Bounded suboptimal agent path planning method and system based on conflict search
By introducing a sorting strategy of cardinal conflict and internal constraints in multi-agent path planning, the conflict resolution process is optimized, and the problem of excessively long and low-efficiency solving of existing algorithms in the case of a large number of agents is solved, and more efficient Path planning.
Patent Information
- Application Number
- CN202411569919.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-06
- Publication Date
- 2024-12-03
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Existing multiagent path planning algorithms based on conflict searches have too long and inefficient solutions when processing a large number of agents, especially when redundant operations and backtracking between advanced searches and low-level searches lead to waste of computing resources.
Introduce a new sorting strategy based on cardinal conflict and internal constraints, optimize the conflict resolution process through FOCAL list, allowing the selection of conflict-free but high-cost nodes as target nodes within the suboptimal solution range, and preferentially extend the cardinal conflict and the number of internal constraints. nodes, reducing redundant backtracking and inefficient verification.
It significantly reduces the number of node expansions and the number of low-level path re-planning times for advanced searches, improves the overall computing efficiency, and can quickly converge to reasonably solve the multi-agent path planning problem.
Smart Images

Figure CN119063741A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of path planning, and in particular to a bounded suboptimal intelligent agent path planning method and system based on conflict search. Background Art
[0002] The task of multi-agent path planning problem is to plan conflict-free paths for multiple agents (such as robots or AGVs) so that they can act simultaneously in a shared environment without conflicts. The key challenge is that as the number of agents increases, the complexity of path planning increases dramatically. Achieving a reasonable trade-off between solution time and solution quality is the key issue faced.
[0003] The conflict-based-search algorithm (CBS) is a classic method for solving multi-agent path planning problems. It adopts a two-level strategy, divided into high-level search and low-level search. The high-level search of the CBS algorithm is responsible for detecting conflicts in the paths of multiple agents and generating new nodes by adding constraints; the low-level search replans the path for each agent while satisfying these constraints, usually using the A* algorithm. The high-level search is responsible for global optimization, and the low-level search is responsible for solving the path of a single agent. The two work together to find a conflict-free multi-agent path solution. CBS guarantees the optimality and completeness of the solution, that is, it can find a feasible solution, and when a solution exists, the algorithm will definitely find it.
[0004] In order to ensure the optimality of the solution, the algorithm faces several major problems: First, the MAPF problem is NP-hard. As a feasible solution, CBS can only handle a limited number of agents, and the solution time to return a feasible solution is often too long. Second, the advanced search verifies the optimality by expanding the node with the lowest cost, which results in repeated addition of the same constraints at different nodes, resulting in a lot of redundancy and backtracking, which prolongs the solution time. Third, CBS is inefficient in dealing with certain conflicts. When there is no solution to the current cost, CBS needs to verify that all paths are infeasible before continuing to expand the node. For conflicts with large cost increases, CBS needs to verify whether all solutions below the cost are infeasible, which consumes a lot of computing resources. Summary of the invention
[0005] In order to solve the above-mentioned problems, the present invention proposes a bounded suboptimal algorithm based on conflict search on the basis of retaining the integrity of CBS, aiming to improve the solution efficiency of the CBS algorithm by relaxing the cost restriction of the target node and introducing a new sorting strategy based on cardinality conflicts and internal constraints.
[0006] In a first aspect, the present invention provides a bounded suboptimal agent path planning method based on conflict search, which adopts the following technical solution: A bounded suboptimal agent path planning method based on conflict search, comprising: Obtain multi-agent road information; Initialize the root node and generate the shortest path based on the multi-agent road information; Check whether there is a conflict in the path. If there is no conflict, directly output the shortest path as the final solution; If there is a conflict, resolve the conflict by inserting the node and generate a new child node; Update the path of the generated new child node; Check new child nodes and repeat conflict detection and update until a feasible solution is found.
