Multi-AGV path planning method and system based on improved CBS algorithm

Abstract

The invention proposes a multi-AGV path planning method and system based on an improved CBS algorithm, and the method comprises the steps: firstly calculating a shortest path set under the condition of no constraint, recognizing a path conflict through a parallel collision detection strategy, and enabling an AGV to carry out the planning of movement according to the shortest path if there is no conflict. And if the conflict exists, generating a global conflict set, and optimizing a conflict path through a weighted priority queue. And carrying out path search by utilizing a pruning strategy, and expanding nodes of which the total cost is lower than a cost threshold value. And if the new path after node expansion has no conflict, updating the new path as a potential optimal path. And re-executing path conflict detection between the AGVs until the shortest paths of the AGVs have no conflict, and obtaining a path planning result. By introducing a weighted priority queue mechanism, paths with serious conflicts or relatively high cost are processed preferentially, so that the calculation efficiency is improved. Meanwhile, a parallel collision detection strategy is adopted, so that the path planning efficiency is improved, and the waiting time is shortened. And in combination with a pruning strategy, redundant nodes in path planning can be avoided, and the search efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the fields of multi-agent systems and path planning, and more particularly to a multi-AGV path planning method and system based on an improved CBS algorithm. Background Art

[0002] Multi-Agent Path Finding (MAPF) is a core issue in the fields of artificial intelligence and robotics, and has important applications especially in industries such as intelligent manufacturing and warehousing logistics. A typical scenario of the MAPF problem is a multi-AGV path planning system, and its core goal is to plan a set of coordinated and conflict-free paths for multiple Automated Guided Vehicles (AGVs).

[0003] However, with the increase in the number of AGVs and the complexity of tasks, traditional path planning algorithms face challenges of excessive computing time and low conflict resolution efficiency when dealing with large-scale multi-AGV tasks. Especially in complex environments and high-density task situations, the problem becomes more prominent. Therefore, how to improve the computing efficiency of path planning while ensuring the path quality has become a difficult problem to be solved in the field of multi-AGV path planning. As a classic conflict-based path planning algorithm, the Conflict-Based Search (CBS) algorithm has achieved remarkable results in multi-robot path planning. However, when facing multi-agent path planning problems with a larger number and more complex environments, the CBS algorithm will encounter problems such as insufficient memory and path solving timeout. Summary of the Invention

[0004] To solve the problems of long computing time, low conflict detection efficiency, and poor path planning quality existing in the existing multi-AGV path planning algorithms when dealing with complex environments and high-density tasks, the present invention proposes a multi-AGV path planning method and system based on an improved CBS algorithm.

[0005] To achieve the above technical effects, the technical solution of the present invention is as follows:

[0006] A multi-AGV path planning method based on an improved CBS algorithm, comprising the following steps:

[0007] Based on the shortest path sets of each AGV under unconstrained conditions, detect whether there are conflicts in the paths between each AGV based on a parallel collision detection strategy;

[0008] If there is no conflict, each AGV moves according to its shortest path plan respectively;

[0009] If there is a conflict, generate a global conflict set, and optimize the conflict paths in the order of decreasing priority based on a weighted priority queue mechanism; where:

[0010] All possible path pairs are obtained according to the set of shortest paths, local conflict information is returned according to the conflicting path pairs, and all local conflict information is merged to obtain a global conflict set;

[0011] Path search is performed based on the global conflict set and pruning strategy. If the total cost of the expanded node is less than or equal to the preset cost threshold, the node is expanded; otherwise, the node is not expanded. When there is no conflict in the path from the starting point to the node after expanding the node, it is used as a potential optimal path and the set of shortest paths is updated;

[0012] The path conflict detection between AGVs is re-executed until there is no conflict in the shortest path of each AGV, and the path planning result is obtained.

