Path planning method and system based on vehicle controller cooperative perception

Abstract

The application provides a path planning method and system based on vehicle controller cooperative perception, relates to the technical field of path planning, and comprises the following steps: acquiring first perception data of a vehicle controller and second perception data of an external controller, performing spatiotemporal alignment processing to generate fused perception data, constructing an environment representation model, and generating a target path in combination with a vehicle dynamics constraint and a communication time delay compensation mechanism. The application expands the perception range through multi-source perception data fusion, improves the safety of path planning, reduces the influence of communication delay through the time delay compensation mechanism, and enhances the robustness of path planning.

Description

Technical Field

[0001] This invention relates to the field of path planning technology, and in particular to a path planning method and system based on collaborative perception of vehicle controllers. Background Technology

[0002] With the rapid development of intelligent connected vehicle technology, vehicle path planning, as a core function of autonomous driving systems, has become a key research focus in both academia and industry. Traditional path planning methods primarily rely on single-vehicle perception systems, using onboard sensors such as LiDAR, cameras, and millimeter-wave radar to collect environmental information and perform obstacle detection, path generation, and obstacle avoidance decisions based on this information. With the development of vehicle-to-everything (V2X) technology, multi-vehicle collaborative perception and path planning have become crucial means to improve the safety and efficiency of autonomous driving, especially in complex traffic scenarios where single-vehicle perception often suffers from blind spots and uncertainties. Summary of the Invention

[0003] The embodiments of the present invention provide a path planning method and system based on vehicle controller cooperative perception, which can solve the problems in the prior art.

[0004] A first aspect of the present invention provides a path planning method based on cooperative perception of a vehicle controller, comprising: The system acquires first perception data collected by the vehicle controller, the first perception data including environmental information acquired by the vehicle's own sensors; and receives second perception data from at least one external controller with which the vehicle has established a communication connection, the second perception data including environmental information acquired by the corresponding sensors of the external controller. The first and second sensing data are spatiotemporally aligned by mapping data points with overlapping sensing areas in the first and second sensing data to a unified spatiotemporal coordinate system to generate fused sensing data. An environmental representation model is constructed based on the fused sensing data, which represents the distribution of obstacles and passable areas in the environment surrounding the vehicle. Based on the obstacle distribution and passable area in the environmental characterization model, and combined with vehicle dynamics constraints and controller communication delay compensation mechanism, a set of candidate paths is generated. A target path that satisfies safety and efficiency constraints is selected from the set of candidate paths. Based on the target path, a vehicle control command is generated and sent to the vehicle actuator to control the vehicle to travel along the target path.

[0005] Spatiotemporal alignment processing is performed on the first and second sensing data. This involves mapping data points with overlapping sensing regions in the first and second sensing data to a unified spatiotemporal coordinate system to generate fused sensing data, including: Extract the timestamp and spatial location information of each data point in the first sensing data, and extract the timestamp and spatial location information of each data point in the second sensing data; Based on the relative pose relationship between the vehicle controller and the external controller, the spatial position information of each data point in the second perception data is transformed into a spatial coordinate system with the vehicle controller as the origin, to obtain the transformed second perception data; Based on the timestamps of data points in the first sensing data and the timestamps of data points in the converted second sensing data, time synchronization processing is performed on data points whose time difference is within a preset time window to obtain time-synchronized first sensing data and time-synchronized second sensing data. The first sensing data after time synchronization and the second sensing data after time synchronization are fused together to generate the fused sensing data.

[0006] Before constructing an environmental representation model based on the fused perception data, the method further includes determining the distribution of obstacles and passable areas in the environment surrounding the vehicle: The fused sensing data is spatially gridded, dividing the vehicle's surrounding environment into multiple spatial grid units. Based on the spatial location information of the data points in the fused sensing data, the spatial grid unit to which each data point belongs is determined, and the number of data points contained in each spatial grid unit is counted. Based on the number of data points contained in each spatial grid cell and the environmental feature information of the data points, the occupancy probability value of each spatial grid cell is calculated. The occupancy probability value represents the probability that the spatial grid cell is occupied by an obstacle. Spatial grid cells with an occupation probability value higher than a preset occupation threshold are marked as obstacle areas, and spatial grid cells with an occupation probability value lower than a preset passable threshold are marked as passable areas. Clustering is performed on adjacent spatial grid cells in the obstacle region to identify obstacle boundaries, and the spatial extent of the obstacle region is determined based on the obstacle boundaries; Based on the spatial distribution of the passable area and the spatial extent of the obstacle area, the environmental representation model is constructed, which includes the topological relationship between the obstacle distribution and the passable area.

[0007] Clustering adjacent spatial grid cells within the obstacle region to identify obstacle boundaries, and determining the spatial extent of the obstacle region based on these boundaries includes: Traverse each spatial grid cell in the obstacle region and determine whether adjacent spatial grid cells are also marked as obstacle regions; when adjacent spatial grid cells are both marked as obstacle regions, classify the two adjacent spatial grid cells into the same obstacle cluster; After clustering all spatial grid cells in the obstacle region, multiple obstacle clusters are obtained, and each obstacle cluster contains several spatially connected spatial grid cells; The spatial grid cells adjacent to the passable area in the obstacle cluster are taken as boundary grid cells, and the spatial coordinate information of the boundary grid cells is extracted; Based on the spatial coordinate information of the boundary grid cells, the boundary contour curve of the obstacle cluster is fitted, and the closed area enclosed by the boundary contour curve is taken as the spatial range of the obstacle region.

[0008] Based on the obstacle distribution and passable area in the environmental characterization model, and combined with vehicle dynamics constraints and controller communication delay compensation mechanisms, a candidate path set is generated, including: Based on the communication delay measurement between the vehicle controller and the external controller, the change in vehicle position and the dynamic change in obstacle distribution during the communication delay period are predicted; Based on the position change, the vehicle's current position is corrected for time delay to obtain the time-delay-compensated vehicle position. Similarly, based on the dynamic change, the spatial location information of the obstacle distribution is corrected for time delay to obtain the time-delay-compensated obstacle distribution. Obtain the vehicle's dynamic parameters, including the vehicle's maximum steering angle rate constraint and maximum acceleration constraint; Based on the vehicle's position after time delay compensation as the starting point and the preset target position as the ending point, multiple initial paths are generated within the passable area; For each initial path, a dynamic feasibility verification is performed. This involves calculating the required steering angle rate and acceleration of the vehicle at each path point on the initial path, and determining whether the steering angle rate and acceleration meet the maximum steering angle rate constraint and the maximum acceleration constraint, respectively. Collision detection is performed on the initial path that satisfies the vehicle dynamics constraints. By determining whether the initial path has spatial intersection with the time-delay-compensated obstacle distribution, the initial path that has spatial intersection with the time-delay-compensated obstacle distribution is eliminated. The initial path that passes the dynamic feasibility verification and the collision detection will be selected as a candidate path in the candidate path set.

