.

According to another aspect of the present invention, the step of generating the super ray includes: determining one of the point cloud data for each segment segment with respect to the point cloud data located between the plurality of segment segments that define the mapping line; And determining the ray of light projected with the point cloud data determined from the sensor origin as the super ray.

According to another aspect of the present invention, the step of generating the super ray includes the steps of: calculating, for each segment segment, the number of point cloud data located in each of the plurality of segment segments; To the weight of the representative light beam.

According to another aspect, the method may further include storing the point cloud data corresponding to the representative ray instead of the plurality of point cloud data corresponding to the same segment period for updating the occupancy map.

According to another aspect, generating the mapping line includes extending the 2D coordinate-based representative ray based on the 3D coordinate system by generating the mapping line for each of the XY plane, the YZ plane, and the ZX plane can do.

According to another aspect, the step of generating the representative ray may generate a representative ray by mapping rays by mapping planes corresponding to the XY plane, the YZ plane, and the ZX plane, which are three mapping planes in the 2D coordinate system have.

According to another aspect of the present invention, the step of generating the representative light ray includes the steps of projecting the edges existing in the initial frustum in the two-dimensional coordinate system to a mapping plane of the three- A plurality of representative regions can be generated for each region by grouping the rays mapped to the regions.

According to another aspect, the rays corresponding to each of the point cloud data belonging to the segment region corresponding to the representative ray can traverse the same cells on the occupation map.

A system for updating an occupancy map based on a grid and an octree, the system comprising: a mapping line generation unit for generating a mapping line based on point clouds data acquired from a sensor; A representative ray generator for generating a super ray by identifying a cell traversing the same cell for a plurality of cells constituting the occupancy map based on the generated mapping line, And a map updater that updates the occupancy map by updating a cell through which the representative ray corresponding to a part of rays associated with the plurality of cells constituting the map passes.

According to an aspect of the present invention, the mapping line generator converts the coordinate system of the point cloud data to a world coordinate system corresponding to the occupancy map, and converts the coordinate system of the plurality of cells constituting the occupancy map, And generates a line segment overlapping between the cell box and the slice associated with the cell box based on the sensor origin on the occupation map, Can be generated as the mapping line.

According to another aspect of the present invention, the mapping line generator determines a plurality of intersection points connecting the both end points of the cell box from the sensor origin on a first slice including the sensor origin, grid point, and divide the initial frustum into a plurality of segments based on the grid points.

According to another aspect of the present invention, the mapping line generation unit generates a mapping line based on a point on a mapping line connecting a mapping line passing through a grid point from a sensor origin on an occupied map, It can be divided into segments.

According to another aspect of the present invention, the representative light generation unit may determine one of the point cloud data for each segment segment with respect to the point cloud data located in each of the plurality of segment segments that define the mapping line, The light beam projected from the data can be determined as the super ray.

According to another aspect of the present invention, the representative light generating unit may calculate the number of point cloud data located in each of the plurality of segment intervals, for each segment segment, and calculate the number of the point cloud data calculated based on the determined weight of the representative light ray Can be set.

According to another aspect, the image processing apparatus may further include a storage unit for storing the point cloud data corresponding to the representative ray in place of the plurality of point cloud data corresponding to the same segment period for updating the occupancy map.

According to another aspect, the mapping line generation unit may extend the 2D coordinate system based on the 3D coordinate system by generating the mapping line for each of the X-Y plane, the Y-Z plane, and the Z-X plane.

According to another aspect, the representative ray expanded on the basis of the 3D coordinate system may be a representative ray generated by mapping rays according to mapping planes corresponding to XY plane, YZ plane, and ZX plane, which are three mapping planes in the 2D coordinate system, Lt; / RTI >

According to another aspect of the present invention, the representative light generating unit may include a plurality of divided light sources for projecting the corners existing in the initial frustum in a two-dimensional coordinate system to a mapping plane of a three-dimensional coordinate system, A representative ray can be generated for each region by grouping the rays mapped to each region on the regions.

According to another aspect, the rays corresponding to each of the point cloud data belonging to the segment region corresponding to the representative ray can traverse the same cells on the occupation map.

A computer program stored on a recording medium for executing a method of updating an occupancy map based on a grid and an octree in combination with a computer implemented electronic device, Includes the steps of generating a mapping line based on point cloud data acquired from a sensor, passing a same cell through a plurality of cells constituting the occupancy map based on the generated mapping line Identifying a traverse cell to generate a super ray; and updating the cell through which the representative ray corresponding to a part of the rays related to the plurality of cells constituting the occupancy map is updated, And updating the occupancy map.

The present invention can update the occupancy map efficiently by updating the point cloud data corresponding to the representative ray (super ray) together with updating the point cloud data that traverses the same cell set as the representative ray. That is, the traversing time and the update time can be reduced to shorten the update time.

Figure 1 is an illustration of an octree map representation of four different rays having different end points in one embodiment of the present invention.
2 is a block diagram illustrating an internal configuration of an occupancy map updating system in an embodiment of the present invention.
3 is a flowchart illustrating an occupancy map updating method in an embodiment of the present invention.
Figure 4 is a diagram provided to illustrate mapping line generation in one embodiment of the present invention.
Figure 5 is a diagram provided to illustrate the operation of generating a representative ray in one embodiment of the present invention.
FIG. 6 is a diagram provided for explaining an operation of expanding the operation of generating a representative ray in 3D, according to an embodiment of the present invention. FIG.
Figure 7 is a graph illustrating the average FPS tested in an octomap and grid map, in one embodiment of the present invention.
Figure 8 is a graph illustrating the number of cells accessed in an octomap and grid map tested in one embodiment of the present invention.
9 is a diagram showing a mapping plane used in 3D representation of an operation for generating representative rays in one embodiment of the present invention.
10 is a view for explaining an operation of bundling rays belonging to the same area among regions included in a mapping plane to generate a representative ray according to an embodiment of the present invention.
11 is a diagram showing the determination of the dendritic light using the mapping single plane and the determination of the representative ray using the mapping line in three three-dimensional planes in accordance with an embodiment of the present invention.
Figure 12 is a diagram provided to illustrate why it is more efficient to generate a representative ray in each dimension in accordance with one embodiment of the present invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

