CN-121999625-A - Multi-vehicle cooperative path planning method and system for autonomous passenger parking system

CN121999625ACN 121999625 ACN121999625 ACN 121999625ACN-121999625-A

Abstract

The disclosure provides a multi-vehicle collaborative path planning method and system for an autonomous valet parking system. The method comprises the steps of defining a multi-vehicle track planning model in a parking lot environment, converting the multi-vehicle track planning model into a grid form to generate an AVP scene map, acquiring parking information of each vehicle to be parked based on the AVP scene map, directly planning and generating complete single vehicle paths from parking start coordinates to parking end coordinates of each vehicle to be parked according to the parking information, detecting potential conflicts among the complete single vehicle paths of all the vehicles to be parked by adopting a conflict searching algorithm driven by a priority mechanism, and carrying out conflict elimination to obtain a collision-free multi-vehicle parking path. The method has the advantages that a path scheme with higher quality can be generated in the structured parking lot environment, the overall running time is remarkably reduced by about 26%, and the running efficiency of the vehicle is improved.

Inventors

WANG GUANGWEI
Jin Bieshu
XIAO ZHONGKUN
ZHAO JIN
SHI QING
XU QING
WANG JIANQIANG

Assignees

贵州大学

Dates

Publication Date: 20260508
Application Date: 20260130

Claims (10)

1. A multi-vehicle cooperative path planning method of an autonomous valet parking system is characterized by comprising the following steps: defining a multi-vehicle track planning model in a parking lot environment, converting the multi-vehicle track planning model into a grid form, and generating an AVP scene map; obtaining parking information of each vehicle to be parked based on the AVP scene map, and directly planning and generating a complete single vehicle path from a parking start point coordinate to a parking end point coordinate of each vehicle to be parked according to the parking information, wherein the parking information comprises a parking mode, a parking end point coordinate, a parking start point coordinate and a vehicle posture to be parked; And detecting potential conflicts among the complete single vehicle paths of all the vehicles to be parked by adopting a conflict searching algorithm driven by a priority mechanism, and carrying out conflict elimination to obtain a collision-free multi-vehicle parking path.
2. The multi-vehicle cooperative path planning method of an autonomous valet parking system according to claim 1, wherein the priority mechanism performs priority allocation based on vehicle IDs, which are numbered sequentially based on the order in which the respective vehicles arrive; the smaller the vehicle ID, the higher the assigned priority, and the larger the vehicle ID, the lower the assigned priority.
3. The method for planning a multi-vehicle cooperative path of an autonomous passenger parking system according to claim 1, wherein a collision search algorithm driven by a priority mechanism is used to detect potential collisions between the complete single vehicle paths of all vehicles to be parked and to perform collision elimination, to obtain a collision-free multi-vehicle parking path, comprising: Searching and obtaining a root node in SearchQueue, and if SearchQueue is not empty, detecting vehicle collision of all complete bicycle paths; If the collision detection result is that no potential collision exists, the path set of all vehicles is a solution of the problem; and regenerating a complete single vehicle path for the low-priority vehicle to be parked based on the priority list set by the priority mechanism to obtain the collision-free multi-vehicle parking path.
4. A multi-vehicle collaborative path planning method for an autonomous guest parking system according to claim 3 further comprising the steps of: Acquiring scene quantization indexes of a parking lot; Comparing the scene quantification index with a preset level threshold, judging whether the scene quantification index meets the preset level threshold and the duration reaches the preset time, if so, dynamically adjusting a priority list set based on a priority mechanism to obtain a dynamic priority list, The scene quantization indexes comprise capacity saturation, key channel occupancy rate and traffic efficiency deviation.
5. The method for planning a multi-vehicle cooperative path of an autonomous guest parking system according to claim 4, wherein the dynamic adjustment of the priority list set based on the priority mechanism comprises the steps of: Comparing the real-time state data of all vehicles to be parked based on a preset tuning screening dimension, and screening out vehicles to be parked meeting any tuning screening dimension to form a vehicle set to be tuned; sorting the vehicle set to be adjusted based on the preset dimension priority to obtain a front adjustment vehicle list; and determining the insertion positions of the priorities of different dimensions in the priority list, and inserting each vehicle to be adjusted corresponding to the priorities of different dimensions into the corresponding positions according to the sequence in the vehicle list before adjustment to obtain a dynamic priority queue.
6. The method for planning a multi-vehicle collaborative path for an autonomous passenger parking system according to claim 5, wherein a complete single-vehicle path is regenerated for low priority vehicles to be parked based on a dynamic priority list to obtain a collision-free multi-vehicle parking path.
7. The method for multi-vehicle collaborative path planning for an autonomous valet parking system according to claim 1, wherein the process of detecting potential conflicts comprises: Traversing all the complete bicycle paths, selecting a time step corresponding to the complete bicycle path with the longest path as a reference time, extending the time steps corresponding to all the vehicles to be parked to the reference time, and expanding the state of the vehicles to be parked corresponding to the complete bicycle path with the time steps smaller than the reference time at the parking end point coordinates to all the subsequent time steps; extracting state information of all vehicles to be parked at the moment on each time step, and constructing an expansion bounding box with a safety buffer zone by utilizing a vehicle geometric transformation function and the state information, wherein the width of the expansion bounding box above two sides of the vehicles to be parked is in direct proportion to the absolute value of a steering angle; and collision detection is carried out on expansion bounding boxes of any two vehicles to be parked through a separation theorem, whether the expansion bounding boxes of the two vehicles to be parked overlap at the same time step is judged, if yes, potential collision is judged, collision information is recorded and returned, and the collision information comprises vehicle IDs, collision occurrence time and collision occurrence positions related to the collision.
8. A multi-vehicle collaborative path planning system for an autonomous valet parking system for implementing the method of any of claims 1-7, comprising the following modules: The model construction module is used for defining a multi-vehicle track planning model in a parking lot environment, converting the multi-vehicle track planning model into a grid form and generating an AVP scene map; The single-vehicle path generation module is connected with the model construction module and is used for acquiring parking information of each vehicle to be parked based on the AVP scene map and generating a complete single-vehicle path of each vehicle to be parked according to the parking information, wherein the parking information comprises a parking mode, a parking terminal point coordinate, a parking starting point coordinate and a vehicle posture to be parked; The multi-vehicle planning module is connected with the single-vehicle path generating module and is used for detecting potential conflicts among the complete single-vehicle paths of all vehicles to be parked by adopting a conflict searching algorithm driven by a priority mechanism and carrying out conflict elimination to obtain a collision-free multi-vehicle parking path.
9. An electronic device comprising a memory, a processor, and a computer program, wherein the computer program is stored in the memory and configured to be executed by the processor to implement the multi-vehicle collaborative path planning method of an autonomous valet parking system of any of claims 1-7.
10. A computer readable storage medium having stored thereon a computer program, the computer program being executable by a processor to implement the multi-vehicle collaborative path planning method of an autonomous passenger parking system of any of claims 1-7.