[0013] The present invention also provides a multi-AGV path planning system based on an improved CBS algorithm, including:

[0014] A conflict detection module, configured to detect whether there is a conflict in the paths between AGVs based on a parallel collision detection strategy according to the set of shortest paths of each AGV without constraints;

[0015] A conflict-free path processing module, configured to output the current set of shortest paths as the path planning result when the conflict detection module outputs a detection result of no conflict;

[0016] A conflict path processing module, configured to generate a global conflict set when the conflict detection module outputs a conflict, and optimize the conflict paths in the order of decreasing priority based on a weighted priority queue mechanism. Among them, path pairs are obtained according to the set of shortest paths, local conflict information is returned according to the conflicting path pairs, all local conflict information is merged to obtain a global conflict set, and path search is performed based on a pruning strategy. If the total cost of the expanded node is less than or equal to the preset cost threshold, the node is expanded; otherwise, the node is not expanded. If there is no conflict in the path from the starting point to the node after expanding the node, a potential optimal path is obtained and the set of shortest paths is updated.

[0017] Compared with the prior art, the beneficial effects of the technical solution of the present invention are:

[0018] The present invention proposes a multi-AGV path planning method and system based on an improved CBS algorithm. By adopting a parallel collision detection strategy, the path conflicts of multiple AGVs can be processed synchronously, reducing time delay, improving path planning efficiency, reducing waiting time, and optimizing resource allocation. By introducing a weighted priority queue, paths with severe conflicts or higher costs are preferentially processed, thereby improving the calculation efficiency, enabling more accurate sorting and processing of conflicting paths, and avoiding inefficiency and resource waste during the path processing. By adopting a pruning strategy for path search to real-time eliminate the invalid search space, it helps to avoid redundant nodes during path planning, reduce the number of search nodes and search time, improve the search efficiency, and ensure the efficient utilization of computing resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 FIG. is the overall flowchart of the multi-AGV path planning method based on the improved CBS algorithm shown in the embodiments of the present invention.

[0020] Figure 2 FIG. is the flowchart of the multi-AGV path planning method based on the improved CBS algorithm shown in the embodiments of the present invention.

[0021] Figure 3 FIG. is the architecture diagram of the multi-AGV path planning system based on the improved CBS algorithm shown in the embodiments of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0022] Here, the exemplary embodiments will be described in detail, and the examples are shown in the drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present invention as detailed in the appended claims.

[0023] The terms used in the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The singular forms "a", "the", and "said" used in the present invention and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0024] It should be understood that although the terms first, second, third, etc. may be used in the present invention to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from each other. For example, without departing from the scope of the present invention, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "when" or "while" or "in response to determining".

[0025] The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0026] Embodiment 1

[0027] This embodiment proposes a multi-AGV path planning method based on an improved CBS algorithm, as Figure 1 shown, which is the overall flowchart of the multi-AGV path planning method based on the improved CBS algorithm in this embodiment.

[0028] In the multi-AGV path planning method based on the improved CBS algorithm proposed in this embodiment, the following steps are included; the flowchart is as Figure 2 shown:

[0029] Based on the shortest path sets of each AGV under unconstrained conditions, detect whether there are conflicts in the paths between AGVs based on a parallel collision detection strategy;

[0030] If there is no conflict, each AGV moves according to its shortest path plan respectively;

[0031] If there is a conflict, generate a global conflict set, and optimize the conflicting paths in the order of decreasing priority based on a weighted priority queue mechanism; where:

[0032] Obtain all possible path pairs according to the shortest path sets, return local conflict information according to the conflicting path pairs, and merge all local conflict information to obtain a global conflict set;

[0033] Perform path search based on the global conflict set and pruning strategy. If the total cost of the expanded node is less than or equal to the preset cost threshold, expand the node, otherwise do not expand the node; when there is no conflict in the path from the starting point to the node after expanding the node, use it as a potential optimal path and update the shortest path set;

[0034] Re-execute the path conflict detection between AGVs until there is no conflict in the shortest paths of each AGV to obtain the path planning result.