The embodiments relate to techniques for updating an occupancy map based on a new concept, super ray. Particularly, in updating the point cloud data on a grid-based occupancy map, the embodiments of the present invention are not limited to updating the occupancy map by individually processing spaces (i.e., cells) passing through all the rays, To a technique for updating an occupancy map using a super ray of rays. That is, by reducing the number of accesses to the occupied map space for updating the point cloud data, updating the occupancy map by performing update using the representative ray while maintaining the expression accuracy of the occupancy map generated by the existing method is improved Lt; / RTI >

In the present embodiments, the 'super ray' may represent rays that always represent rays passing through the same space (i.e., the same cell) on a grid-based occupancy map. That is, the representative ray may be used to update the occupancy map at the same time, by bundling the points (pints) that update the same space (i.e., the same cell) on the occupancy map. In other words, it can be used to update the same point as the set of cells accessed during the map update process. For example, when the point cloud data corresponding to the representative ray is updated, the point cloud data (i.e., the same cell cluster) corresponding to the rays passing through the same cell as the cell through which the representative ray passes (i.e., the lattice space) have.

In the present embodiments, 'point clouds' may represent sensor data captured or sensed by a sensor or laser range finder, and point cloud data itself may be used to represent the occupancy map . At this time, the point cloud data may be acquired from the sensor every frame. When 'point cloud' is obtained from the sensor, that is, when the point group data is reported by the sensor, the space between the sensor origin and the grid point corresponding to the point cloud data is emptied . Accordingly, a representative ray is generated by connecting a ray to a lattice point corresponding to the point cloud data reported from the sensor starting from the sensor origin, and searching the map to the end point of connecting the ray, and based on the representative ray The occupancy map can be updated.

In these embodiments, the occupancy map update system may represent a fixed terminal implemented by a computer or an electronic device such as a mobile terminal. For example, the electronic device may include a smart phone, a mobile phone, a navigation device, a computer, a notebook, a terminal for digital broadcasting, a personal digital assistant (PDA), a portable multimedia player (PMP)

In these embodiments, a " mapping line " can be used to construct a representative ray based on the input point, i.e., the point cloud data obtained from the sensor. For example, a mapping line may be used to update a two-dimensional map, and may be used to construct a representative ray by extending it to a three-dimensional map such as a grid and an octree.

In the present embodiments, a 'frustum' can represent a portion between two planes when cutting a solid into two flat planes. In addition, the update can be performed by adding a weight to the number of points belonging to the occupancy probability that the representative ray updates.

In these embodiments, " traversing " means a process of searching through a cell and arriving at an end point until a specific ray arrives at the end point from the sensor origin on an occupancy map divided into a plurality of cells can do.

은 아래의 수학식 1과 같이 표현될 수 있다.In the present embodiments, the 'ray' may have two types of state information: an occupancy state for a space and a free state. Since the sensor reports some objects at a specific point that is sensed, the end point of the ray can be in an occupied state. At this time, the other spaces through which the rays pass can be in a free state. Such state information such as free state or occupancy state can be used as important information in an application program such as motion planning. Occupancy maps may be expressed based on the information collected from the sensors, wherein the data collected by the sensors may include various levels of noise. Occupancy probability can be used to account for such noise, and occupancy probability

Can be expressed as Equation (1) below.

[Equation 1]

은 셀의 점유 상태 n, 초기 시간 1(step 1)부터 현재 시간 t(step t) 동안 센싱된 센서 측정값

및 베이즈 법칙(Bayes rule)에 기초하여 계산될 수 있다. 위의 비특허 문헌 [3] Alberto Elfes , "Using occupancy grids for mobile robot perception and navigation", Computer, vol. 22, no.6, pp. 46-57, 1989.에서는 베이즈 법을 이용하여 점유 확률을 모델링하는 방법을 제시하고 있다.In Equation (1), the occupancy probability

Is the occupied state n of the cell, sensor measured value from the initial time 1 (step 1) to the current time t (step t)

And Bayes rule. ≪ / RTI > [3] Alberto Elfes , " Using occupancy grids for mobile robot perception and navigation ", Computer, vol. 22, no. 6, pp. 46-57, 1989. In this paper, we propose a method of modeling occupancy probability using Bayes' method.

The above Equation (1) is derived from the above non-patent document [4] Hans P Moravec and Alberto Elfes , " High resolution maps from wide angle sonar ", in Robotics and Automation. Proceedings. 1985 IEEE International Conference on. IEEE, 1985, vol. 2, pp. 116-121. Can be expressed as Equation (2) below using the log-odd notation given in Equation (2).

&Quot; (2) "

은 역 센서 모델을 나타내는 것으로서, 아래의 수학식 3과 같이 정의될 수 있다.In Equation (2)

Represents an inverse sensor model, and can be defined as Equation (3) below.

&Quot; (3) "

Over-confidence problems in which new input data that conflicts with the current state of a cell can not change state immediately, if the cell has a cumulative occupancy probability over a predefined period of time, can occur frequently in a dynamic environment have. At this time, the overconfidence can be solved by using the clamping policy which limits the occupancy probability of the cell based on the boundary between the minimum state and the maximum state. That is, the state of a cell bounded by one of the two boundaries can be considered completely free, or fully occupied by a high occupancy probability.

Figure 1 is an illustration of an octree map representation of four different rays having different end points in one embodiment of the present invention.