[0007] Furthermore, the initialization of the root node includes setting the constraint set Root.constraints of the root node to an empty set, which is used to indicate that the current node has no constraints, thereby ensuring that subsequent searches are performed under unconstrained conditions.
[0008] Furthermore, the generating of the shortest path based on multi-agent road information includes generating an initial path for each agent through a low-level search, including calling the low-level search A^* algorithm, calculating the shortest path from the starting point to the end point for k agents in turn, and storing the search results in Root.solution.
[0009] Furthermore, the detection of whether there is a conflict in the path includes determining whether two agents and Are they at the same time step? Occupies the same vertex or the same side , if the conflict occurs at a vertex, the conflict is represented as , if the conflict occurs on an edge, the conflict is represented as .
[0010] Furthermore, the conflict resolution and generation of new child nodes through node insertion include, when there is a conflict, inserting the Root node into the OPEN list, checking whether the node meets the suboptimal limit and inserting it into the FOCAL list, and selecting the best node in the FOCAL list for expansion, and generating new child nodes for the conflicting agents.
[0011] Furthermore, the path updating of the generated child nodes includes adding the constraints of the conflicting agent a_i to the constraint set A.constraints of the new node, and updating its solution A.solution, and generating a new path for the conflicting agent by calling the low-level search again.
[0012] Furthermore, the checking of new child nodes and repeated conflict detection and updating until a feasible solution is found includes checking the cost value of the newly generated child node. If the cost is less than infinity, the new child node is inserted into the OPEN list, and conflict detection and path updating are repeated until a conflict-free solution is found.
[0013] In the second aspect, a bounded suboptimal agent path planning system based on conflict search includes: A data acquisition module is configured to acquire multi-agent road information; The path generation module is configured to initialize the root node and generate the shortest path according to the multi-agent road information; The conflict detection module is configured to detect whether there is a conflict in the path. If there is no conflict, directly output the shortest path as the final solution; A conflict resolution module, configured to, if a conflict exists, resolve the conflict by inserting a node and generate a new child node; The path updating module is configured to update the path of the generated new child node; The update iteration module,is configured to,check new child nodes and repeat the conflict detection,update until a feasible solution is found.
[0014] In a third aspect, the present invention provides a computer-readable storage medium storing a plurality of instructions, wherein the instructions are suitable for being loaded by a processor of a terminal device and executing the bounded suboptimal agent path planning method based on conflict search.
[0015] In a fourth aspect, the present invention provides a terminal device comprising a processor and a computer-readable storage medium, wherein the processor is used to implement various instructions; the computer-readable storage medium is used to store multiple instructions, wherein the instructions are suitable for being loaded by the processor and executing the bounded suboptimal intelligent agent path planning method based on conflict search.
[0016] In summary, the present invention has the following beneficial technical effects: The present invention proposes a bounded suboptimal agent path planning method based on conflict search, and optimizes the conflict resolution process by introducing the FOCAL list and an improved node priority sorting strategy. This method allows the selection of conflict-free but high-cost nodes as target nodes within the suboptimal solution range, and gives priority to expanding nodes with fewer cardinality conflicts and a large number of internal constraints, avoiding redundant backtracking operations and inefficient verification. High-level search selects nodes for expansion through the FOCAL list, and low-level search updates the optimal path that meets the constraints through path replanning. This method not only ensures completeness, but also achieves rapid convergence of solutions within acceptable suboptimal limits, significantly reduces the number of node expansions and low-level path replanning times of high-level search, effectively reduces the overall amount of calculation, and can efficiently find reasonable solutions for multi-agent path planning problems. BRIEF DESCRIPTION OF THE DRAWINGS
[0017] Figure 1 Schematic diagram of a bounded suboptimal agent path planning method based on conflict search according to Embodiment 1 of the present invention; Figure 2 This is a schematic diagram of the multi-agent initial shortest path of Example 2 of the present invention; Figure 3 This is a schematic diagram of generating a partial constraint tree in Embodiment 2 of the present invention; Figure 4 This is a schematic diagram of the initial shortest path of multiple agents in Example 3 of the present invention; Figure 5 Schematic diagram of constraint tree generation based on the original CBS algorithm according to Embodiment 3 of the present invention; Figure 6 This is a schematic diagram of the multi-agent initial shortest path of Example 4 of the present invention; Figure 7 This is a schematic diagram of generating a partial constraint tree according to Embodiment 4 of the present invention. DETAILED DESCRIPTION
[0018] The present invention is further described in detail below in conjunction with the accompanying drawings.