[0009] The target path selected from the candidate path set that satisfies both security and efficiency constraints includes: For each candidate path in the candidate path set, calculate the minimum safe distance between the candidate path and the time-delay-compensated obstacle distribution, and calculate a safety assessment value based on the minimum safe distance; Obtain the path length and path curvature change rate of the candidate path, and calculate the efficiency evaluation value based on the path length and path curvature change rate; The security assessment value and the efficiency assessment value are weighted and combined to obtain a comprehensive assessment value; the candidate path with the best comprehensive assessment value is selected from the candidate path set as the target path.

[0010] Generating vehicle control commands based on the target path includes: The target path is discretized into multiple path segments, and the expected arrival time and expected travel speed are calculated for each path segment. Based on the spatial relationship between the vehicle's current position and the next path segment on the target path, the required steering angle command for the vehicle is generated; based on the vehicle's current speed and the expected speed of the next path segment, the required acceleration command for the vehicle is calculated. The steering angle command and the acceleration command are combined to generate the vehicle control command; The vehicle control command is sent to the vehicle actuator to control the vehicle to travel along the target path.

[0011] A second aspect of the present invention provides a path planning system based on cooperative perception of vehicle controllers, comprising: The first unit is configured to acquire first perception data collected by the vehicle controller, the first perception data including environmental information acquired by the vehicle's own sensors; and receive second perception data from at least one external controller that has established a communication connection with the vehicle, the second perception data including environmental information acquired by the corresponding sensors of the external controller. The second unit is used to perform spatiotemporal alignment processing on the first and second sensing data. By mapping data points with overlapping sensing areas in the first and second sensing data to a unified spatiotemporal coordinate system, fused sensing data is generated. Based on the fused sensing data, an environmental representation model is constructed, which represents the distribution of obstacles and passable areas in the environment surrounding the vehicle. The third unit is used to generate a set of candidate paths based on the obstacle distribution and passable area in the environmental characterization model, combined with vehicle dynamics constraints and inter-controller communication delay compensation mechanism, and select a target path from the set of candidate paths that meets safety and efficiency constraints. Based on the target path, a vehicle control command is generated and sent to the vehicle actuator to control the vehicle to travel along the target path.

[0012] A third aspect of the embodiments of the present invention, An electronic device is provided, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0013] Fourth aspect of the present invention, A computer-readable storage medium is provided, having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0014] The beneficial effects of this application are as follows: By acquiring the first perception data from the vehicle's own sensors and the second perception data from the external controller's sensors, and performing spatiotemporal alignment processing to generate fused perception data, the perception range is effectively expanded, the comprehensiveness and accuracy of environmental perception are improved, and the problem of limited perception capabilities of a single vehicle is solved.

[0015] The environmental representation model built based on fused perception data can comprehensively reflect the distribution of obstacles and passable areas in the environment around the vehicle, providing a more reliable environmental information basis for path planning and improving the safety and robustness of path planning.

[0016] By combining vehicle dynamics constraints with a communication delay compensation mechanism between controllers to generate a set of candidate paths, and selecting a target path that satisfies both safety and efficiency constraints, the system takes into account the limitations of actual vehicle driving performance and compensates for the impact of communication delays during multi-controller collaboration, thereby achieving safer and more efficient autonomous vehicle driving control. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the path planning method based on collaborative perception of the vehicle controller according to an embodiment of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0020] Figure 1 This is a flowchart illustrating the path planning method based on cooperative perception of the vehicle controller according to an embodiment of the present invention, as shown below. Figure 1 As shown, the method includes: The system acquires first perception data collected by the vehicle controller, the first perception data including environmental information acquired by the vehicle's own sensors; and receives second perception data from at least one external controller with which the vehicle has established a communication connection, the second perception data including environmental information acquired by the corresponding sensors of the external controller. The first and second sensing data are spatiotemporally aligned by mapping data points with overlapping sensing areas in the first and second sensing data to a unified spatiotemporal coordinate system to generate fused sensing data. An environmental representation model is constructed based on the fused sensing data, which represents the distribution of obstacles and passable areas in the environment surrounding the vehicle. Based on the obstacle distribution and passable area in the environmental characterization model, and combined with vehicle dynamics constraints and controller communication delay compensation mechanism, a set of candidate paths is generated. A target path that satisfies safety and efficiency constraints is selected from the set of candidate paths. Based on the target path, a vehicle control command is generated and sent to the vehicle actuator to control the vehicle to travel along the target path.

[0021] In one optional implementation, spatiotemporal alignment processing is performed on the first sensing data and the second sensing data. This is achieved by mapping data points with overlapping sensing regions in the first and second sensing data to a unified spatiotemporal coordinate system to generate fused sensing data, including: Extract the timestamp and spatial location information of each data point in the first sensing data, and extract the timestamp and spatial location information of each data point in the second sensing data; Based on the relative pose relationship between the vehicle controller and the external controller, the spatial position information of each data point in the second perception data is transformed into a spatial coordinate system with the vehicle controller as the origin, to obtain the transformed second perception data; Based on the timestamps of data points in the first sensing data and the timestamps of data points in the converted second sensing data, time synchronization processing is performed on data points whose time difference is within a preset time window to obtain time-synchronized first sensing data and time-synchronized second sensing data. The first sensing data after time synchronization and the second sensing data after time synchronization are fused together to generate the fused sensing data.

[0022] In the technical solution of spatiotemporal alignment processing of the first and second perception data, the vehicle controller can be the vehicle's own sensor system control unit, which collects the first perception data of the surrounding environment; the external controller can be a roadside unit or the sensor system of other vehicles, which collects and provides the second perception data. The perception data collected by the two may have time differences and different spatial coordinate systems, requiring spatiotemporal alignment processing for effective fusion.

[0023] The vehicle controller extracts the timestamp and spatial location information of each data point from the first sensing data, and simultaneously extracts the timestamp and spatial location information of each data point from the second sensing data. Taking LiDAR point cloud data as an example, each point cloud data point typically contains three-dimensional coordinate information (x, y, z) and an acquisition timestamp t. For example, a data point in the first sensing data might be (2.3m, 1.5m, 0.5m, 1623456789.123s), representing the spatial location of the point in the vehicle controller's coordinate system and its acquisition time. Similarly, a data point in the second sensing data might be (5.2m, 3.7m, 0.2m, 1623456789.234s), representing the spatial location of the point in the external controller's coordinate system and its acquisition time.