1, it can be seen that although the four different rays traverse the same cells on the octree map, each ray has a different end point. In this case, if the point cloud data corresponding to each of the rays are updated one by one, the performance may be deteriorated due to redundant operations when traversing the same cell. Accordingly, any one of the plurality of light rays 110 passing through the same cells can be generated as a representative light ray 101 to update the map. By updating the map using the representative light ray, You can improve the performance of map updates by eliminating calculations.

FIG. 2 is a block diagram illustrating an internal configuration of an occupancy map updating system in an embodiment of the present invention, and FIG. 3 is a flowchart illustrating an occupancy map updating method in an embodiment of the present invention.

The occupancy map updating system 200 according to the present embodiment may include a sensor 210, a mapping line generator 220, a representative light generator 230, a map updater 240 and a storage 250 have. The sensor 210, the mapping line generator 220, the representative light generator 230, the map updater 240 and the storage 250 may be configured to perform the steps 310 through 340 of FIG. . In the present embodiments, an occupation map for point cloud data is represented on the basis of a uniform grid. However, this corresponds to the embodiment, and the size of the grid may be uneven. That is, the sizes of the lattice spaces constituting the occupation map (that is, the respective cells) may be different from each other. The mapping line generation unit 220, the representative light generation unit 230, the map update unit 240, and the storage unit 250 are configured to process basic computer arithmetic, logic, and input / output operations, Lt; / RTI > processor. For example, the processor may be configured to execute instructions received in accordance with program code stored in a recording device, such as a memory.

The sensor 210 senses objects constituting a surrounding environment to be sensed and provides the sensed data, that is, the point cloud data, to the occupancy map updating system 200. Here, the sensor 210 may be included in the occupancy map updating system 200 or separately from the occupancy map updating system 200. For example, the sensor 210 may provide the occupancy map update system 200 with point cloud data sensed via a near field wireless communication module or a wired communication module.

For example, the sensor 210 may sense a robot or an environment surrounding the autonomous vehicle, and may provide point cloud data to the mapping line generator 220 for generating occupancy maps. At this time, the sensor 210 may provide the point cloud data sensed for a predetermined period of time, and may transmit the point cloud data sensed for each frame or every predetermined time unit to the mapping line generator 220 .

In step 310, the mapping line generation unit 220 may generate a mapping line based on the point cloud data acquired from the sensor 210. [ At this time, since the coordinate system of the point cloud data acquired from the sensor 210 is defined in the sensor coordinate system, it is necessary to convert the coordinate system into a world coordinate system for representation as an occupancy map.

In step 311, the mapping line generation unit 220 may convert the coordinate system of the point cloud data acquired from the sensor 210 into a world coordinate system. Here, assuming that the position and direction of the sensor are already known on the world coordinate system, the coordinate system of the point cloud data can be converted from the sensor coordinate system to the world coordinate system. The update of the occupancy map can then be processed using the converted point cloud data based on the world coordinate system.

In step 312, the mapping line generation unit 220 generates a virtual cell box including all the point cloud data corresponding to any one of the plurality of cells constituting the occupation map, with respect to the point cloud data of which the coordinate system has been converted can do.

In step 313, a mapping line may be created that overlaps the cell box created based on the sensor origin on the occupation map and the slice associated with the cell box. Here, the detailed operation for generating the mapping line will be described later with reference to FIG.

In step 320, when a mapping line is generated, the representative light generating unit 230 identifies a cell that traverses the same cell on a plurality of cells constituting the occupation map based on the generated mapping line, Can be generated.

In step 321, the representative ray generating unit 230 may determine any one of the point cloud data for each segment segment with respect to the point cloud data located in each of the plurality of segment segments for identifying the mapping line.

In step 322, the representative light generating unit 230 may determine a super ray as a light ray projected (or projected) from the sensor origin point with the determined point cloud data. At this time, the representative light generating unit 230 may calculate the number of the pointing data located in the plurality of segment intervals, for each segment segment, and set the number of the point cloud data to the weight of the representative light. Here, the detailed operation for generating the representative ray will be described later with reference to FIG.

In step 330, the storage unit 250 may store the point cloud data corresponding to the representative ray in place of the plurality of point cloud data corresponding to the same segment period, to update the occupancy map.

In step 340, the map updating unit 240 can update the occupancy map by updating the cell through which the representative ray corresponding to a part of the plurality of cells constituting the occupancy map passes. That is, instead of updating the cells through which all rays pass, instead of performing redundant updating for the same cell, a cell in which a representative ray passing through the same cell passes among a plurality of rays is updated to remove redundant operations, Can be improved.

Figure 4 is a diagram provided to illustrate mapping line generation in one embodiment of the present invention.

Referring to FIG. 4, the occupancy map 410 may be divided into a plurality of cells (i.e., a plurality of lattice spaces). 4, it is assumed that each of the lattice spaces constituting the occupation map 410 has a uniform size.

The mapping line generation unit 220 generates a mapping map including all the point cloud data belonging to the cell C from the sensor origin point 401 to any one of the plurality of cells constituting the occupation map 410 A virtual cell box 402 can be created. The mapping line generation unit 220 may configure an initial frustum from the sensor origin point 401 to the cell box 420.

At this time, the initial frustum can be divided into a plurality of pieces in the processing direction 412 for the mapping line generation, and can be processed in units of slices. Either x-axis, y-axis, or z-axis can be selected for processing in slice units, and the selected axis direction can be the processing direction. In FIG. 4, mapping lines are generated by using the x-axis direction as a processing direction. However, this corresponds to the implementation, and the y-axis or z-axis direction may be the processing direction. In Fig. 4, the slice may comprise a plurality of cells. For example, slice 1 421 may include four cells, and slice 2 422 may include cells included in slice 1 and four cells neighboring in the processing direction. FIG. 4 shows an example in which occupancy maps include four slices.

At this time, in order to identify the point cloud data mapped to the same representative ray, a line segment overlapping between the slice including the cell C and the cell C (i.e., the slice 3 423) may be generated as a mapping line.