[0035] In this embodiment, based on the shortest path sets of each AGV under unconstrained conditions, a parallel collision detection strategy is used to detect whether there are conflicts in the paths between AGVs. By adopting the parallel collision detection strategy, the paths of multiple AGVs can be calculated simultaneously during the conflict detection stage, thus greatly improving the calculation efficiency. Through parallel computing, the path conflicts between AGVs can be processed synchronously, reducing the time delay caused by serial computing. If there are no conflicts, each AGV moves according to its shortest path plan; if there are conflicts, a global conflict set is generated, and based on the weighted priority queue mechanism, the conflict paths are optimized in the order of decreasing priority. By introducing the weighted priority queue mechanism, the priority of conflict handling is dynamically adjusted to effectively reduce the calculation time of path planning. Specifically, the algorithm assigns different weights to paths according to the cost of each path, the number of conflicts, and the number of times the expanded nodes are generated, so as to prioritize the processing of paths with serious conflicts or high calculation costs. Through this method, the algorithm can more effectively concentrate resources to process critical paths, avoiding the inefficient way of treating all paths with the same priority in traditional algorithms; among them: all possible path pairs are obtained according to the shortest path set, local conflict information is returned according to the conflicting path pairs, and all local conflict information is merged to obtain a global conflict set; based on the global conflict set and pruning strategy, path search is performed to further optimize the performance of the algorithm by eliminating invalid search spaces. During the path planning process, the algorithm can identify and exclude search spaces that cannot generate valid paths in real time, avoiding meaningless calculations and ensuring the efficient use of computing resources. If the total cost of the expanded node is less than or equal to the preset cost threshold, the node is expanded; otherwise, the node is not expanded; when there is no conflict in the path from the starting point to the node after expanding the node, it is used as a potential optimal path and the shortest path set is updated; the path conflict detection between AGVs is re-executed until there are no conflicts in the shortest paths of each AGV, and the path planning result is obtained, significantly improving the efficiency of multi-AGV path planning and the conflict resolution ability.

[0036] In an alternative embodiment, the weighted priority queue mechanism dynamically adjusts the priority of conflict handling according to the cost of the path, the number of conflicts, and the number of times the expanded nodes are generated.

[0037] In an alternative embodiment, the weighted priority queue mechanism includes the following steps:

[0038] Calculate the weights of each path, and sort each path in the weighted priority queue from smallest to largest weight, and process the paths in the order of decreasing priority;

[0039] After conflict resolution, update the cost, conflict information, and the number of times the expanded nodes are generated for the path, and recalculate the weight of the path;

[0040] Insert the updated path into the conflict - free path set. Adjust the priorities of the paths in the weighted priority queue according to the recalculated weights until the paths in the weighted priority queue are empty, and finally output the conflict - free shortest path set.

[0041] In an alternative embodiment, when calculating the weights of each path, the following steps are included: Calculate the weight of the path according to the cost of the path, the number of conflicts, and the generation times of the expanded nodes. The expression is:

[0042] ω i =α·C i +β·D i +γ·G i

[0043] Among them, ω i represents the weight of path P i ; C i represents the number of conflicts on path P i ; D i represents the cost of path P i ; G i represents the generation times of the expanded nodes involved in path P i , and α, β, γ represent weight coefficients.

[0044] Exemplarily, first, extract path P i one by one from the path set P, and calculate the parameters related to this path: the number of conflicts C i , the path cost D i , and the generation times of the expanded nodes G i . The number of conflicts C i is used to record the number of times that the path may conflict with other paths. The path cost D i is used to evaluate the overall cost of the path, such as the path length or time consumption. The generation times of the expanded nodes G i is used to count the generation times of the nodes involved in the path during the search process. According to the extracted parameters C i , D i , G i , calculate the path weight through the formula ω i =α·C i +β·D i +γ·G i . Here, the weight coefficients α, β, γ respectively control the influence ratios of the number of conflicts, the path cost, and the generation times of the expanded nodes on the weight, so as to dynamically reflect the importance of the path. Attach the calculated weight value ω i to path P i to form path data containing priority information.

