CN-116842025-B - Occupancy probability grid map updating method, device, equipment and medium

CN116842025BCN 116842025 BCN116842025 BCN 116842025BCN-116842025-B

Abstract

The embodiment of the disclosure discloses an occupation probability grid map updating method, device, equipment and medium, which comprise the steps of obtaining a hit point and a boundary miss point associated with single-frame point cloud data, carrying out association statistics on the light projection data of the hit point and the boundary miss point and a counting grid subgraph by utilizing a light projection algorithm to obtain a statistical result, and updating an occupation probability grid subgraph to be updated according to the statistical result, wherein a coordinate system of the counting grid subgraph coincides with a coordinate system of the occupation probability grid subgraph, the statistical result comprises types of grids in the counting grid subgraph and times of occurrence of hit events or miss events of the grids, and the types of the grids comprise hit grids and miss grids. The present disclosure improves update accuracy of occupancy probability grid maps.

Inventors

LIN XIN
FENG JINGYI
ZHANG DAN

Assignees

驭势科技（北京）有限公司

Dates

Publication Date: 20260512
Application Date: 20230706

Claims (9)

1. A method of updating an occupancy probability grid map, the method comprising: acquiring a hit point and a boundary miss point associated with single-frame point cloud data; Performing association statistics on light projection data of the hit point and the boundary miss point and a counting grid sub-graph by utilizing a light projection algorithm to obtain a statistical result, wherein the statistics comprises the steps of matching the hit point with an occupation probability grid sub-graph according to an original point of a laser radar for acquiring single-frame point cloud data to obtain a pose of the hit point under a coordinate system of the occupation probability grid sub-graph; regarding the hit point and the boundary miss point as laser ray endpoints, traversing each laser ray endpoint, marking a line segment formed by the current laser ray endpoint and an origin of the laser radar as a first line segment aiming at the current laser ray endpoint, and marking grids in a counting grid subgraph through which the first line segment passes as miss grids; Traversing each laser ray endpoint, traversing the mass grids sequentially according to the direction from the current laser ray endpoint to the laser radar origin aiming at the current laser ray endpoint, and when the current laser ray endpoint is a ground ray endpoint and the current traversed mass grid is a non-ground mass grid, assigning the count value corresponding to the current traversed mass grid as a large negative number if the count value corresponding to the current traversed mass grid is a positive value or zero, and subtracting the count value corresponding to the current traversed mass grid by one if the count value corresponding to the current traversed mass grid is a negative value; when the current laser ray end point is a non-ground ray end point, updating a count value corresponding to the pass grid based on a second strategy, and when the current traversed pass grid is a ground pass grid, directly entering the next operation without updating the count value corresponding to the current traversed pass grid; And updating the occupied probability grid subgraph to be updated according to the statistical result, wherein the coordinate system of the counting grid subgraph coincides with the coordinate system of the occupied probability grid subgraph, the statistical result comprises the type of each grid in the counting grid subgraph and the times of occurrence of hit events or miss events of each grid, and the types of the grids comprise hit grids and miss grids.
2. The method of claim 1, wherein the obtaining the hit point and the boundary miss point associated with the single-frame point cloud data comprises: determining points in the acquisition boundary in the single-frame point cloud data as hit points; And determining an intersection point of the laser beam corresponding to a point outside the acquisition boundary in the single-frame point cloud data and the acquisition boundary as a boundary miss point.
3. The method of claim 1, wherein the performing statistics on the ray casting data of the hit point and the boundary miss point and the counting grid subgraph by using a ray casting algorithm to obtain statistics results, further comprises: obtaining the pose confidence corresponding to the pose; And if the pose confidence is greater than a preset positioning confidence threshold, carrying out association statistics on the ray projection data of the hit point and the boundary MIss point and the counting grid subgraph by utilizing a ray projection algorithm based on the pose of the hit point under the coordinate system of the grid subgraph to be updated, and obtaining a statistical result.
4. The method of claim 1, wherein the step of determining the position of the substrate comprises, The updating the count value corresponding to the Miss grid based on the second policy includes: And if the count value corresponding to the currently traversed MIS grid is a negative value, subtracting 1 from the count value corresponding to the currently traversed MIS grid.
