CN-121979187-A - Unmanned cleaning vehicle driving safety method and device based on occupation grid prediction

CN121979187ACN 121979187 ACN121979187 ACN 121979187ACN-121979187-A

Abstract

The scheme includes that three-dimensional point cloud data of the surrounding environment of a target unmanned sweeping vehicle in a working state are obtained in real time based on a laser radar sensor and stored according to time sequence to obtain a three-dimensional point cloud data sequence, the three-dimensional point cloud data sequence is subjected to rasterization processing in sequence and mapped to a two-dimensional plane coordinate system to obtain a two-dimensional occupied grid sequence, a predicted occupied grid map corresponding to future motion tracks of dynamic obstacles in the surrounding environment of the target unmanned sweeping vehicle is generated through a space-time prediction network based on the two-dimensional occupied grid sequence, the predicted occupied grid map is used for representing position distribution and motion trend of the dynamic obstacles in future time at the current moment, a dynamic safety domain model is built based on the predicted occupied grid map and the current motion state of the target unmanned sweeping vehicle, and corresponding active safety decisions are executed according to calculation results.

Inventors

LIANG SHUANG
DOU FENGQIAN
BAI YUNLONG

Assignees

云创智行科技(湖州)有限公司

Dates

Publication Date: 20260505
Application Date: 20251226

Claims (9)