[0045] Secondly, insert all the paths with calculated weights into the weighted priority queue WPQ one by one. Each path is initially sorted according to its weight value, and the paths with smaller weights are processed first. Perform a global sorting on all the paths in the weighted priority queue WPQ based on the weight values. Ensure that the paths with serious conflicts or higher costs are at the front of the weighted priority queue and are given priority for resolution. The sorted weighted priority queue WPQ indicates the priority distribution of the most critical paths in the current path planning task, providing an efficient processing basis for subsequent conflict resolution steps.

[0046] Finally, extract the path P with the highest priority, i.e., the smallest weight, from the weighted priority queue WPQ. i Perform conflict detection on the extracted path, mark the conflict points, and resolve the path conflicts through constraints or adjustments. For the path P after conflict resolution, i recalculate its conflict count C, i path cost D, i number of extended node generations G. i The updated weight ω i reflects the change in the importance of the current path. Insert the updated path into the conflict-free path set, and re-sort the paths in the weighted priority queue to adjust the priorities. This process is continuously iterated until the weighted priority queue is empty, and finally, the conflict-free path set Ω is output. This dynamic adjustment mechanism ensures that the critical paths are resolved first, optimizing the overall path planning efficiency.

[0047] In this embodiment, introducing the weighted priority queue mechanism can sort the paths and assign dynamic priorities to the paths, thus effectively handling the conflict paths, and avoiding inefficiencies and resource waste in the path management process. The dynamic adjustment mechanism ensures that the algorithm can concentrate resources to prioritize the key issues, thereby effectively improving the computing efficiency. In contrast, traditional methods often rely on basic queues or lists to manage paths, and their priorities are only based on the order in which the paths are added to the queue, lacking targeted priority sorting. This results in the path processing order being largely random and unable to be dynamically adjusted according to the severity of the conflicts, thereby significantly reducing the computing efficiency when dealing with large-scale problems. Through the weighted priority queue, we can ensure that the paths are processed in a more reasonable order, optimizing the overall path planning process.

[0048] In an alternative embodiment, the parallel collision detection strategy includes the following steps:

[0049] Generate all possible path pairs according to the shortest path set, divide the path pairs into multiple groups of tasks and distribute them to multiple computing units in parallel, and each computing unit judges the conflict detection of a group of path pairs;

[0050] When there is a conflict in the path pair, record the information of the path pair into the local conflict set;

[0051] After all computing units complete their tasks, the returned local conflict information is aggregated to form a global conflict set; the global conflict set includes detailed information on conflicts that occur for all path pairs at each time step.

[0052] Exemplarily, first, all possible path pairs (P i , P j ) are generated from the path set P. The total number N of path pairs is calculated by the formula N = n·(n - 1) / 2, where n is the number of AGVs. Each path pair formed by combining each path with other paths is the basic unit for subsequent conflict detection. The generated path pairs are divided into m groups according to the uniform principle, and the size of each group is approximately N / m. By such a task division method, it is ensured that the task loads of the computing units are uniform, and individual units are prevented from becoming performance bottlenecks due to excessive task amounts. The path pair tasks divided for each group are assigned to an independent computing unit. Each computing unit will independently be responsible for conflict detection of the path pairs within its assigned group, preparing for subsequent parallel processing.

[0053] Secondly, each computing unit starts to independently process the path pair tasks assigned to it. For each pair of paths (P i , P j ), at each time step t, it is checked whether the spatial positions of the two paths overlap at time t. If it is detected that at a certain time step the path pair (P i (t) = P j (t)), it is determined as a conflict, and the conflict path pair and related information are recorded in the local conflict set. This information includes the identifier of the conflict path pair and the time step at which the conflict occurs. After each computing unit completes the conflict detection task for the path pairs assigned to it, it returns the local conflict set containing the conflict information, providing input for subsequent global aggregation.