For example, the mapping line generation unit 220 generates a mapping line generation unit 220 for generating a plurality of connection lines connecting the both end points 403 and 404 of the cell box 402 from the sensor origin point 401 on the slice 1 4110 including the sensor origin point 401. The intersection points 405 and 406 can be determined. Referring to the occupancy map 420, the mapping line generation unit 220 generates a grid point (g1, 426) located between the determined intersection points 424 and 425 on the slice 1 4110 including the sensor origin point 401 Can be determined. The mapping line generation unit 220 may divide the initial mapping line into a plurality of segment segments by dividing the initial splitting frame into a plurality of segments based on the determined point 426. [ For example, in the occupancy map 410, when connecting the sensor origin 401 and both end points 403 and 404 of the cell box 402, the two line segments 407 and 408 on the slice 3 including the cell C 403 and cell box 402 can be generated as initial mapping lines. The initial mapping line may be divided into a plurality of segment segments by dividing the initial frustum. The initial mapping line starts with a single line segment representing a representative ray. As the initial line segment is divided into a plurality of segments and the initial mapping line is divided into a plurality of segment segments, a segment corresponding to each of the segment segments is divided into a plurality of representative rays Respectively.

For example, in the occupancy map 420, a line segment connecting a point (427) that passes through a grid point (g1, 426) from a sensor origin to an initial mapping line is divided into a plurality of segment intervals . That is, the line segment connecting the point 427 based on the lattice point 426 can divide the initial frustum into two, and the cell through which the ray corresponding to the line segment passes may be changed by the lattice point 426 have. Referring to the occupancy map 430, the mapping line generation unit 220 repeatedly performs an operation of dividing the initial mapping line into a plurality of segment intervals, for the slice 2 431 to obtain an initial mapping line divided into two segments, Segment segments. For example, referring to occupancy map 440, the initial mapping line may be divided into four segments 441, 442, 443, and 444. At this time, it may be important to efficiently determine a grid point in the frustum so as to divide the initial mapping line into a plurality of segment intervals (for example, segment interval 1 to segment interval 4).

와 같이 표현될 수 있다. 이처럼, 격자점은 초기 절두체 또는 현재 절두체를 m+1개의 서브 절두체로 분할하는 점을 나타낼 수 있으며, 결국 초기 맵핑 라인에서 m+1개의 세그먼트가 생성(즉, 구분)될 수 있다. 이처럼, 점유맵의 이산적 특성으로 인해 격자점을 쉽게 결정하고, 값비싼 정렬방법을 사용하지 않고도 상기 격자점을 기반으로 초기 맵핑 라인을 복수개의 세그먼트들로 빠르게 구분할 수 있다. 셀 C와 관련하여 맵핑 라인을 생성하는 의사 코드(pseudo code) 알고리즘은 아래의 표 1과 같을 수 있다.For example, referring to the occupancy map 420, out ¹ may represent a line farther from the processing slice (i.e., slice 1) from the processing direction, and the mapping line generation unit 220 may generate a mapping slice The grid point can be determined based on the intersection point (int _max , int _min ) between frustums. That is, the lattice points 426 existing between the intersections 424 and 425 between the slice 1 and the initial frustum can be determined as lattice points in the frustum. Here, if the first slice including the sensor origin is slice 1 and the slice including cell C is defined as the last slice, N, the mapping line generation unit 220 generates a mapping line on each slice from slice 1 to slice N-1 The initial mapping line can be divided into a plurality of segments by determining the intersections and repeating the operation of determining the lattice points located between the intersections to the slice N-1. For example, assuming that there are m grid points,

Can be expressed as Thus, the lattice point may represent the point at which the initial frustrum or the current frustum is divided into m + 1 subfractors, so that m + 1 segments in the initial mapping line may be generated (i.e., separated). Thus, due to the discrete nature of the occupancy map, the lattice points can be easily determined and the initial mapping line can be quickly segmented into a plurality of segments based on the lattice points without using expensive alignment methods. The pseudo code algorithm for generating a mapping line with respect to cell C may be as shown in Table 1 below.

 Figure 5 is a diagram provided to illustrate the operation of generating a representative ray in one embodiment of the present invention.

Referring to FIG. 5, all the point cloud data 512 corresponding to the cell C 511 acquired from the sensor may be mapped on the occupancy map 510. For example, five point cloud data 512 belonging to the cell C 511 may be represented on the occupancy map 510. At this time, the representative light generating unit 230 may determine a light ray corresponding to any one point group data among the point cloud data corresponding to the plurality of segment segments divided for the mapping line as a representative light ray. For example, referring to the occupancy map 520, the mapping line 521 can be divided into four segments 522, 523, 524, and 525 based on the lattice points of the slice 1 and the lattice points of the slice 2. Then, in the occupancy map 530, the representative light generating unit 230 generates the representative point light from among the point cloud data positioned between the light rays 532 and 533 passing through both end points of the segment 1 522 from the sensor origin 531 The light corresponding to the data can be determined as a representative light ray. For example, a ray corresponding to the first point cloud data 534 (i.e., a ray passing through the first point cloud data from the sensor origin) may be determined as the representative ray 535. In the same manner, the representative light generating unit 230 can determine the representative ray 536 of the segment segment 2 and the representative ray 537 of the segment segment 3. At this time, there may be no representative ray due to the absence of the point cloud data mapped in the segment section 4.