[0024] Since the first and second sensing data originate from different controllers, their spatial coordinate systems are typically different. The system transforms the spatial position information of each data point in the second sensing data to a spatial coordinate system with the vehicle controller as the origin, based on the relative pose relationship between the vehicle controller and the external controller. The relative pose relationship can be obtained via GPS, RTK, or V2X communication, including relative translation vectors and rotation matrices. For example, if the external controller's position offset relative to the vehicle controller is (10m, 5m, 0m) with no rotation, then the point with coordinates (5.2m, 3.7m, 0.2m) in the second sensing data will be transformed to (15.2m, 8.7m, 0.2m) in the vehicle controller's coordinate system. In this way, the transformed second sensing data and the first sensing data are in the same spatial coordinate system.

[0025] Even if the spatial coordinates have been unified, there may still be differences in the acquisition times of the two sets of data. The system performs time synchronization processing on the data points whose time differences are within a preset time window based on the timestamps of the data points in the first perception data and the timestamps of the data points in the converted second perception data. The preset time window can be set to 100 milliseconds, which means that if the timestamp difference between two data points does not exceed 100 milliseconds, they are considered to be time-synchronizable. The time synchronization method can adopt interpolation or extrapolation techniques. Taking interpolation as an example, if the vehicle controller acquires positions p1 and p3 at time points t1 and t3, and needs to synchronize with the data acquired by an external controller at time point t2 (t1 < t2 < t3), then the position p2 at time point t2 can be calculated by linear interpolation as p2 = p1 + (p3 - p1) * (t2 - t1) / (t3 - t1). Through this method, the system generates the first perception data after time synchronization and the second perception data after time synchronization.

[0026] The first perception data after time synchronization and the converted second perception data are fused to generate fused perception data. During the fusion process, for the data points in the overlapping perception area, methods such as weighted average and Kalman filtering can be used for fusion. For example, if the position of a certain obstacle in the first perception data is (20.5m, 15.3m) with a confidence level of 0.85, and the position of the same obstacle in the second perception data is (20.7m, 15.1m) with a confidence level of 0.90, then the fused position can be calculated as: (20.5 * 0.85 + 20.7 * 0.90) / (0.85 + 0.90) = 20.6m, (15.3 * 0.85 + 15.1 * 0.90) / (0.85 + 0.90) = 15.2m, that is, (20.6m, 15.2m), and the fused confidence level is 0.85 + 0.90 - 0.85 * 0.90 = 0.985, which is higher than the confidence level of the perception data from a single source. For the data points in the non-overlapping area, they are directly added to the fused perception data to expand the perception range.

[0027] Through the above spatio-temporal alignment and fusion processing, the system effectively integrates the perception data from different sources to generate fused perception data with a wider coverage range and higher accuracy, providing more comprehensive and accurate environmental perception information for vehicle control decisions, and improving the safety and reliability of the autonomous driving system.

[0028] In an optional implementation manner, before constructing an environmental representation model based on the fused perception data, the method further includes determining the obstacle distribution and passable areas in the vehicle's surrounding environment: The fused sensing data is spatially gridded, dividing the vehicle's surrounding environment into multiple spatial grid units. Based on the spatial location information of the data points in the fused sensing data, the spatial grid unit to which each data point belongs is determined, and the number of data points contained in each spatial grid unit is counted. Based on the number of data points contained in each spatial grid cell and the environmental feature information of the data points, the occupancy probability value of each spatial grid cell is calculated. The occupancy probability value represents the probability that the spatial grid cell is occupied by an obstacle. Spatial grid cells with an occupation probability value higher than a preset occupation threshold are marked as obstacle areas, and spatial grid cells with an occupation probability value lower than a preset passable threshold are marked as passable areas. Clustering is performed on adjacent spatial grid cells in the obstacle region to identify obstacle boundaries, and the spatial extent of the obstacle region is determined based on the obstacle boundaries; Based on the spatial distribution of the passable area and the spatial extent of the obstacle area, the environmental representation model is constructed, which includes the topological relationship between the obstacle distribution and the passable area.

[0029] Spatial meshing of fused sensing data is a fundamental step in achieving environmental perception. Specifically, the three-dimensional space surrounding the vehicle is divided into multiple mesh cells, each with dimensions of 0.2m × 0.2m × 0.2m. This partitioning method ensures both spatial representation accuracy and computational efficiency. In practical applications, the mesh size can be adjusted according to the accuracy requirements of the sensing task and computational resource limitations. For each spatial mesh cell, the system assigns an occupancy status flag, indicating whether an obstacle exists within that mesh cell.

[0030] Based on the spatial location information of each data point in the fused sensing data, the spatial grid cell to which each data point belongs is determined. For example, for a radar point cloud data point with coordinates (3.25, 4.15, 0.85), if the grid size is 0.2 meters, then this point belongs to the grid cell with index (16, 20, 4). The system counts the number of data points falling into each grid cell; this number is an important basis for determining the occupancy status of the grid cell. For example, when a grid cell contains 15 LiDAR point cloud data points, the grid is more likely to be occupied by an obstacle; while when the number of point cloud data points is only 2, the grid is less likely to be occupied.

[0031] Based on the number of data points contained in each spatial grid cell and its environmental characteristics, the occupancy probability value of each grid cell is calculated. Environmental characteristics include, but are not limited to, the reflection intensity, height distribution, and geometric features of the data points. For example, for a given grid cell containing 10 data points, with an average reflection intensity of 85 (assuming an intensity range of 0-100) and a height variation standard deviation of 0.02 meters, the occupancy probability value of that grid cell can be calculated by combining these characteristics. The specific calculation process involves converting the number of data points into a base occupancy probability; for example, 10 points correspond to a base probability of 0.6. Then, a weighted adjustment is made based on the reflection intensity and height characteristics. If the reflection intensity is high and the height distribution is concentrated, the probability of an obstacle is considered higher, and the occupancy probability value can be adjusted to 0.85. Conversely, if the reflection intensity is low and the height distribution is scattered, it may be due to sensor noise, and the occupancy probability value is adjusted to a lower value, such as 0.35.

[0032] In practical systems, the calculation of occupancy probability values ​​also considers the influence of historical frame data, and improves the stability of probability estimation through temporal filtering. For example, if the occupancy probability of a grid calculated in the current frame is 0.75, while the occupancy probability of the same grid in the previous frame was 0.65, then an exponential filtering method with a weight of 0.8 can be used to update the occupancy probability, resulting in a new occupancy probability of 0.73.