[0054] Finally, the local conflict sets returned by all computing units are collected and aggregated to form a global conflict set C. The global set contains complete information on conflicts that occur for all path pairs at all time steps. The global conflict set C is used to update the conflict status of the paths, providing comprehensive and accurate data support for the path adjustment and planning in the next stage. Through the parallelization strategy, the time complexity of collision detection is reduced from the traditional O(n 2 ·T) to O((n 2 ·T) / m). This method significantly improves the computational efficiency of multi-AGV path planning and meets the requirements of large-scale tasks.

[0055] In this embodiment, the parallel collision detection strategy generates all possible path pairs and evenly distributes them to multiple computing units. Each unit independently detects potential conflicts in the path pairs. This method not only ensures the balanced distribution of the task load, avoiding overloading a single computing unit, but also significantly improves the detection efficiency through parallel operations, significantly enhancing the computing efficiency while reducing resource waste.

[0056] In an alternative embodiment, the pruning strategy includes the following steps:

[0057] Expand nodes during path search; where:

[0058] Calculate the total cost of each expanded node. If the total cost of the expanded node exceeds the preset cost threshold, then do not adopt this expanded node;

[0059] During the process of expanding nodes, for the child nodes of each current node, if they already exist in the closed list, directly skip this node; the closed list includes all the nodes that have been processed;

[0060] Each time a node is expanded, if the path from the starting point to the current node has no conflict, stop expanding and return the current path as the potential optimal path.

[0061] In an alternative embodiment, the calculation of the total cost of each expanded node further includes the following steps:

[0062] For nodes whose total cost exceeds the preset cost threshold, skip the expansion operation and do not add the nodes to the open list; the open list includes all the nodes that have been discovered but not processed;

[0063] When skipping nodes whose cost exceeds the preset cost threshold, concentrate the computing resources on nodes with lower total costs.

[0064] Exemplarily, first, each time a node is expanded, calculate the total cost of the expanded node v and compare it with the preset cost threshold θ. If the total cost of the expanded node exceeds the cost threshold, then determine that this node is not expandable. For the expanded node v whose total cost exceeds the cost threshold, directly skip its expansion operation and do not add it to the open list. For the expanded node v whose total cost is less than or equal to the cost threshold, add it to the open list. In this embodiment, by removing the nodes whose total path cost exceeds the preset cost threshold, the upper limit of the path search cost is restricted, and the pruning operation is realized, which can effectively reduce the expansion of invalid nodes and save computing resources. While skipping the nodes with excessive cost, concentrate the computing resources on nodes with lower total costs. This resource optimization strategy helps to quickly find a better feasible path.

[0065] Secondly, when expanding a node each time, check whether there are conflicts in the path from the starting node to the current node v. If the path has no conflicts at all, mark it as a potentially optimal path. For a conflict-free path, directly return this path as the current optimal solution without continuing to expand other nodes. Terminate the expansion operation in advance to avoid redundant calculations and quickly output the result. Add the conflict-free path to the final path set, providing a reference for the subsequent path planning with known optimal solutions, and at the same time reducing the scale of the search space.

[0066] Finally, for each child node u of the current node, check whether it already exists in the closed list. If it already exists, directly skip it and do not repeat the expansion of this node. In the path search tree, if a certain branch cannot find any valid path under the current constraints, throw an exception or directly terminate the exploration of this branch. Combining the total cost check and the node status check, remove all nodes that cannot generate valid paths from the open list, further optimizing the search efficiency and ensuring that computing resources focus on the parts that are likely to generate optimal paths.

[0067] In this embodiment, the pruning mechanism further reduces the meaningless search space and redundant calculations by filtering nodes with excessive total costs, terminating conflict-free paths in advance, and removing invalid nodes.

[0068] Embodiment 2

[0069] This embodiment proposes a multi-AGV path planning system based on the improved CBS algorithm, applying the multi-AGV path planning method based on the improved CBS algorithm proposed in Embodiment 1. As Figure 3 shown, it is the architecture diagram of the multi-AGV path planning system based on the improved CBS algorithm in this embodiment.