The representative ray generating unit 230 may calculate the number of point group data mapped (i.e., assigned) for each segment period. For example, the number of mapped point group data can be calculated such that two segment segment 1, one segment segment 2, and two segment segments 3, and the representative light generation unit 230 calculates the number of point grid data Can be set as the weight of the representative ray of the segment segment. That is, the weight of the representative ray 1 541 may be set to 2, the weight of the representative ray 2 542 may be set to 1, and the weight of the representative ray 3 543 may be set to 2. Then, the map updating unit 240 may update the occupancy map by assigning weights corresponding to the weight values set for each representative ray in updating the occupancy map, that is, the number of the point cloud data bundled in occupancy probabilities. The cell through which the representative ray 1 541 passes and the cell through which at least one ray belonging to the segment pass may be the same. That is, the rays corresponding to the representative light rays belonging to the same segment as the representative light rays per each segment segment can traverse the same cells on the occupation map. Accordingly, by updating the cell through which the representative ray passes, not all rays on the occupied map, it is possible to increase the map update speed by avoiding duplication of overlapping cells as the remaining rays pass.

FIG. 6 is a diagram provided for explaining an operation of expanding the operation of generating a representative ray in 3D, according to an embodiment of the present invention. FIG.

To extend to 3D, the mapping lines of 2D planes can be created in three different 2D planes and expanded to 3D. For example, the mapping line generation unit 220 may generate a mapping line of the 2D plane for each of the XY plane, the YZ plane, and the ZX plane, and the representative ray based on the 2D coordinate system may be extended to the 3D coordinate system .

First, for the 3D representation, the bounding volume containing the point cloud data can be calculated, and an initial frustum that traverses the volume as in 2D can be constructed. Then, a representative ray that divides the frustum into sub-frustums and accesses the same set of cells as each divided sub-frustrum can be determined.

Referring to FIG. 5, it can be seen that two rays 602 and 603 from the sensor origin 601 are divided based on the lattice points 604 in each plane, so that the two rays access different cells. That is, the access pattern of the two rays may be different from the grid point projected onto the Y-Z plane 610. [ Thus, by projecting a grid point on each of the YZ plane 610, the ZX plane 620 and the XY plane 630 and verifying that the corresponding lattice point is divided into lattice points in the 2D plane, By creating a mapping line, mapping line creation can be extended to 3D. When a mapping line is generated in 3D, all point group data belonging to a specific cell acquired from the sensor are mapped (or projected) to the mapping lines of the respective planes, and if points corresponding to the same segment in all three mapping lines are determined, The point may be generated as a representative ray having the same access pattern. A pseudo-code algorithm that generates a representative ray for a cell in 3D may be as shown in Table 2 below.

Table 3 below shows representative weights and weights for each representative beam.

According to Table 3, in the case of the resolution 0.6 m, it is confirmed that the average point group data of the indoor point group data is grouped by 17.5, and the average point group data of the outdoor point group data is grouped by 3.4. That is, it can be seen that the rays corresponding to the average of 17.5 point cloud data pass through the same cell in the room, and the rays corresponding to 3.4 point cloud data on the average pass through the same cell outdoors. The higher the grouping rate, the more the number of cells traversing during the update of the maps in the indoor and outdoor data sets can be reduced relatively than when the ratio is smaller.

를 고려하여 수정된 역 센서 모델을 이용하여 점유 맵이 업데이트될 수 있다. 위의 수학식 3의 역 센서 모델은 아래의 수학식 4와 같이 수정될 수 있다.Since the representative ray is generated for various point cloud data even in one cell, the weight of the representative ray

The occupancy map may be updated using the modified inverse sensor model. The inverse sensor model of Equation (3) can be modified as shown in Equation (4) below.

&Quot; (4) "

를 이용하여 대표 광선을 생성할 수도 있다. 여기서, 임계값

는 특정 셀 내에 포함되는 최소 점군 데이터의 개수를 나타낼 수 있다.On the other hand, when the occupancy map is updated using the representative light, the representative light ray may have only a single point when the point cloud data in the cell is small. Thus, only the single point may increase the overhead. Accordingly, the representative light generating unit 230 generates the representative light

May be used to generate a representative ray. Here,

May represent the number of minimum point group data included in a specific cell.

보다 작은 경우, 대표 광선 생성부(230)는 셀에 포함된 모든 점들 각각을 대상으로 대표 광선을 생성할 수 있다. 즉, 셀 내의 모든 점군 데이터들을 개별적으로 처리할 수 있다. 그리고, 셀 내에 포함된 점군 데이터의 개수가 임계값

이상인 경우, 대표 광선 생성부(230)는 세그먼트 구간에 속하는 복수의 광선들 중 어느 하나를 대표 광선으로 결정하고, 맵 업데이트부(240)는 대표 광선에 기초하여 점유맵을 업데이트할 수 있다. 이처럼, 점유맵을 업데이트함에 있어서, 상기 임계값

는 특정 셀을 대상으로 대표 광선을 생성하는 것이 해당 점유맵에 유익한지 아닌지 여부를 확인하기 위해 이용될 수 있으며, 특정 셀 내의 점군 데이터의 개수가 임계값 k 이상인 경우, 대표 광선을 생성하는 것으로 결정하고, k보다 작은 경우 대표 광선을 생성하지 않고 셀 내 모든 점군 데이터에 대해 개별적으로 점유맵을 업데이트하는 것으로 결정될 수 있다. 그리고, 결정된 방법에 따라 점유맵이 업데이트될 수 있다. For example, if the number of the point cloud data included in the cell exceeds the threshold

The representative light generating unit 230 may generate a representative light beam on each of all the points included in the cell. That is, all the point cloud data within the cell can be processed individually. When the number of point cloud data included in the cell is smaller than the threshold value

The representative light generating unit 230 may determine any one of the plurality of light beams belonging to the segment section as a representative light beam and the map updating unit 240 may update the occupancy map based on the representative light beam. As described above, in updating the occupancy map,

May be used to determine whether generating a representative ray for a particular cell is beneficial to the occupation map or not, and if the number of point cloud data in a particular cell is greater than or equal to the threshold value k, And to update the occupancy map separately for all the point cloud data in the cell without creating a representative ray if k is less than k. Then, the occupancy map may be updated according to the determined method.