[0033] Spatial grid cells with an occupancy probability value higher than a preset occupancy threshold are marked as obstacle areas. In a specific implementation, the occupancy threshold can be set to 0.7, meaning that when the occupancy probability value of a grid cell exceeds 0.7, the grid cell is determined to be occupied by an obstacle. Similarly, spatial grid cells with an occupancy probability value lower than a preset passable threshold are marked as passable areas. The passable threshold can be set to 0.3, meaning that when the occupancy probability value is lower than 0.3, the grid cell is considered passable. Grid cells with an occupancy probability value between 0.3 and 0.7 are marked as uncertain areas, indicating that the system cannot determine whether there is an obstacle in that area.

[0034] In specific application scenarios, preset thresholds can be dynamically adjusted to adapt to different environmental conditions. For example, on well-lit open roads, the occupancy threshold can be increased to 0.75 and the passable threshold can be decreased to 0.25 to reduce uncertain areas; while in complex environments such as tunnels or rainy / snowy weather, the occupancy threshold can be decreased to 0.65 and the passable threshold increased to 0.35 to increase the system's safety margin.

[0035] Through the above steps, the system can effectively identify the distribution of obstacles and passable areas in the vehicle's surrounding environment, providing a foundation for subsequent environmental representation model construction. For example, in one perception process, the system processed a total of 32,768 spatial grid cells, of which 8,192 were marked as obstacle areas, accounting for 25%; 20,480 were marked as passable areas, accounting for 62.5%; and the remaining 4,096 were marked as uncertain areas, accounting for 12.5%. These marking results are directly used to construct the environmental representation model, assisting the vehicle in path planning and obstacle avoidance.

[0036] Obstacle zones and uncertain areas are typically treated as no-entry zones in path planning, with vehicles prioritizing passage through passable areas. For example, when a 3m x 2m obstacle zone is detected 5 meters ahead, the system plans a detour route to ensure driving safety. This grid-based environmental representation method effectively supports the vehicle's autonomous navigation capabilities.

[0037] In one optional implementation, clustering adjacent spatial grid cells within the obstacle region to identify obstacle boundaries, and determining the spatial extent of the obstacle region based on the obstacle boundaries, includes: Traverse each spatial grid cell in the obstacle region and determine whether adjacent spatial grid cells are also marked as obstacle regions; when adjacent spatial grid cells are both marked as obstacle regions, classify the two adjacent spatial grid cells into the same obstacle cluster; After clustering all spatial grid cells in the obstacle region, multiple obstacle clusters are obtained, and each obstacle cluster contains several spatially connected spatial grid cells; The spatial grid cells adjacent to the passable area in the obstacle cluster are taken as boundary grid cells, and the spatial coordinate information of the boundary grid cells is extracted; Based on the spatial coordinate information of the boundary grid cells, the boundary contour curve of the obstacle cluster is fitted, and the closed area enclosed by the boundary contour curve is taken as the spatial range of the obstacle region.

[0038] When clustering adjacent spatial grid cells within an obstacle region, the system first acquires fused sensing data based on pre-divided spatial grid cells. These spatial grid cells can be cubes in three-dimensional space or square grids on a two-dimensional plane, with side lengths ranging from 0.2 meters to 0.5 meters, adjustable according to the specific application scenario. The system marks each spatial grid cell as either an obstacle region or a passable region, based on the detected point cloud density, radar reflection intensity, or camera image features within that grid cell. For example, if the number of point clouds within a grid cell exceeds a predetermined threshold (e.g., 20 points), the grid cell is marked as an obstacle region; otherwise, it is marked as a passable region.

[0039] The clustering process employs a region growing algorithm to group adjacent grid cells marked as obstacle regions. Specifically, a queue is set up, initially selecting an obstacle grid cell as a seed point and adding it to the queue. Each time, a grid cell is taken from the queue, and its six adjacent grid cells (up, down, left, right, front, and back) are checked. If an adjacent grid cell is also marked as an obstacle region and has not yet been assigned to any cluster, it is added to the current cluster and then to the queue. This process continues until the queue is empty, completing the processing of one cluster. The system then continues to select unprocessed obstacle grid cells as new seed points, repeating the above process until all obstacle grid cells have been assigned to their respective clusters.

[0040] In real-world applications, such as a 10m x 10m area, there may be 5-10 different obstacle clusters, each representing an independent obstacle entity, such as a wall, pillar, or vehicle. To eliminate the impact of noise, the system can set a minimum cluster size threshold. For example, if a cluster contains fewer than 3 grid cells, it can be considered noise and ignored.

[0041] When identifying obstacle boundaries, the system performs a boundary extraction algorithm for each obstacle cluster. This algorithm examines the neighborhood state of each grid cell in the cluster. If an obstacle grid cell has at least one neighboring grid cell marked as a passable area, then that grid cell is considered a boundary grid cell. The system saves the coordinates of all boundary grid cells as a set of boundary points for the obstacle. For each obstacle cluster, dozens to hundreds of boundary points may be identified, depending on the size and shape complexity of the obstacle.

[0042] To determine the geometry and spatial extent of the obstacle region, the system applies a polygon fitting algorithm to the set of boundary points. This algorithm first calculates the convex hull of the boundary point set, then optimizes the convex hull, removing redundant points to form a polygon that accurately represents the obstacle's shape. For obstacles with complex shapes, the system can use a piecewise polygon fitting method, decomposing the obstacle into a combination of multiple simple geometric shapes. For example, an L-shaped obstacle can be decomposed into a combination of two rectangles. The system records the vertex coordinates, center position, length, width, and height information of each obstacle polygon, forming a geometric representation of the obstacle.

[0043] In determining the obstacle space range, the system considers the actual size of the obstacle and a safety margin. The safety margin is typically set to 0.3 to 0.5 meters to ensure a safe distance between the robot or mobile device and the obstacle. For example, if an obstacle is detected that is 1.8 meters long and 0.8 meters wide, after considering a safety margin of 0.3 meters, its space range will expand to a rectangular area that is 2.4 meters long and 1.4 meters wide.

[0044] When constructing the environmental representation model, the system comprehensively considers the topological relationship between obstacle distribution and traversable areas. The system first represents the entire environment as a two-dimensional or three-dimensional mesh map in a global coordinate system. In this map, the location, shape, and extent of each obstacle are marked, while the distribution of traversable areas is determined. Traversable areas are typically represented as a collection of free mesh cells, or as polygonal regions containing no obstacles.