[0019] Example 1 Reference Figure 1 , a bounded suboptimal agent path planning method based on conflict search in this embodiment includes: Obtain multi-agent road information; Initialize the root node and generate the shortest path based on the multi-agent road information; Check whether there is a conflict in the path. If there is no conflict, directly output the shortest path as the final solution; If there is a conflict, resolve the conflict by inserting the node and generate a new child node; Update the path of the generated new child node; Check new child nodes and repeat conflict detection and update until a feasible solution is found.
[0020] Specifically: A bounded suboptimal agent path planning method based on conflict search includes the following steps: Step 1, initialize the root node. Set the initial constraints of the root node to empty. Specifically, set the constraint set Root.constraints of the root node to an empty set to indicate that the current node has no constraints. This step is the initial step of the algorithm to ensure that subsequent searches are performed under unconstrained conditions.
[0021] Step 2: Generate the initial path of the agent. Generate the initial path for each agent through low-level search. Specifically: call the low-level search algorithm ( Algorithm), calculate the shortest path from the starting point to the end point for each of the k agents in turn, and store the search results in Root.solution. This step is used to plan an independent path for each agent.
[0022] Step 3: Check if there is a conflict in the path. In this step, the definition of conflict is: and Try to do it at the same time step Occupies the same vertex or the same side If the conflict occurs at a vertex, the conflict is represented as If the conflict occurs on an edge, the conflict is represented as . Determine whether there is a path conflict between agents in Root.solution. If there is no conflict, directly return Root.solution as the final solution and the algorithm ends; if there is a conflict, proceed to the next step.
[0023] Step 4: Insert the node into the OPEN list. If there is a conflict in the node, insert it into the OPEN list and prepare for conflict resolution. Specifically: insert the Root node into the OPEN list to start the subsequent conflict resolution process.
[0024] Step 5: Check whether the node meets the suboptimal limit and insert it into the FOCAL list. Check whether the cost value of the node in the OPEN list meets the suboptimal limit, that is, determine whether the cost of the node meets the formula If the condition is met, the node is inserted into the FOCAL list. The FOCAL list is used to store nodes within the suboptimal boundary to provide support for fast search for the most reasonable suboptimal solution.
[0025] Step 6: Select the best node in the FOCAL list for expansion. Select the node with the highest priority from the FOCAL list for expansion. The priority is as follows: first, sort the nodes in FOCAL according to the number of cardinality conflicts of the nodes, and expand the nodes with fewer cardinality conflicts first; if the number of cardinality conflicts is the same, sort the nodes according to the number of internal constraints, and expand the nodes with more constraints first. This step aims to quickly approach the optimal or suboptimal solution through reasonable node priority sorting.
[0026] Step 7: Handle conflicts and generate new child nodes. Identify conflicts in node P and generate new child nodes for conflicting agents. Specifically: If a conflict C is detected in node P (such as agent and At a certain locationv and time t If a conflict occurs, a new child node is generated for each conflicting agent. This step eliminates the conflict by adding additional constraints and provides new nodes for subsequent searches. Constraints refer to restrictions imposed to avoid path conflicts between multiple agents. For example, if an agent and In Location v and time t If a conflict occurs, constraints will be imposed on the two agents: Add constraints to prohibit it from t Access Location v . For the agent Add constraints to prohibit it from t Access Location v .