Description

Multi-vehicle cooperative path planning method and system for autonomous passenger parking system Technical Field The disclosure relates to the technical field of automatic driving, in particular to a multi-vehicle collaborative path planning method and system of an autonomous passenger parking system. Background With the acceleration of the urban process and the vigorous development of intelligent traffic technology, an autonomous passenger parking (Automated VALET PARKING, AVP) system is taken as a leading edge technology in the intelligent traffic field, and gradually becomes a revolutionary scheme for solving the urban parking problem. The AVP system can realize the full-automatic parking process from the appointed get-off point to the parking space, not only remarkably improves the parking efficiency and the space utilization rate, but also provides more convenient and comfortable parking experience for users. In contrast, the traditional parking lot management system only provides the information of the number of the remaining parking spaces, and lacks intelligent scheduling capability, so that the problems of low utilization rate of the parking spaces, overlong locating time and the like are caused. However, as AVP systems expand from single car applications to multiple car scenarios, multi-car collaborative path planning has become a core technical challenge that restricts its large-scale deployment. In a complex scene that multiple vehicles run simultaneously, because the planning process is limited to a single vehicle body, the real-time sensing and coordination of the motion states of other vehicles are lacking, and the phenomena of parking space control, mutual blocking, even local deadlock and the like are easily caused. From a theoretical point of view, this problem essentially falls into the category of a Multi-vehicle trajectory planning model (Multi-AGENT PATH FINDING, MAPF), a proven NP-hard problem whose computational complexity grows exponentially with the number of agents. Therefore, how to design an efficient multi-vehicle collaborative path planning algorithm, and plan collision-free and efficient optimal paths for multiple autonomous vehicles simultaneously on the premise of ensuring safety, has become a key challenge for AVP system research. In view of the above challenges, existing studies are largely divided into three categories: The first type is a centralized planning method, wherein a vehicle dynamics optimization model is built, and a multi-vehicle cooperative parking track is solved uniformly; the second type is a distributed coordination strategy, and the deep reinforcement learning is utilized to realize coordination of parking space allocation and multi-agent negotiation; the third class is a priority-based hybrid approach, such as priority inheritance backtracking algorithms, perform well in terms of efficiency and convergence guarantees, and priority searching achieves significant computational efficiency improvement in intersection coordination. Although the existing method achieves a certain effect under a specific scene, the method has obvious limitations that firstly, the centralized method such as a scheme based on integer linear programming has high computational complexity and is difficult to apply in real time, and the improved genetic algorithm and the time enhancement A are combinedThe multi-objective scheduling strategy of the algorithm is easy to be trapped in local optimum, and the computational complexity is increased sharply in a high-density environment. Second, while the distributed architecture reduces the computational burden, the first-come-first-serve strategy may result in long-time queuing and deadlock at high-density parking. Most importantly, these methods are mostly validated in simple environments, and lack special treatment for complex environments with characteristics of narrow space, complex layout, high vehicle density and the like, such as parking lots. Meanwhile, the parking lot environment also needs to consider incomplete constraint and accurate geometric shape of vehicles, which puts higher demands on multi-vehicle collaborative path planning. Therefore, there is an urgent need to develop a multi-vehicle collaborative path planning method and system for an autonomous passenger parking system to solve the above problems. Disclosure of Invention The invention provides a multi-vehicle cooperative path planning method and system of an autonomous passenger parking system, which are used for solving the problem that an optimal multi-vehicle cooperative path cannot be rapidly and accurately given out in a complex parking area environment (such as narrow space, complex layout, high vehicle density and the like). According to a first aspect of the present disclosure, a multi-vehicle collaborative path planning method for an autonomous valet parking system is provided. The method comprises the following steps: defining a multi-vehicle track pla