[0045] The system further analyzes the spatial relationships between obstacles, identifying features such as passageways, intersections, and dead ends. A passageway is a narrow, passable area between two obstacles. When the distance between two obstacles is less than a preset threshold (e.g., 1.5 meters) but greater than the minimum passable width (e.g., 0.8 meters), the area is marked as a passageway. An intersection is the junction of multiple passageways, typically appearing as an open, passable area surrounded by multiple passageways. The system calculates the area and connectivity of each intersection; for example, an intersection with an area of ​​4 square meters may connect passageways in three different directions.

[0046] The environmental representation model is ultimately presented in the form of a topological map and a geometric map. The topological map describes the connectivity of traversable areas, where nodes represent intersections or key areas, and edges represent pathways connecting these areas. The geometric map contains a precise geometric representation of all obstacles and traversable areas in the environment, including boundary coordinates, area size, and shape features. The combination of these two representations provides comprehensive environmental information to support subsequent path planning and navigation decisions.

[0047] In one optional implementation, based on the obstacle distribution and passable area in the environmental characterization model, and combined with vehicle dynamics constraints and a controller-to-controller communication delay compensation mechanism, a candidate path set is generated, including: Based on the communication delay measurement between the vehicle controller and the external controller, the change in vehicle position and the dynamic change in obstacle distribution during the communication delay period are predicted; Based on the position change, the vehicle's current position is corrected for time delay to obtain the time-delay-compensated vehicle position. Similarly, based on the dynamic change, the spatial location information of the obstacle distribution is corrected for time delay to obtain the time-delay-compensated obstacle distribution. Obtain the vehicle's dynamic parameters, including the vehicle's maximum steering angle rate constraint and maximum acceleration constraint; Based on the vehicle's position after time delay compensation as the starting point and the preset target position as the ending point, multiple initial paths are generated within the passable area; For each initial path, a dynamic feasibility verification is performed. This involves calculating the required steering angle rate and acceleration of the vehicle at each path point on the initial path, and determining whether the steering angle rate and acceleration meet the maximum steering angle rate constraint and the maximum acceleration constraint, respectively. Collision detection is performed on the initial path that satisfies the vehicle dynamics constraints. By determining whether the initial path has spatial intersection with the time-delay-compensated obstacle distribution, the initial path that has spatial intersection with the time-delay-compensated obstacle distribution is eliminated. The initial path that passes the dynamic feasibility verification and the collision detection will be selected as a candidate path in the candidate path set.

[0048] In an autonomous vehicle path planning method, an environmental representation model is first established, which includes information on obstacle distribution and passable areas. The system acquires environmental data through onboard sensors, detects and locates obstacles, determines their spatial position, speed, and orientation, and simultaneously identifies elements such as road boundaries and lane lines to delineate passable areas.

[0049] To address the communication latency issue between the vehicle controller and external controllers, this implementation introduces a latency compensation mechanism. The system continuously monitors the difference between the sending and receiving timestamps of communication data packets and calculates the real-time communication latency measurement. Assuming the currently measured communication latency is 150 milliseconds, the system predicts that the vehicle will move approximately 1.25 meters during this latency period based on the vehicle's current motion state (e.g., speed 30 km / h, acceleration 1.2 m / s²). Simultaneously, for dynamic obstacles (e.g., an oncoming vehicle with a relative speed of 20 km / h), the system calculates that its position change during the latency period is approximately 0.83 meters.

[0050] The system combines the vehicle's current position coordinates (e.g., [100.0, 50.0] meters in the local coordinate system) with the predicted position change to obtain the time-delay-compensated vehicle position [101.25, 50.0] meters. Similarly, for the original obstacle position [120.0, 51.5] meters, the system calculates the time-delay-compensated obstacle position [119.17, 51.5] meters. This compensation ensures that path planning is based on the accurate environmental conditions at future execution times.

[0051] Obtaining vehicle dynamics parameters is a crucial step in ensuring the generation of an executable path. The system reads vehicle specification parameters, including maximum steering rate constraints (e.g., 30 degrees / second) and maximum acceleration constraints (e.g., 2.5 m / s²). These constraints are used for subsequent path verification to ensure that the generated path is executable within the vehicle's physical capabilities.

[0052] Based on the time-delay compensated starting position [101.25, 50.0] meters and the preset ending position (e.g., [200.0, 52.0] meters), the system generates multiple initial paths within the passable area. Path generation employs an improved RRT algorithm, using the vehicle's current position as the root node, and exploring the environmental space through random sampling and expanding the tree structure. The system generates five initial candidate paths, each consisting of a series of discrete points with a 0.5-meter interval between points, ensuring that the path accuracy meets control requirements.

[0053] The system is based on a pure tracking control model, calculating the required steering angle rate and acceleration at each point along the path. Taking the first path as an example, the system detects a sharp turn 8.5 meters from the starting point and calculates a required steering angle rate of 35 degrees / second, exceeding the vehicle's maximum capacity of 30 degrees / second. Therefore, this path is marked as dynamically infeasible. Similarly, the second path requires an acceleration of 2.8 m / s² at 15 meters from the starting point, exceeding the maximum acceleration constraint of 2.5 m / s², and is also eliminated.

[0054] For paths that pass dynamic verification, the system further performs collision detection. An extended vehicle boundary method is used to extend the vehicle outline by a safe distance of 0.3 meters, forming a vehicle-occupied area. The system samples points along each path at 0.5-meter intervals, calculating the minimum distance between the vehicle-occupied area and the time-delay-compensated obstacle distribution at each point. When this distance is less than the preset safety threshold of 0.5 meters, a collision risk is identified. In the remaining three paths, the third path has a minimum distance of 0.3 meters to a static obstacle at a distance of 23 meters from the starting point, which is below the safety threshold and is therefore identified as an unsafe path.

[0055] After dynamic feasibility verification and collision detection screening, only the fourth and fifth initial paths met all the constraints. These two paths were retained as a candidate path set for use by the subsequent path evaluation and selection module. The path data is stored in the form of a three-dimensional array, containing information such as path point coordinates, path point tangential angles, and path curvature, providing a complete reference for subsequent control execution.

[0056] This implementation method generates a high-quality candidate path set by comprehensively considering communication latency compensation, vehicle dynamics constraints, and safety requirements, providing a reliable guarantee for the safe and efficient navigation of autonomous vehicles in complex environments. The system fully applies vehicle control theory and path planning technology during processing, ensuring the real-time nature and executability of the planning results.