[0027] Step 8: Update the generated child nodes. Specifically: The constraints of the new node are added to the constraint set A.constraints of the new node and its solution A.solution is updated to generate a new path for the conflicting agents by calling the low-level search again. In addition, the cost A.cost of the new node is calculated. This step ensures that the paths of the agents are replanned under the new constraints. In this step, the cost of a node refers to the sum of individual costs, that is, the sum of the time steps required for each agent to reach its goal position.
[0028] Step 9, check the cost value of the newly generated child node. If its cost is less than infinity (a feasible solution exists), insert it into the OPEN list for subsequent search and expansion. Specifically: if A.cost < ∞, insert node A into the OPEN list and repeat steps 5 to 8. This step ensures that the algorithm's search process can continue until a conflict-free solution is found.
[0029] In step 10, the algorithm declares that a feasible solution has been found and outputs the result.
[0030] The present invention proposes a strategy based on suboptimal solutions to solve the problem that the number of CBS agents is limited and the solution time is too long. SCBS relaxes the restrictions on the target node and is no longer limited to finding a global feasible solution. It allows the selection of nodes that are not the minimum cost but have no conflicts as the target node. Through this improvement, the present invention balances the quality of the solution and the speed of the solution while ensuring completeness.
[0031] The present invention aims at the problems of high-level search node redundancy and low efficiency in the CBS algorithm, and proposes a new high-level search node sorting strategy. Unlike the traditional CBS that only expands nodes based on cost, in addition to considering the cost of the node, the number of cardinality conflicts and the number of internal constraints are introduced as additional sorting criteria. Specifically, nodes with fewer cardinality conflicts are expanded first. If the number of cardinality conflicts of multiple nodes is the same, nodes with more internal constraints are expanded first. In this way, while ensuring rationality and suboptimality, the lower bound can be expanded faster, redundant backtracking can be reduced, and the possibility of repeated addition of constraints can be reduced, thereby improving the solution efficiency.
[0032] In order to further improve the search efficiency of the algorithm, the present invention introduces the FOCAL list to store nodes whose cost is less than the preset suboptimal limit. When performing path expansion, the present invention preferentially selects nodes that meet the conditions from the FOCAL list, rather than verifying the feasibility of all cost nodes one by one. The nodes in the FOCAL list are based on the currently known minimum cost nodes, and are screened according to the set suboptimal range.
[0033] Example 2 This embodiment illustrates that the node expansion strategy of the present invention according to cardinality conflict sorting can effectively improve the efficiency of solving.
[0034] Input known map information, a group of agents , the initial position of the agent , the target position of the agent .
[0035] Step 1: Initialize the root node. There are no constraints in the root node.
[0036] Step 2, without constraints, use The algorithm plans a path for each agent and obtains a set of paths for agent A. Root.solution =( , , , ), and record the cost of the root node as 12.
[0037] Step 3: Check whether there is a conflict between nodes and determine the type of conflict. Figure 2 As shown in the figure, it is a schematic diagram of the MAPF problem in case 1. S1-S6 represents the starting position of the agent, G1-G6 represents the target position of the agent, and the arrow represents the action route of the agent from the starting position to the target position in the root node. In this case, there are 3 cardinality conflicts in the path of the root node, that is, at time step 1, the agent , A conflict occurs at position B2, and the agent , The position conflicts at E2, and the agent , A conflict occurs at position B2, recorded as , , .
[0038] Step 4: Since there is a conflict in the root node, insert the root node into the OPEN list.
[0039] Step 5, insert the node into the FOCAL list and select the cost in the OPEN list 's node and inserts it into the FOCAL list.
[0040] Step 6, select the best node in the FOCAL list for expansion. At this time, there is only one root node in the FOCAL list.