Hereinafter, the performance comparison between the conventional updating method using the 3D light source and the 3D digital differential analyzer (3DDDA) algorithm will be described. Here, the performance comparison may include both the generation time of the representative ray and the update time of the map having the representative ray. That is, whether or not the performance of the representative light beam is improved can be considered not only in the update of the map but also in the generation time of the representative light ray. The map update method based on the 3DDDA algorithm is described in Non-Patent Document [5] Kai M Wurm , Armin Hornung , Maren Bennewitz , Cyrill Stachniss , and Wolfram Burgard, " Octomap : A probabilistic, flexible, and compact 3d map representation for robotic systems ", in Proc . of the ICRA 2010 workshop on best practice in 3D perception and modeling for mobile manipulation, 2010, vol. In this paper ,

Figure 7 is a graph illustrating the average FPS tested in an octomap and grid map, in one embodiment of the present invention.

In FIG. 7, a graph 710 shows a simulation result performed on point cloud data in an indoor environment, and a graph 720 can show a simulation result on a cloud cloud data in an outdoor environment. Table 4 below shows the representative light generation time and map update time as the simulation result of Fig.

According to FIG. 7, in the case where the occupancy map is updated by generating a representative ray more than when using the 3D Digital Differential Analyzer (3DDDA) algorithm in both the grid map and the octagon map, , 712, 721, 722). That is, the average number of frames per second (FPS) is improved. According to Table 4 above, the performance is 2.5 times faster on the indoor point cloud data and 1.6 times faster on the outdoor cloud point data than on the 3D digital differential analyzer algorithm. Also, it can be seen that the FPS increases when the corresponding degree is higher (0.8 m) than when the resolution is lower (0.2 m). That is, it can be confirmed that the occupancy map update rate is rapidly processed (that is, the update time is shortened) while maintaining the high resolution. Accordingly, the autonomous vehicle, the robot, or the like can quickly detect a collision and quickly cope with objects (e.g., obstacles, etc.) that change in a surrounding situation.

Figure 8 is a graph illustrating the number of cells accessed in an octomap and grid map tested in one embodiment of the present invention.

That is, FIG. 8 may represent a graph illustrating the total time required to generate a representative ray and update an occupancy map (e.g., an octomap and a grid map) based on the generated representative ray.

According to FIG. 8, it can be confirmed that the total time is shortened by an average of 5.6 times and a maximum of 20 times when using the representative light beam, both in indoor and outdoor environments, compared with the use of the 3DDDA algorithm . It can be seen that the time required to generate the representative ray differs depending on the resolution and the point cloud data set, but the overall performance is improved over the Brentham algorithm. For example, for an indoor data set consisting of 89K point cloud data, it may take about 16.6ms to generate a 25.1K representative light at 0.2m resolution using a representative light. As a result, it can be confirmed that a representative ray is generated at 1.51K per second. Since each representative ray has an average of 2.6 point cloud data, it may take 8.6ms to generate a 2.1K representative ray at a resolution of 1m. That is, it can generate a representative ray of 0.24K per second, and the representative ray may have 43.1 point cloud data. As a result, the number of cells traversed for occupancy map update is reduced at a high rate, so that the performance of the system can be improved.

As described above, updating the occupancy map based on the representative ray greatly reduces the number of cells traversing in the map update process, so that the update time can be reduced, i.e., the updater speed can be improved.

9 is a diagram showing a mapping plane used in 3D representation of an operation for generating representative rays in one embodiment of the present invention.

는 평면 z=d에 매핑/투영(projection)된 모서리(911)가 가지는 기울기 값을 나타낼 수 있다. 즉, 2차원에서 센서의 원점으로부터 모든 그리드 포인트를 매핑 라인에 매핑/투영(prokjection)시킨 것과 마찬가지로, 3차원에서 센서의 원점으로부터 모든 셀(cell)의 모서리가 매핑 평면(910)에 매핑/투영(prokjection)될 수 있다.9, z = d represents an arbitrary plane in three dimensions (an XY plane having a z = d value)

Can represent the slope value of the edge 911 mapped / projected on the plane z = d. That is, as in the case where all the grid points are mapped / projected from the origin of the sensor in two dimensions to the mapping line, the edges of all the cells from the origin of the sensor in three dimensions are mapped / projected onto the mapping plane 910 (prokjection).

A mapping line is used to generate the representative light in two dimensions. In FIG. 6, the concept of generating a representative light in two dimensions is extended to three dimensions. At this time, a mapping plane can be used when expanding to three dimensions, and concepts for generating a representative ray in each of two-dimensional and three-dimensional directions can be as shown in Table 5 below.

If the grid point divides the traversal pattern of light rays in two dimensions, the edge of the cell can distinguish the traversal pattern of light rays in three dimensions. Accordingly, a data structure that generates a representative ray is classified into three groups by classifying a traversal pattern of rays existing in an initial frustum and using a mapping plane to select a representative ray. It can be expanded to a dimension.

For example, one segment is divided into segments by using all the grid points present in the seed frustum in two dimensions. However, in the three-dimensional segment, all the edges 901 and 902 existing in the seed frustum, 911 to the mapping plane 910 to divide them into regions. Here, each of the plurality of regions divided on the mapping plane 910 may represent a two-dimensional region in which three-dimensional rays, which pass through the same space, can always be mapped.

Accordingly, the mapping plane 910 may be composed of several regions, and one region in the mapping plane 910 is matched to one traversal pattern, Can be generated as a representative ray by bundling rays that are mapped to the same region. Here, the operation of projecting the edges 911 onto the mapping plane 910 and dividing the edges into a plurality of areas may be performed by the mapping line generation unit 220 of FIG. 2, May be performed in a mapping plane processing section (not shown) that separately processes the process for expanding representative ray generation. When performed in the mapping plane processing unit, the occupied map updating system 200 of FIG. 2 may further include a mapping plane processing unit (not shown).

10 is a view for explaining an operation of bundling rays belonging to the same area among regions included in a mapping plane to generate a representative ray according to an embodiment of the present invention.