[0057] In one optional implementation, selecting a target path from the candidate path set that satisfies both security and efficiency constraints includes: For each candidate path in the candidate path set, calculate the minimum safe distance between the candidate path and the time-delay-compensated obstacle distribution, and calculate a safety assessment value based on the minimum safe distance; Obtain the path length and path curvature change rate of the candidate path, and calculate the efficiency evaluation value based on the path length and path curvature change rate; The security assessment value and the efficiency assessment value are weighted and combined to obtain a comprehensive assessment value; the candidate path with the best comprehensive assessment value is selected from the candidate path set as the target path.

[0058] For each candidate path in the candidate path set, the system first calculates the minimum safe distance between the candidate path and the time-delay-compensated obstacle distribution. Specifically, the system obtains the current vehicle position and speed information, including position coordinates (x0, y0) and speed v0, for example, the current position is (10.5m, 5.2m) and the speed is 15m / s. The system obtains the position, speed, and acceleration information of surrounding obstacles from the perception module, such as obstacle A being located at (30.2m, 4.8m), with a speed of 10m / s and an acceleration of 1.2m / s². Considering the system processing delay, assumed to be 200ms, the system performs time-delay compensation on the obstacle position, updating the predicted position of obstacle A to (32.4m, 4.8m). For path 1 in the candidate path set, the system uses a discrete point sampling method to sample 20 points at equal intervals on the path, calculates the Euclidean distance from each sampling point to the obstacle contour after time delay compensation, finds the minimum value, and records it as the minimum safe distance d_min. For example, d_min is 3.5m.

[0059] When calculating the safety assessment value based on the minimum safe distance, the system sets the safety threshold d_safe to 3.0m. When d_min is greater than d_safe, the safety assessment value s_safety increases with d_min, but the rate of increase gradually decreases. Specifically, the calculation of the safety assessment value takes into account the influence of vehicle speed, and the safety distance requirement is higher at high speeds. For example, when the vehicle speed is 15m / s, the minimum safe distance for path 1 is 3.5m, and the calculated safety assessment value s_safety is 0.78.

[0060] When obtaining the path length and curvature change rate of a candidate path, the system calculates the total length of the candidate path, l_path, for example, the length of path 1 is 45.6m. Simultaneously, the system calculates the curvature values ​​at discrete points along the path, obtaining the curvature at each point by fitting a circle to three adjacent points, and calculating the sum of squared differences between adjacent curvatures to obtain the path curvature change rate, c_rate. For example, the curvature change rate of path 1 is 0.024.

[0061] When calculating the efficiency evaluation value based on path length and the rate of change of path curvature, the system sets the reference path length l_ref to 40.0m and the reference rate of change of curvature c_ref to 0.020. The efficiency evaluation value s_efficiency decreases as the path length increases and as the rate of change of curvature increases. For example, for path 1, the calculated efficiency evaluation value s_efficiency is 0.72.

[0062] When the security assessment value and the efficiency assessment value are weighted and combined to obtain the comprehensive assessment value, the system sets the security weight w_safety to 0.6 and the efficiency weight w_efficiency to 0.4, and calculates the comprehensive assessment value s_total = w_safety × s_safety + w_efficiency × s_efficiency. For path 1, the comprehensive assessment value s_total is 0.756.

[0063] The system performs the above calculations on all paths in the candidate path set. For example, for path 2 (length 42.1m, minimum safe distance 4.2m, curvature change rate 0.018), the calculated safety evaluation value is 0.85, the efficiency evaluation value is 0.81, and the comprehensive evaluation value is 0.834. For path 3 (length 47.5m, minimum safe distance 2.8m, curvature change rate 0.030), the calculated safety evaluation value is 0.62, the efficiency evaluation value is 0.65, and the comprehensive evaluation value is 0.632.

[0064] When selecting the candidate path with the best overall evaluation value from the candidate path set as the target path, the system compares the overall evaluation values ​​of all candidate paths and selects the path with the highest value. In this example, path 2 has the highest overall evaluation value of 0.834, and is therefore selected as the target path.

[0065] In practical applications, the system can dynamically adjust the weights of safety and efficiency based on different road environments. For example, in narrow roads or high-speed driving situations, the safety weight can be increased to 0.7 or higher; in open roads or scenarios requiring rapid passage, the efficiency weight can be increased to 0.5 or higher. The system can also set hard thresholds for safety constraints, such as directly excluding candidate paths from the comprehensive evaluation stage if the minimum safe distance is lower than the absolute safety threshold (e.g., 2.0m).

[0066] When calculating the efficiency evaluation value, the system also considers the smoothness index of the path, which is quantified by the change in direction angle between adjacent points on the path. The smoother the path, the smaller the change in direction angle, and the higher the corresponding efficiency evaluation value. For example, for a path with a total change in direction angle of 45 degrees, its smoothness score is 0.85; for a path with a total change in direction angle of 75 degrees, its smoothness score is 0.65.

[0067] In actual deployment, the system adjusts the evaluation parameters according to the characteristics of different vehicle models. For example, for vehicles with large turning radius limitations, the system increases the weight of the rate of curvature change in the efficiency evaluation; for vehicles with good acceleration performance, the system correspondingly reduces the weight of path length in the efficiency evaluation. Through this adaptive parameter adjustment mechanism, the system can adapt to different vehicle characteristics and generate the optimal path best suited to the current vehicle dynamics.

[0068] In one optional implementation, generating vehicle control commands based on the target path includes: The target path is discretized into multiple path segments, and the expected arrival time and expected travel speed are calculated for each path segment. Based on the spatial relationship between the vehicle's current position and the next path segment on the target path, the required steering angle command for the vehicle is generated; based on the vehicle's current speed and the expected speed of the next path segment, the required acceleration command for the vehicle is calculated. The steering angle command and the acceleration command are combined to generate the vehicle control command; The vehicle control command is sent to the vehicle actuator to control the vehicle to travel along the target path.

[0069] This embodiment describes in detail a method for generating vehicle control commands based on a target path. This method realizes the conversion process from the target path to vehicle control commands, enabling the vehicle to travel precisely along a predetermined trajectory.

[0070] This method first discretizes the target path, converting the continuous target path into multiple discrete path segments. Specifically, an equidistant sampling method can be used, setting sampling points along the target path at fixed intervals (e.g., 0.5 meters or 1 meter), with adjacent sampling points forming a path segment. For example, for a target path with a total length of 100 meters, if a sampling interval of 1 meter is used, it can be divided into 100 path segments. Each path segment is defined by the position coordinates of its start and end points, and can be represented in the form of [(x1, y1), (x2, y2)].