[0041] Step 7: Handle conflicts and generate new child nodes. Select one of the conflicts to handle, generate two child nodes, and add constraints to the child nodes to constrain the agents respectively. and At time step 1, the position B1 is occupied and the constraints in the two child nodes are recorded as and .
[0042] Step 8: Re-plan the path for the generated child nodes. The algorithm updates the paths in the two nodes under the new constraints and repeats steps 3 to 8 until all conflict-free paths of the group of agents A are obtained.
[0043] Compared with other optimal CBS algorithms, the present invention has the advantage of optimizing the order of node expansion and reducing unnecessary backtracking operations. Figure 3 As shown, the CBS algorithm is used to solve Figure 2 In the simplified constraint tree generated by Case 1, in each node, the numbers 12-15 on the left represent the cost of the current node path, and a1-a6 on the right represent the agents with constraints in the current node. The traditional optimal CBS algorithm uses cost as the only sorting basis when selecting nodes for expansion. When the algorithm expands to the node with a cost of 14 in the third layer of the constraint tree, the optimal CBS algorithm will backtrack to the previous layer and select a node with a cost of 13 for expansion to explore feasible solutions. Although this backtracking operation can ensure the optimality of the solution, it will lead to a large number of redundant calculations and increase the solution time.
[0044] The improved strategy proposed in the present invention uses the number of cardinality conflicts as the first order of expansion. The number of cardinality conflicts of the third-layer nodes is 1, while the number of cardinality conflicts of the second-layer nodes is 2. The present invention gives priority to expanding the third-layer nodes with fewer cardinality conflicts. In this way, the rapid increase in cost during node expansion is achieved, the repeated expansion of low-priority nodes is avoided, redundant backtracking operations are reduced, and the size of the constraint tree is effectively pruned. Among them, the nodes selected by the line box represent the redundant nodes that can be pruned by the present invention compared to the traditional CBS algorithm.
[0045] Only part of the expansion process of the CBS algorithm is shown. After expanding to the node with a cost of 15, the search for the optimal CBS algorithm is not over. This is because there are still nodes with a cost less than 15 in the constraint tree. According to the optimality principle of the CBS algorithm, these low-cost nodes still need to be expanded until all nodes with a cost of 14 are expanded and verified, and then it can be finally confirmed that the feasible solution to the problem is 14. This one-by-one verification and expansion process leads to a long solution time, which is the part optimized by the pruning strategy of the present invention.
[0046] Example 3 Step 1: Initialize the root node. There are no constraints in the root node.
[0047] Step 2, without constraints, use The algorithm plans a path for each agent and obtains a set of paths for agent A. Root.solution =( , ),in, , , and record the cost of the root node as 5.
[0048] Step 3: Check whether there is a conflict between nodes and determine the type of conflict. In this case, Arrives at the target vertex at time step 1 , and stay there until the other agents reach the target location. At time step 3, , so the agent and At time step 3 Conflicts arise.
[0049] Step 4-6: Since there is a conflict in the root node, the root node is inserted into the OPEN list, and the nodes that meet the conditions are inserted into the FOCAL list. (Same as Implementation Case 2).
[0050] Step 7: Handle the conflict and generate a new child node. At this time, there is only one conflict in the node, that is, the conflict , two child nodes are generated at this time and ,like Figure 4 As shown in Figure 2, it is a schematic diagram of the MAPF problem in Case 2, where S1 and S2 represent the starting positions of the agent, G1 and G2 represent the target positions of the agent, and the arrows represent the action routes of the agent from the starting position to the target position in the root node. Constraints are added in the two nodes to restrict the agent. At time step 3 , recorded as and .
[0051] Step 8: Re-planning at the generated sub-nodes and path. To give the agent Added constraints , call The algorithm is rewritten Plan the path, the new path is , Remain unchanged, at this time the calculation node The cost value is cost=8. To give the agent Added constraints , call The algorithm is rewritten Plan the path, the new path is , Remain unchanged. At this time, the calculation node The cost value of is cost=6.