10, light rays 1002 and 1003 are mapped to the same area 1011 among a plurality of areas included in the mapping plane 1010, and the light ray 1001 is mapped to the other area 1012 ). ≪ / RTI > Since the rays 1002 and 1003 mapped to the same area 1011 pass through the same space, they can be bundled into one representative ray. The representative light beam generator 230 bundles the light beams 1002 and 1003 belonging to the region 1011 into one and the same representative light beam and the light beam 1001 mapped to the other region 1012, (1002, 1003), as shown in FIG. For example, the representative light generating unit 230 may determine a light ray 1001 as a representative light ray representative of the light rays mapped to the corresponding region 1012, and determine the light ray 1003 as a representative light ray representative of the light rays mapped to the corresponding region 1011 It can be decided as a representative ray.

11 is a diagram showing the determination of the dendritic light using the mapping single plane and the determination of the representative ray using the mapping line in three three-dimensional planes in accordance with an embodiment of the present invention.

Referring to FIG. 11, the three-dimensional extension of generating a representative light ray includes determining / generating a representative ray using a mapping plane in three dimensions, determining a representative ray using mapping lines in three two- There are two ways to create / create. At this time, mapping lines 1111, 1121, and 1131 are generated in each of the two two-dimensional planes (XY, ZX, ZY, 1110, 1120, and 1130) to generate representative light, It can be more efficient.

11, a grid point 1140 may be a grit point corresponding to the same segment in all of the mapping lines of each of the two two-dimensional planes 1110, 1120, 1130.

Figure 12 is a diagram provided to illustrate why it is more efficient to generate a representative ray in each dimension in accordance with one embodiment of the present invention.

Rather than creating a single mapping plane in three dimensions, the process of creating two three-dimensional mapping planes can be fast. And, the calculation time of mapping the rays into the representative rays can be matched.

Referring to FIG. 12, when it is desired to obtain lines 1210 and 1220 on a mapping plane, referring to 1201, a corner of a cell is mapped / projected onto a mapping plane in three dimensions, ) Can be calculated. At this time, a total of three corners can be considered. Referring to 1202, a line segment 1220 to be obtained can be obtained by considering a total of one grid point in the process of matching a segment of the mapping line in the Z-X plane. That is, in the case of 1201 (i.e., when one mapping plane is created in three dimensions), three edges must be considered in the process of calculating lines for dividing each region in the mapping plane, and 1202 (i.e., Dimensional mapping plane), it is more efficient to generate a representative ray by obtaining a mapping line of three sub-planes (XY, ZX, ZY), since there is a difference in considering one grid point .

The method according to an embodiment may be implemented in the form of a program command that can be executed through various computer means and recorded in a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, and the like, alone or in combination. The program instructions to be recorded on the medium may be those specially designed and configured for the embodiments or may be available to those skilled in the art of computer software. Examples of computer-readable media include magnetic media such as hard disks, floppy disks and magnetic tape; optical media such as CD-ROMs and DVDs; magnetic media such as floppy disks; Magneto-optical media, and hardware devices specifically configured to store and execute program instructions such as ROM, RAM, flash memory, and the like. Examples of program instructions include machine language code such as those produced by a compiler, as well as high-level language code that can be executed by a computer using an interpreter or the like.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed exemplary embodiments. For example, it is to be understood that the techniques described may be performed in a different order than the described methods, and / or that components of the described systems, structures, devices, circuits, Lt; / RTI > or equivalents, even if it is replaced or replaced.

Therefore, other implementations, other embodiments, and equivalents to the claims are also within the scope of the following claims.

Claims (23)