[0071] For each path segment, its expected arrival time and expected speed need to be calculated. The expected arrival time can be determined based on the time planning during path planning, or it can be dynamically calculated by considering factors such as road speed limits and traffic conditions. For example, for the i-th path segment, its expected arrival time at the starting point can be set as ti, and its expected speed can be set as vi. Specifically, if the planned average vehicle speed is 10 meters per second, then the expected arrival time of the 10th path segment (approximately 10 meters from the starting point) is 1 second, the expected arrival time of the 20th path segment is 2 seconds, and so on.

[0072] The calculation of desired driving speed needs to take into account factors such as road curvature and gradient. At curves, the desired speed should be appropriately reduced to ensure safety. For example, when the radius of curvature of a road segment is detected to be less than 30 meters, the desired speed can be reduced from 10 m / s to 5 m / s. For straight road sections, a higher desired speed, such as 10 m / s, can be maintained. Furthermore, the desired speed can be appropriately reduced on uphill sections and appropriately increased on downhill sections to optimize energy consumption and driving comfort.

[0073] After discretization, the required steering angle command is calculated based on the spatial relationship between the vehicle's current position and the next path segment on the target path. The specific steps are as follows: First, obtain the vehicle's current position (xc, yc) and orientation angle θc using the vehicle positioning system. Then, find the path segment closest to the current position on the discretized target path, denoted as the current path segment. Next, determine the path segment to be reached, typically the nth path segment after the current path segment (n can be set as a pre-aiming distance parameter, such as 3 or 5), with its endpoint coordinates (xn, yn).

[0074] Calculate the relative coordinates (Δx, Δy) of the preview point (xn, yn) with respect to the vehicle's current position, where Δx = xn - xc and Δy = yn - yc. Transform the relative coordinates from the world coordinate system to the vehicle coordinate system, obtaining (Δxv, Δyv). Calculate the steering angle command based on the transformed coordinates. For example, a simplified forward-looking model can be used, where the steering angle δ can be approximated as the ratio of Δyv to the preview distance L, i.e., δ = arctan(2Lcar·Δyv / L²), where Lcar is the vehicle's wheelbase and L is the straight-line distance from the current position to the preview point.

[0075] In practical applications, the calculation of steering angle commands can be further optimized. For example, when it is predicted that the vehicle deviates from the target path by more than 0.3 meters, the steering angle can be increased to quickly return to the path; when the vehicle speed exceeds 20 meters per second, the steering angle can be appropriately reduced to avoid excessive lateral acceleration.

[0076] Simultaneously, based on the vehicle's current speed and the expected speed for the next path segment, the required acceleration command is calculated. The vehicle's current speed, vc, and the expected speed, vn, for the next path segment are obtained. The speed difference Δv = vn - vc is calculated. Based on the speed difference and the expected time interval Δt (e.g., 1 second) to reach the desired speed, the required acceleration a = Δv / Δt is calculated.

[0077] To avoid sudden acceleration or deceleration, the calculated acceleration can be limited. For example, the acceleration can be limited to the range of [-2, 2] m / s², meaning the maximum deceleration cannot exceed 2 m / s² and the maximum acceleration cannot exceed 2 m / s². Furthermore, the acceleration limit can be dynamically adjusted based on road conditions and vehicle load. For instance, on slippery surfaces, the acceleration limit can be reduced to [-1, 1] m / s².

[0078] The calculated steering angle command δ and acceleration command a are combined to generate vehicle control commands. Specifically, a control command data packet can be constructed, containing information such as the steering angle value, acceleration value, timestamp, and checksum. For example, when the calculated steering angle is 0.1 radians (approximately 5.7 degrees) and the acceleration is 0.5 m / s², the generated control command data packet could be {steering_angle: 0.1, acceleration: 0.5, timestamp: 1589960558, checksum:0xABCD}.

[0079] Finally, the generated vehicle control commands are sent to the vehicle actuators via the in-vehicle communication network. The vehicle actuators include the steering system, braking system, and drive system. The steering system receives steering angle commands and controls the steering mechanism; the braking system and drive system work together according to acceleration commands to achieve vehicle acceleration and deceleration control. For example, when the acceleration command is positive (e.g., 0.5 m / s²), the drive system increases its output torque; when the acceleration command is negative (e.g., -1.0 m / s²), the braking system applies appropriate braking force.

[0080] Through the above steps, the conversion from the target path to vehicle control commands is realized, enabling the vehicle to travel smoothly and accurately along the planned target path.

[0081] This invention provides a path planning system based on vehicle controller cooperative perception, comprising: The first unit is configured to acquire first perception data collected by the vehicle controller, the first perception data including environmental information acquired by the vehicle's own sensors; and receive second perception data from at least one external controller that has established a communication connection with the vehicle, the second perception data including environmental information acquired by the corresponding sensors of the external controller. The second unit is used to perform spatiotemporal alignment processing on the first and second sensing data. By mapping data points with overlapping sensing areas in the first and second sensing data to a unified spatiotemporal coordinate system, fused sensing data is generated. Based on the fused sensing data, an environmental representation model is constructed, which represents the distribution of obstacles and passable areas in the environment surrounding the vehicle. The third unit is used to generate a set of candidate paths based on the obstacle distribution and passable area in the environmental characterization model, combined with vehicle dynamics constraints and inter-controller communication delay compensation mechanism, and select a target path from the set of candidate paths that meets safety and efficiency constraints. Based on the target path, a vehicle control command is generated and sent to the vehicle actuator to control the vehicle to travel along the target path.

[0082] A third aspect of the present invention provides an electronic device, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0083] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0084] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0085] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A path planning method based on cooperative perception of vehicle controllers, characterized in that, include: The system acquires first perception data collected by the vehicle controller, the first perception data including environmental information acquired by the vehicle's own sensors; and receives second perception data from at least one external controller with which the vehicle has established a communication connection, the second perception data including environmental information acquired by the corresponding sensors of the external controller. The first and second sensing data are spatiotemporally aligned by mapping data points with overlapping sensing areas in the first and second sensing data to a unified spatiotemporal coordinate system to generate fused sensing data. An environmental representation model is constructed based on the fused sensing data, which represents the distribution of obstacles and passable areas in the environment surrounding the vehicle. Based on the obstacle distribution and passable area in the environmental characterization model, and combined with vehicle dynamics constraints and controller communication delay compensation mechanism, a set of candidate paths is generated. A target path that satisfies safety and efficiency constraints is selected from the set of candidate paths. Based on the target path, a vehicle control command is generated and sent to the vehicle actuator to control the vehicle to travel along the target path.