5. The method of claim 4, wherein after updating the count value corresponding to the currently traversed boss grid, the performing, based on the pose of the hit point in the coordinate system of the grid sub-graph to be updated, correlation statistics on the ray casting data of the hit point and the boundary boss point and the count grid sub-graph by using a ray casting algorithm to obtain a statistical result, further comprising: If the count value corresponding to the currently traversed MISS grid is smaller than a first threshold value, ending the operation of traversing the MISS grid in sequence according to the direction from the current laser ray end point to the laser radar origin point so as to traverse the next laser ray end point; and if the count value corresponding to the currently traversed MIS grid is greater than or equal to the first threshold value, continuing to traverse the next MIS grid.
6. The method of claim 1, wherein updating the occupancy probability grid subgraph based on the statistics comprises: Traversing all grids in all counting grid subgraphs, and counting the sum of current grid count values in all counting grid subgraphs containing the current grid aiming at the current grid; If the sum of the calculated values of the current grid in all the counting grid subgraphs is positive, determining whether the calculated values of the current grid in all the counting grid subgraphs are larger than a second threshold value, and if so, carrying out N times of binary Bayesian filtering state updating on the corresponding grid of the current grid in the occupancy probability grid subgraphs; if the sum of the calculated values of the current grid in all the counting grid subgraphs is negative, determining whether the calculated values of the current grid in all the counting grid subgraphs are smaller than a third threshold value, and if so, updating the occupied state of the corresponding grid of the current grid in the occupied probability grid subgraphs to be idle.
7. An occupancy probability grid map updating apparatus, comprising: the acquisition module is used for acquiring hit points and boundary miss points associated with single-frame point cloud data; The statistics module is used for carrying out association statistics on the light projection data of the hit point and the boundary miss point and a counting grid subgraph by utilizing a light projection algorithm to obtain a statistics result, and comprises the steps of matching the hit point with an occupation probability grid subgraph according to the original point of a laser radar for acquiring the single-frame point cloud data to obtain the pose of the hit point under the coordinate system of the occupation probability grid subgraph; regarding the hit point and the boundary miss point as laser ray endpoints, traversing each laser ray endpoint, marking a line segment formed by the current laser ray endpoint and an origin of the laser radar as a first line segment aiming at the current laser ray endpoint, and marking grids in a counting grid subgraph through which the first line segment passes as miss grids; Traversing each laser ray endpoint, traversing the mass grids sequentially according to the direction from the current laser ray endpoint to the laser radar origin aiming at the current laser ray endpoint, and when the current laser ray endpoint is a ground ray endpoint and the current traversed mass grid is a non-ground mass grid, assigning the count value corresponding to the current traversed mass grid as a large negative number if the count value corresponding to the current traversed mass grid is a positive value or zero, and subtracting the count value corresponding to the current traversed mass grid by one if the count value corresponding to the current traversed mass grid is a negative value; when the current laser ray end point is a non-ground ray end point, updating a count value corresponding to the pass grid based on a second strategy, and when the current traversed pass grid is a ground pass grid, directly entering the next operation without updating the count value corresponding to the current traversed pass grid; The updating module is used for updating the occupied probability grid subgraph to be updated according to the statistical result, wherein the coordinate system of the counting grid subgraph coincides with the coordinate system of the occupied probability grid subgraph, the statistical result comprises the types of grids in the counting grid subgraph and the times of occurrence of hit events or miss events of the grids, and the types of the grids comprise hit grids and miss grids.
8. An electronic device, the electronic device comprising: one or more processors; a storage means for storing one or more programs; The one or more programs, when executed by the one or more processors, cause the one or more processors to implement the method of any of claims 1-6.
9. A computer readable storage medium, on which a computer program is stored, characterized in that the program, when being executed by a processor, implements the method according to any of claims 1-6.