[0070] In the multi-AGV path planning system based on the improved CBS algorithm proposed in this embodiment, it includes:

[0071] A conflict detection module, which is used to detect whether there are conflicts in the paths of each AGV based on the parallel collision detection strategy according to the set of shortest paths of each AGV without constraints;

[0072] A conflict-free path processing module, which is used to output the current set of shortest paths as the path planning result when the conflict detection module outputs a detection result of no conflict;

[0073] The conflict path processing module is used to generate a global conflict set when the conflict detection module outputs that there is a conflict, and optimize the conflict paths in the order of decreasing priority based on the weighted priority queue mechanism; among them, path pairs are obtained from the shortest path set, local conflict information is returned according to the conflicting path pairs, the local conflict information is merged to obtain a global conflict set, and path search is performed based on a pruning strategy. If the total cost of the expanded node is less than or equal to the preset cost threshold, the node is expanded; otherwise, the node is not expanded. If there is no conflict in the path from the starting point to the node after expanding the node, a potential optimal path is obtained and the shortest path set is updated.

[0074] Embodiment 3

[0075] This embodiment provides a computer device, including a memory and a processor. A computer-readable instruction is stored in the memory. When the computer-readable instruction is executed by the processor, the processor executes all or part of the steps of a multi-AGV path planning method based on an improved CBS algorithm proposed in Embodiment 1.

[0076] Embodiment 4

[0077] This embodiment provides a storage medium, on which a computer-readable instruction is stored. When the computer-readable instruction is executed by a processor, all or part of the steps of a multi-AGV path planning method based on an improved CBS algorithm proposed in Embodiment 1 are implemented.

[0078] Exemplarily, the storage medium includes, but is not limited to, various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs.

[0079] Exemplarily, the instructions, programs, code sets, or instruction sets can be implemented using conventional programming languages.

[0080] Exemplarily, the processor includes, but is not limited to, smartphones, personal computers, servers, network devices, etc., and is used to execute all or part of the steps of a multi-AGV path planning method based on the improved CBS algorithm described in Embodiment 1.

[0081] Each embodiment in the present invention is described in a progressive manner. For the same or similar parts among the embodiments, reference can be made to each other. Each embodiment focuses on the differences from other embodiments. In particular, for the device embodiments, since they are basically similar to the method embodiments, they are described relatively simply. For the relevant parts, reference can be made to the description of the method embodiments. The device embodiments described above are merely exemplary. The modules described as separate components may or may not be physically separated. When implementing the solution of the present invention, the functions of the modules can be realized in one or more software and / or hardware. It is also possible to select some or all of the modules according to actual needs to achieve the purpose of the solution of this embodiment.

[0082] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, rather than limiting the implementation manners of the present invention. For those of ordinary skill in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to enumerate all the implementation manners here. Any modifications, equivalent replacements, and improvements made within the spirit and principle of the present invention shall be included in the protection scope of the claims of the present invention.

Claims

1. A multi-AGV path planning method based on an improved CBS algorithm, characterized in that: The following steps are involved: According to the unconstrained shortest path set of each AGV, a parallel collision detection strategy is used to detect whether there is a conflict between the paths of each AGV; If there is no conflict, each AGV moves according to its shortest path planning; If there is a conflict, a global conflict set is generated, and the conflicting paths are optimized in descending order of priority based on the weighted priority queue mechanism; where: Obtain all possible path pairs according to the shortest path set, return local conflict information according to the conflicting path pairs, and obtain a global conflict set after merging all local conflict information; Perform path search based on the global conflict set and pruning strategy. If the total cost of the expanded node is less than or equal to the preset cost threshold, the node is expanded, otherwise the node is not expanded. When there is no conflict in the path from the starting point to the node after the node is expanded, it is used as a potential optimal path and the shortest path set is updated. Re-execute the path conflict detection between AGVs until the shortest paths of each AGV are conflict-free, and obtain the path planning result.

2. According to claim 1, a multi-AGV path planning method based on an improved CBS algorithm is characterized in that: The weighted priority queue mechanism dynamically adjusts the priority of conflict handling according to the cost of the path, the number of conflicts, and the number of times the extended node is generated.