2. The method according to claim 1, characterized in that, Spatiotemporal alignment processing is performed on the first and second sensing data. This involves mapping data points with overlapping sensing regions in the first and second sensing data to a unified spatiotemporal coordinate system to generate fused sensing data, including: Extract the timestamp and spatial location information of each data point in the first sensing data, and extract the timestamp and spatial location information of each data point in the second sensing data; Based on the relative pose relationship between the vehicle controller and the external controller, the spatial position information of each data point in the second perception data is transformed into a spatial coordinate system with the vehicle controller as the origin, to obtain the transformed second perception data; Based on the timestamps of data points in the first sensing data and the timestamps of data points in the converted second sensing data, time synchronization processing is performed on data points whose time difference is within a preset time window to obtain time-synchronized first sensing data and time-synchronized second sensing data. The first sensing data after time synchronization and the second sensing data after time synchronization are fused together to generate the fused sensing data.

3. The method according to claim 1, characterized in that, Before constructing an environmental representation model based on the fused perception data, the method further includes determining the distribution of obstacles and passable areas in the environment surrounding the vehicle: The fused sensing data is spatially gridded, dividing the vehicle's surrounding environment into multiple spatial grid units. Based on the spatial location information of the data points in the fused sensing data, the spatial grid unit to which each data point belongs is determined, and the number of data points contained in each spatial grid unit is counted. Based on the number of data points contained in each spatial grid cell and the environmental feature information of the data points, the occupancy probability value of each spatial grid cell is calculated. The occupancy probability value represents the probability that the spatial grid cell is occupied by an obstacle. Spatial grid cells with an occupation probability value higher than a preset occupation threshold are marked as obstacle areas, and spatial grid cells with an occupation probability value lower than a preset passable threshold are marked as passable areas. Clustering is performed on adjacent spatial grid cells in the obstacle region to identify obstacle boundaries, and the spatial extent of the obstacle region is determined based on the obstacle boundaries; Based on the spatial distribution of the passable area and the spatial extent of the obstacle area, the environmental representation model is constructed, which includes the topological relationship between the obstacle distribution and the passable area.

4. The method according to claim 3, characterized in that, Clustering adjacent spatial grid cells within the obstacle region to identify obstacle boundaries, and determining the spatial extent of the obstacle region based on these boundaries includes: Traverse each spatial grid cell in the obstacle region and determine whether adjacent spatial grid cells are also marked as obstacle regions; when adjacent spatial grid cells are both marked as obstacle regions, classify the two adjacent spatial grid cells into the same obstacle cluster; After clustering all spatial grid cells in the obstacle region, multiple obstacle clusters are obtained, and each obstacle cluster contains several spatially connected spatial grid cells; The spatial grid cells adjacent to the passable area in the obstacle cluster are taken as boundary grid cells, and the spatial coordinate information of the boundary grid cells is extracted; Based on the spatial coordinate information of the boundary grid cells, the boundary contour curve of the obstacle cluster is fitted, and the closed area enclosed by the boundary contour curve is taken as the spatial range of the obstacle region.

5. The method according to claim 1, characterized in that, Based on the obstacle distribution and passable area in the environmental characterization model, and combined with vehicle dynamics constraints and controller communication delay compensation mechanisms, a candidate path set is generated, including: Based on the communication delay measurement between the vehicle controller and the external controller, the change in vehicle position and the dynamic change in obstacle distribution during the communication delay period are predicted; Based on the position change, the vehicle's current position is corrected for time delay to obtain the time-delay-compensated vehicle position. Similarly, based on the dynamic change, the spatial location information of the obstacle distribution is corrected for time delay to obtain the time-delay-compensated obstacle distribution. Obtain the vehicle's dynamic parameters, including the vehicle's maximum steering angle rate constraint and maximum acceleration constraint; Based on the vehicle's position after time delay compensation as the starting point and the preset target position as the ending point, multiple initial paths are generated within the passable area; For each initial path, a dynamic feasibility verification is performed. This involves calculating the required steering angle rate and acceleration of the vehicle at each path point on the initial path, and determining whether the steering angle rate and acceleration meet the maximum steering angle rate constraint and the maximum acceleration constraint, respectively. Collision detection is performed on the initial path that satisfies the vehicle dynamics constraints. By determining whether the initial path has spatial intersection with the time-delay-compensated obstacle distribution, the initial path that has spatial intersection with the time-delay-compensated obstacle distribution is eliminated. The initial path that passes the dynamic feasibility verification and the collision detection will be used as a candidate path in the candidate path set.

6. The method according to claim 5, characterized in that, The target path selected from the candidate path set that satisfies both security and efficiency constraints includes: For each candidate path in the candidate path set, calculate the minimum safe distance between the candidate path and the time-delay-compensated obstacle distribution, and calculate a safety assessment value based on the minimum safe distance; Obtain the path length and path curvature change rate of the candidate path, and calculate the efficiency evaluation value based on the path length and path curvature change rate; The security assessment value and the efficiency assessment value are weighted and combined to obtain a comprehensive assessment value; the candidate path with the best comprehensive assessment value is selected from the candidate path set as the target path.

7. The method according to claim 1, characterized in that, Generating vehicle control commands based on the target path includes: The target path is discretized into multiple path segments, and the expected arrival time and expected travel speed are calculated for each path segment. Based on the spatial relationship between the vehicle's current position and the next path segment on the target path, the required steering angle command for the vehicle is generated; based on the vehicle's current speed and the expected speed of the next path segment, the required acceleration command for the vehicle is calculated. The steering angle command and the acceleration command are combined to generate the vehicle control command; The vehicle control command is sent to the vehicle actuator to control the vehicle to travel along the target path.

8. A path planning system based on vehicle controller cooperative perception, used to implement the method as described in any one of claims 1-7, characterized in that, include: The first unit is configured to acquire first perception data collected by the vehicle controller, the first perception data including environmental information acquired by the vehicle's own sensors; and receive second perception data from at least one external controller that has established a communication connection with the vehicle, the second perception data including environmental information acquired by the corresponding sensors of the external controller. The second unit is used to perform spatiotemporal alignment processing on the first and second sensing data. By mapping data points with overlapping sensing areas in the first and second sensing data to a unified spatiotemporal coordinate system, fused sensing data is generated. Based on the fused sensing data, an environmental representation model is constructed, which represents the distribution of obstacles and passable areas in the environment surrounding the vehicle. The third unit is used to generate a set of candidate paths based on the obstacle distribution and passable area in the environmental characterization model, combined with vehicle dynamics constraints and inter-controller communication delay compensation mechanism, and select a target path from the set of candidate paths that meets safety and efficiency constraints. Based on the target path, a vehicle control command is generated and sent to the vehicle actuator to control the vehicle to travel along the target path.

9. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 7.

Images