Description

Occupancy probability grid map updating method, device, equipment and medium Technical Field The disclosure relates to the technical field of occupancy probability grid maps, and in particular relates to an occupancy probability grid map updating method, an occupancy probability grid map updating device, an occupancy probability grid map updating equipment and an occupancy probability grid map updating medium. Background The occupancy probability grid map is a map commonly used in the field of artificial intelligence (such as automatic driving vehicles, robots and the like), and consists of grids with certain precision, and when path planning is performed based on the occupancy probability grid map, the probability of each grid in the occupancy probability grid map being occupied is estimated according to environmental information returned by a laser radar, so that a basis is provided for path planning. There are two kinds of states of the grids in the occupancy probability grid map, one is occupied, the other is idle, wherein the occupied grid represents that there is an obstacle at the position represented by the grid, such a grid is also called hit grid, and the idle grid represents that there is no obstacle at the position represented by the grid unit, such a grid is also called miss grid. The traditional occupation probability grid map updating method mainly comprises the steps of determining whether each grid in the occupation probability grid map to be updated contains an obstacle according to laser point cloud data fed back by a laser radar, if the grid is hit by the point cloud, indicating that a hit event occurs in the grid, increasing the occupation probability of the grid, and if the grid is crossed by a laser beam of the point cloud, indicating that a miss event occurs in the grid, and reducing the occupation probability of the grid. In the prior art, however, the current frame point cloud is generally directly inserted into the used occupancy probability grid map in combination with the current positioning pose, then a ray casting algorithm is used for calculating a hit event and a miss event based on each point in the point cloud to obtain a hit grid and a miss grid, and finally a binary Bayesian filter is used for updating the hit grid and the miss grid. First, in the prior art, the calculation errors may exist in the positioning pose used when the current frame point cloud is inserted into the occupied probability grid map, and the use of the pose with errors may cause errors in the hit grid and the miss grid acquired based on each point through the ray casting algorithm, resulting in reduced accuracy of the updated map. Secondly, in the prior art, only the points within the acquisition boundary in the point cloud of the current frame are used for updating the map, but the points outside the acquisition boundary are not used, so when some obstacle grids exist in the occupation probability grid map to be updated, and only laser rays corresponding to the points exceeding the acquisition boundary pass through the obstacle grids when the map is updated, the obstacle grids cannot be updated only by using the points within the acquisition boundary, so that the accuracy of the updated map is reduced, specifically, for example, a certain grid exists in the acquisition boundary of the previous frame, the next frame has no obstacle, the grid needs to be updated in the next frame, and the points outside the acquisition boundary are directly deleted in the prior art and do not participate in map updating. Third, in the prior art, when the incident angle of the laser beam to the ground is too large (approaching 90 °), the ground grid is easily killed by mistake (i.e., the ground grid is determined as a miss grid), resulting in that the ground is thinned and hollowed out, and the accuracy of the updated map is reduced. Fourth, in the prior art, when the occupancy probability grid map is updated, the probability value of the corresponding hit grid is increased because the dynamic obstacle is also hit by the laser ray, so that a smear or ghost is formed in the map, and the dynamic obstacle is added to the occupancy probability grid map, so that the accuracy of the updated map is reduced. Fifth, in the prior art, the hit and the miss grids to be updated need to be obtained by calculating the hit event and the miss event for each point by using a complete ray projection algorithm, and because the laser radar rays are radial, the number of times of passing through the grids near the origin of the laser radar is more, the corresponding miss events are more, that is, the miss events in the miss grid are more, and all the miss grids passing through the ray need to be traversed for each ray, so that the calculated amount is larger, the map updating is time-consuming, and the efficiency is reduced. In view of this, the present invention has been made. Disclosure of Invention In order to solve the above technical p