3. The multi-AGV path planning method based on the improved CBS algorithm according to claim 2 is characterized in that: The weighted priority queue mechanism comprises the following steps: Calculate the weight of each path, sort the paths in a weighted priority queue from small to large weights, and process the paths in descending order of priority; After the conflict is resolved, the path cost, conflict information, and number of extended node generation are updated, and the path weight is recalculated; The updated paths are inserted into the conflict-free path set, and the paths in the weighted priority queue are prioritized according to the recalculated weights until the paths in the weighted priority queue are empty, and finally the conflict-free shortest path set is output.

4. The multi-AGV path planning method based on the improved CBS algorithm according to claim 3 is characterized in that: The calculation of the weight of each path includes the following steps: calculating the weight of the path according to the cost of the path, the number of conflicts, and the number of times the extended node is generated, and the expression is: oh i =α·C i +β·D i +γ·G i Among them, ω i Represents the path P i The weight of C i Represents the path P i The number of conflicts on i Represents the path P i The cost, G i Represents the path P i The number of times the extended nodes involved are generated, and α, β, and γ represent weight coefficients.

5. The multi-AGV path planning method based on the improved CBS algorithm according to claim 1 is characterized in that: The parallel collision detection strategy includes the following steps: Generate all possible path pairs according to the shortest path set, divide the path pairs into multiple groups of tasks and distribute them to multiple computing units in parallel, each computing unit determines the conflict detection of a group of path pairs; When there is a conflict in a path pair, the information of the path pair is recorded in the local conflict set; When all computing units complete their tasks, the returned local conflict information is summarized to form a global conflict set; the global conflict set includes detailed information on conflicts occurring in all path pairs at each time step.

6. The multi-AGV path planning method based on the improved CBS algorithm according to claim 1 is characterized in that: The pruning strategy includes the following steps: Expands nodes during path search; where: Calculate the total cost of each expansion node. If the total cost of the expansion node exceeds the preset cost threshold, the expansion node will not be used. During the node expansion process, if each child node of the current node already exists in the closed list, the node is directly skipped; the closed list includes all nodes that have been processed; Each time a node is expanded, if there is no conflict in the path from the starting point to the current node, the expansion stops and the current path is returned as the potential optimal path.

7. The multi-AGV path planning method based on the improved CBS algorithm according to claim 6 is characterized in that: The calculation of the total cost of each extended node also includes the following steps: For nodes whose total cost exceeds a preset cost threshold, the expansion operation is skipped and the node is not added to the open list; the open list includes all nodes that have been discovered but not processed; When nodes whose cost exceeds the preset cost threshold are skipped, computing resources are concentratedly allocated to nodes with lower total cost.

8. A multi-AGV path planning system based on an improved CBS algorithm, applying the multi-AGV path planning method based on an improved CBS algorithm as claimed in any one of claims 1 to 7, characterized in that: include: The conflict detection module is used to detect whether there is a conflict between the paths of each AGV based on the shortest path set of each AGV under unconstrained conditions and the parallel collision detection strategy; A conflict-free path processing module is used to output the current shortest path set as a path planning result when the conflict detection module outputs a detection result that there is no conflict; The conflict path processing module is used to generate a global conflict set when the conflict detection module outputs a conflict, and optimize the conflict paths in descending order of priority based on the weighted priority queue mechanism; wherein, a path pair is obtained according to the shortest path set, local conflict information is returned according to the conflicting path pair, the global conflict set is obtained after the local conflict information is merged, and a path search is performed based on the pruning strategy; if the total cost of the extended node is less than or equal to the preset cost threshold, the node is extended, otherwise the node is not extended; If there is no conflict in the path from the starting point to the node after the node is expanded, a potential optimal path is obtained and the shortest path set is updated.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the multi-AGV path planning method based on the improved CBS algorithm as described in any one of claims 1 to 7 is implemented.

10. A computer, characterized in that: Comprising a computer medium according to claim 9.

Images