1. The unmanned sweeping vehicle driving safety method based on occupation grid prediction is characterized by comprising the following steps of: Acquiring three-dimensional point cloud data of the surrounding environment when the target unmanned sweeping vehicle is in a working state in real time based on a laser radar sensor arranged on the target unmanned sweeping vehicle, and storing the three-dimensional point cloud data according to the time sequence acquired by the laser radar to obtain a three-dimensional point cloud data sequence within a preset time period before the current moment to be analyzed; the three-dimensional point cloud data sequence is subjected to rasterization in sequence and mapped to a two-dimensional plane coordinate system so as to obtain a two-dimensional occupation grid sequence consistent with the time sequence; Generating a predicted occupation grid map corresponding to a future motion trail of a dynamic obstacle in the surrounding environment of the target unmanned sweeping vehicle through a space-time prediction network based on the two-dimensional occupation grid sequence, wherein the predicted occupation grid map is used for representing the position distribution and the motion trend of the dynamic obstacle in a future period of time at the current moment; and constructing a dynamic safety domain model based on the predicted occupancy grid map and the current motion state of the target unmanned sweeping vehicle, and executing a corresponding active safety decision according to the calculation result of the dynamic safety domain model.
2. The unmanned cleaning vehicle active safety method based on occupancy grid prediction of claim 1, wherein the sequentially rasterizing and mapping the three-dimensional point cloud data sequence to a two-dimensional planar coordinate system to obtain a two-dimensional occupancy grid sequence consistent with the chronological order comprises: sequentially carrying out noise reduction treatment on each frame of point cloud data in the three-dimensional point cloud data sequence to obtain a denoised point cloud data sequence; Sequentially mapping the denoised point cloud data of each frame to a two-dimensional plane, projecting three-dimensional point cloud coordinates to a two-dimensional grid coordinate system through coordinate conversion, and distributing occupation state values for each grid to generate a corresponding two-dimensional occupation grid; And storing the generated two-dimensional occupation grids into a sequence according to the time sequence, and constructing and forming the two-dimensional occupation grid sequence with consistent time sequence.
3. The unmanned cleaning vehicle driving safety method based on occupied grid prediction according to claim 2, wherein the noise reduction processing is voxel noise reduction, specifically comprising dividing each frame of point cloud data into cubic grids with preset volumes, reserving the geometric center of point clouds in each grid as a representative point, and eliminating the rest redundant points to realize noise reduction; The operation of assigning an occupied state value to each grid is realized through binarization processing, and specifically comprises the steps of marking a certain grid as an occupied state if at least one projection point exists in the certain grid, and marking the grid as an unoccupied state if the projection point exists in the certain grid.
4. The unmanned sweeping vehicle active safety method of claim 1, wherein generating a predicted occupancy grid map corresponding to a future motion trajectory of a dynamic obstacle in the surrounding environment of the target unmanned sweeping vehicle through a spatiotemporal prediction network based on the two-dimensional occupancy grid sequence comprises: Taking the two-dimensional occupation grid sequences with consistent time sequences as input, and sending the two-dimensional occupation grid sequences into a trained space-time prediction network; extracting space-time characteristics contained in the two-dimensional occupation grid sequence through a characteristic extraction module in the space-time prediction network; inferring, by a prediction module in the spatio-temporal prediction network, a motion vector of the dynamic obstacle based on the extracted spatio-temporal features; generating a series of prediction grid probability maps in a future period of time according to the motion vector; And carrying out fusion processing on the series of prediction grid probability maps to generate a unified prediction occupation grid map.
5. The unmanned cleaning vehicle active safety method based on occupancy grid prediction of claim 4, wherein the two-dimensional occupancy grid sequence is a time sequence of consecutive multiframes with a predetermined time interval between adjacent frames, the predetermined time interval matching the acquisition frequency of the lidar sensor.
6. The unmanned cleaning vehicle active safety method based on occupancy grid prediction of claim 1, wherein the constructing a dynamic safety domain model based on the predicted occupancy grid map and the current motion state of the target unmanned cleaning vehicle, and executing a corresponding active safety decision according to the calculation result of the dynamic safety domain model, comprises: calculating a basic safety distance according to the current speed, the maximum deceleration and the system response time of the target unmanned sweeper; calculating a dynamic risk margin based on the predicted obstacle movement trend represented by the occupancy grid map; combining the basic safety distance with the dynamic risk margin to construct the dynamic safety domain model so as to determine the comprehensive safety distance; Calculating the actual distance between the target unmanned sweeper and surrounding obstacles in real time, and comparing the actual distance with the comprehensive safety distance; triggering a hierarchical security response strategy according to the comparison result, wherein the hierarchical security response strategy comprises the following steps: When the actual distance is larger than a preset first safety threshold value, executing path smooth optimization; When the actual distance is smaller than or equal to the first safety threshold value and larger than a preset second safety threshold value, performing deceleration and local path re-planning; triggering emergency braking when the actual distance is less than or equal to the second safety threshold; wherein the first safety threshold is greater than the second safety threshold.
7. The unmanned cleaning vehicle active safety method based on occupancy grid prediction of claim 6, wherein the base safety distance is calculated based on the following formula: ; Wherein the symbols are Representing the basic safety distance, symbol Sign for indicating current speed of the target unmanned sweeping vehicle Indicating the maximum deceleration achievable by the vehicle, sign Representing system response time, sign Representing the risk margin.
8. The unmanned cleaning vehicle active safety method based on occupancy grid prediction of claim 6, wherein the dynamic risk margin The calculation formula of (2) is as follows: ; Wherein the symbols are Representing the instantaneous speed of a predicted obstacle, sign Indicating the acceleration of the obstacle, sign Representing the direction change rate of the obstacle, representing the maneuvering risk index and the sign Representing the dynamic obstacle occupancy probability integral.
9. An unmanned cleaning vehicle actuation safety device based on occupancy grid prediction, the device comprising: The system comprises a point cloud data acquisition and storage module, a target unmanned sweeping vehicle, a laser radar sensor and a control module, wherein the point cloud data acquisition and storage module is used for acquiring three-dimensional point cloud data of the surrounding environment when the target unmanned sweeping vehicle is in a working state in real time based on the laser radar sensor arranged on the target unmanned sweeping vehicle, and storing the three-dimensional point cloud data according to the time sequence acquired by the laser radar, so as to obtain a three-dimensional point cloud data sequence within a preset time length before the current moment to be analyzed; The rasterization processing and sequence construction module is used for sequentially rasterizing the three-dimensional point cloud data sequence and mapping the three-dimensional point cloud data sequence to a two-dimensional plane coordinate system so as to obtain a two-dimensional occupation grid sequence consistent with the time sequence; The system comprises a prediction occupation grid map generation module, a prediction occupation grid map generation module and a control module, wherein the prediction occupation grid map generation module is used for generating a prediction occupation grid map corresponding to a future motion trail of a dynamic obstacle in the surrounding environment of the target unmanned sweeping vehicle through a space-time prediction network based on the two-dimensional occupation grid sequence, and the prediction occupation grid map is used for representing the position distribution and the motion trend of the dynamic obstacle in a future period of time at the current moment; The dynamic security domain construction and decision execution module is used for constructing a dynamic security domain model based on the predicted occupation grid map and the current motion state of the target unmanned sweeping vehicle, and executing a corresponding active security decision according to the calculation result of the dynamic security domain model.