A method for updating an occupancy map based on a grid and an octree,
Generating a mapping line based on the point cloud data obtained from the sensor;
Generating a super ray by identifying cells traversing the same cell on a plurality of cells constituting the occupancy map based on the generated mapping line; And
Updating the occupancy map by updating a cell through which the representative ray corresponding to some of the rays associated with the plurality of cells constituting the occupancy map passes,
Lt; / RTI >
Wherein the generating the mapping line comprises:
Converting the coordinate system of the point cloud data into a world coordinate system corresponding to the occupancy map;
Generating a virtual cell box including all the point cloud data corresponding to any one of the plurality of cells constituting the occupancy map among the point cloud data of which the coordinate system is transformed; And
Generating a line segment overlapping between the cell box and a slice associated with the cell box on the mapping line based on a sensor origin on the occupancy map
/ RTI > delete The method according to claim 1,
The step of generating the overlapping line segment with the mapping line includes:
Determining a plurality of intersection points connecting the two end points of the cell box from the sensor origin on a first slice comprising the sensor origin;
Determining a grid point located between the determined intersection points; And
Dividing the initial frustum into a plurality of segments based on the lattice points;
/ RTI > The method according to claim 1,
Wherein the generating the mapping line comprises:
Dividing the mapping line into a plurality of segment segments based on a point on a mapping line connecting a mapping line passing through a grid point from a sensor origin on an occupancy map
/ RTI > delete A method for updating an occupancy map based on a grid and an octree,
Generating a mapping line based on the point cloud data obtained from the sensor;
Generating a super ray by identifying cells traversing the same cell on a plurality of cells constituting the occupancy map based on the generated mapping line; And
Updating the occupancy map by updating a cell through which the representative ray corresponding to some of the rays associated with the plurality of cells constituting the occupancy map passes,
Lt; / RTI >
The step of generating the super ray comprises:
Determining one of the point cloud data for each of the segment segments with respect to the point cloud data located in each of the plurality of segment segments that define the mapping line;
Determining a ray of light projected on the point cloud data determined from the sensor origin as the representative ray (super ray);
Calculating the number of point cloud data located in each of the plurality of segment intervals on a segment segment basis; And
Setting the number of calculated point cloud data to a weight of the determined representative light ray
/ RTI > The method according to claim 6,
Storing the point cloud data corresponding to the representative ray in place of the plurality of point cloud data corresponding to the same segment period for updating the occupancy map
The occupancy map updating method further comprising: The method according to claim 1,
Wherein the mapping line comprises:
By being generated for each of the XY plane, YZ plane, and ZX plane, it is possible to extend the representative ray based on the 2D coordinate system on the basis of the 3D coordinate system
And updating the occupancy map. delete A method for updating an occupancy map based on a grid and an octree,
Generating a mapping line based on the point cloud data obtained from the sensor;
Generating a super ray by identifying cells traversing the same cell on a plurality of cells constituting the occupancy map based on the generated mapping line; And
Updating the occupancy map by updating a cell through which the representative ray corresponding to some of the rays associated with the plurality of cells constituting the occupancy map passes,
Lt; / RTI >
Wherein the step of generating the representative light beam comprises:
In a two-dimensional coordinate system, a plurality of regions divided according to projections of edges existing in an initial frustum on a mapping plane of a three-dimensional coordinate system, To create a single representative ray for each region
And updating the occupancy map. The method according to claim 1,
The rays corresponding to each of the point cloud data belonging to the segment segment corresponding to the representative light ray are traversed on the same cells on the occupation map
And updating the occupancy map. A system for updating an occupancy map based on a grid and an octree,
A mapping line generation unit for generating a mapping line based on point cloud data acquired from a sensor;
A representative ray generator for generating a super ray by identifying a cell traversing the same cell for a plurality of cells constituting the occupancy map based on the generated mapping line; And
A map updating unit updating the occupancy map by updating a cell through which the representative ray corresponding to a part of the rays related to the plurality of cells constituting the occupancy map passes,
Lt; / RTI >
Wherein the mapping line generation unit comprises:
A coordinate system of the point cloud data is converted into a world coordinate system corresponding to the occupation map and all point cloud data corresponding to any one of a plurality of cells constituting the occupancy map among the point cloud data, And generating a line segment overlapping between the cell box and the slice associated with the cell box as the mapping line based on the sensor origin on the occupancy map
The occupancy map updating system comprising: delete 13. The method of claim 12,
Wherein the mapping line generation unit comprises:
Determining a plurality of intersection points connecting the two end points of the mapping line from the sensor origin on a first slice including the sensor origin, determining a grid point located between the determined intersection points, And dividing the initial frustum into a plurality of segments
The occupancy map updating system comprising: 13. The method of claim 12,
Wherein the mapping line generation unit comprises:
A mapping line is divided into a plurality of segment segments based on a point on a mapping line connecting a mapping line passing through a grid point from a sensor origin on an occupancy map
The occupancy map updating system comprising: delete A system for updating an occupancy map based on a grid and an octree,
A mapping line generation unit for generating a mapping line based on point cloud data acquired from a sensor;
A representative ray generator for generating a super ray by identifying a cell traversing the same cell for a plurality of cells constituting the occupancy map based on the generated mapping line; And
A map updating unit updating the occupancy map by updating a cell through which the representative ray corresponding to a part of the rays related to the plurality of cells constituting the occupancy map passes,
Lt; / RTI >
The representative light beam generator may include:
Wherein the point cloud data is determined for each segment segment with respect to the point cloud data located in each of the plurality of segment segments that define the mapping line, and the light reflected by the point cloud data determined from the sensor origin is referred to as the super ray ), ≪ / RTI >
The number of point cloud data located in each of the plurality of segment intervals is calculated for each segment period and the number of the calculated point cloud data is set to the determined weight of the representative light ray
The occupancy map updating system comprising: 18. The method of claim 17,
A storage unit for storing the point cloud data corresponding to the representative rays in place of the plurality of point cloud data corresponding to the same segment period for updating occupancy maps;
The occupancy map updating system further comprising: 13. The method of claim 12,
Wherein the mapping line comprises:
By being generated for each of the XY plane, YZ plane, and ZX plane, it is possible to extend the representative ray based on the 2D coordinate system on the basis of the 3D coordinate system
The occupancy map updating system comprising: delete A system for updating an occupancy map based on a grid and an octree,
A mapping line generation unit for generating a mapping line based on point cloud data acquired from a sensor;
A representative ray generator for generating a super ray by identifying a cell traversing the same cell for a plurality of cells constituting the occupancy map based on the generated mapping line; And
A map updating unit updating the occupancy map by updating a cell through which the representative ray corresponding to a part of the rays related to the plurality of cells constituting the occupancy map passes,
Lt; / RTI >
The representative light beam generator may include:
In a two-dimensional coordinate system, a plurality of regions divided according to projections of edges existing in an initial frustum on a mapping plane of a three-dimensional coordinate system, To create a single representative ray for each region
The occupancy map updating system comprising: 13. The method of claim 12,
The rays corresponding to each of the point cloud data belonging to the segment segment corresponding to the representative light ray are traversed on the same cells on the occupation map
The occupancy map updating system comprising: A computer program stored on a recording medium for executing a method for updating an occupancy map based on a grid and an octree in combination with a computer-implemented electronic device,
A method for updating occupancy map,
Generating a mapping line based on the point cloud data obtained from the sensor;
Generating a super ray by identifying cells traversing the same cell on a plurality of cells constituting the occupancy map based on the generated mapping line; And
Updating the occupancy map by updating a cell through which the representative ray corresponding to some of the rays associated with the plurality of cells constituting the occupancy map passes,
Lt; / RTI >
Wherein the generating the mapping line comprises:
Converting the coordinate system of the point cloud data into a world coordinate system corresponding to the occupancy map;
Generating a virtual cell box including all the point cloud data corresponding to any one of the plurality of cells constituting the occupancy map among the point cloud data of which the coordinate system is transformed; And
Generating a line segment overlapping between the cell box and a slice associated with the cell box on the mapping line based on a sensor origin on the occupancy map
And a computer program product.

Images