Description

Unmanned cleaning vehicle driving safety method and device based on occupation grid prediction Technical Field The invention relates to the technical field of unmanned sweeping vehicles, in particular to an unmanned sweeping vehicle driving safety method and device based on occupation grid prediction. Background With the continued popularity of autopilot technology, the concept of autopilot has been widely penetrated into a variety of industries. Wherein, combine together autopilot technique and traditional sanitation motor sweeper to replace traditional manual driving mode, can practice thrift financial resources, material resources, human cost through intelligent control's mode, effectively improve the efficiency of cleaning operation simultaneously. In order to realize autonomous unmanned driving of the sweeper, the safety of the sweeper and other traffic participants is important, so that the research and development of the active safety technology of the vehicle have important significance. In the existing active safety technical field of unmanned sweeping vehicles, ultrasonic radars are mostly adopted, or laser radar point clouds are directly utilized to construct static occupied grid maps (Occupancy Grid Map, OGM), and emergency braking or local path rescheduling operation is triggered by detecting the position of an obstacle at the current moment. For example, using point cloud data generated by lidar, the point cloud is projected onto a two-dimensional plane to construct a real-time occupancy grid map, and a fixed safety threshold, such as 1.5m, is set, which triggers a deceleration or parking action when an obstacle within the grid enters the threshold range. However, the above-mentioned existing methods still have some limitations in practical application. For example, such methods can only react according to the obstacle position at the current moment, and cannot effectively predict the movement trend of the obstacle, especially the dynamic obstacle. Disclosure of Invention The present disclosure provides a method and apparatus for vehicle safety of an unmanned cleaning vehicle based on occupancy grid prediction, which are used to overcome at least one technical problem existing in the related art. According to a first aspect of embodiments of the present specification, there is provided an unmanned cleaning vehicle actuation safety method based on occupancy grid prediction, comprising: Acquiring three-dimensional point cloud data of the surrounding environment when the target unmanned sweeping vehicle is in a working state in real time based on a laser radar sensor arranged on the target unmanned sweeping vehicle, and storing the three-dimensional point cloud data according to the time sequence acquired by the laser radar to obtain a three-dimensional point cloud data sequence within a preset time period before the current moment to be analyzed; the three-dimensional point cloud data sequence is subjected to rasterization in sequence and mapped to a two-dimensional plane coordinate system so as to obtain a two-dimensional occupation grid sequence consistent with the time sequence; Generating a predicted occupation grid map corresponding to a future motion trail of a dynamic obstacle in the surrounding environment of the target unmanned sweeping vehicle through a space-time prediction network based on the two-dimensional occupation grid sequence, wherein the predicted occupation grid map is used for representing the position distribution and the motion trend of the dynamic obstacle in a future period of time at the current moment; and constructing a dynamic safety domain model based on the predicted occupancy grid map and the current motion state of the target unmanned sweeping vehicle, and executing a corresponding active safety decision according to the calculation result of the dynamic safety domain model. In some optional embodiments, the rasterizing the three-dimensional point cloud data sequence sequentially and mapping the three-dimensional point cloud data sequence to a two-dimensional plane coordinate system to obtain a two-dimensional occupation grid sequence consistent with the chronological order includes: sequentially carrying out noise reduction treatment on each frame of point cloud data in the three-dimensional point cloud data sequence to obtain a denoised point cloud data sequence; Sequentially mapping the denoised point cloud data of each frame to a two-dimensional plane, projecting three-dimensional point cloud coordinates to a two-dimensional grid coordinate system through coordinate conversion, and distributing occupation state values for each grid to generate a corresponding two-dimensional occupation grid; And storing the generated two-dimensional occupation grids into a sequence according to the time sequence, and constructing and forming the two-dimensional occupation grid sequence with consistent time sequence. In some optional embodiments, the noise reduction processi