[0052] At this time, the detection node and Is there a conflict in the path of node The path has no conflicts and its cost is 8; the node There is a conflict in the path, and its cost is 6; since the node If there is no conflict in the path, the algorithm declares that it has found a solution. In this problem, cost=8 is a feasible solution to the problem. However, some optimal solution algorithms based on CBS still need to continue to expand nodes with a cost value of 6. To find a better solution or verify the optimality of the solution until it is expanded to the node Only then can we finally verify that the feasible solution to this problem is 14.
[0053] The present invention avoids the redundant search operation performed in the CBS algorithm to verify the feasible solution by selecting a feasible but non-optimal node as the target node during operation, thereby effectively improving the solution efficiency. Under the premise of ensuring the rationality of the solution, the operation selects a feasible non-optimal node as the target node to shorten the calculation time and reduce the backtracking operation of the constraint tree.
[0054] Example 4 In multi-agent path planning problems, corridor conflict is a common type of path conflict, which refers to multiple agents trying to enter or pass through the same narrow corridor or bottleneck area at the same time. When dealing with such conflicts, the traditional CBS optimal solution algorithm often needs to perform multiple backtracking and expansions, and thus performs poorly in terms of solution efficiency. Especially in scenarios with more corridor conflicts, the algorithm may face redundant calculations and large-scale node expansion, resulting in a sharp increase in the time to solve the problem.
[0055] Figure 5 This is a schematic diagram of constraint tree generation based on the original CBS algorithm. The simplified constraint tree generated by using the CBS algorithm to solve Case 2. In each node, N1-N6 on the left represents the name of the node, and the numbers 1-8 on the right represent the path cost of the current node.
[0056] This embodiment illustrates that the present invention can effectively improve the efficiency of solving such problems. Figure 6 As shown in the figure, it is a schematic diagram of the MAPF problem in Case 3, where S1 and S2 represent the starting positions of the agents, G1 and G2 represent the target positions of the agents, and the arrows represent the action routes of the agents in the root node from the starting position to the target position. Case 3 gives a corridor map scenario with a length of 4. The optimal strategy for this problem is that one agent crosses the corridor while the other agent waits for 4 time steps at the starting position. However, to find such a solution, CBS must verify that all solutions with an increased cost of less than 4 are invalid. For the above example, the root node and To avoid the conflict of passing through (B3, C3) in the corridor at the same time, each AGent has several shortest paths, such as The path [A4, A3, B3, B3, C3, D3, D4] and There are paths [D2, D2, D3, C3, A3, A2], all of which have a waiting action and differ only in the time steps taken in the waiting action. However, these paths are still conflicting and CBS needs to be performed at least four times to find a solution. More nodes need to be expanded to verify the optimality of the solution.
[0057] When using the CBS algorithm to resolve corridor conflicts, as the corridor depth k increases, the number of expanded nodes increases exponentially, reaching .
[0058] In view of this problem, the present invention proposes an optimization strategy, which can effectively reduce the number of required node expansions, thereby improving the efficiency of resolving corridor conflicts.
[0059] Specifically, the CBS algorithm solves path conflicts by adding constraints. The number of constraints directly affects the number of node expansions. The more constraints in a node, the farther the distance from the root node to the node is, and the complexity of expansion increases accordingly. In the same layer of the constraint tree, the number of conflicts for all nodes is the same, and the CBS algorithm adopts the best-first search strategy, giving priority to expanding the node with the lowest current cost, which easily causes too much backtracking on the constraint tree.
[0060] Taking Example 3 as an example, at the same level of the constraint tree, the number of constraints between nodes is the same, and a feasible solution can usually only be obtained after the fourth level. Each branch starting from the root node has the potential to produce a feasible solution. If each branch is fully expanded, up to 16 leaf nodes containing feasible solutions may be generated, but only two of them are feasible solutions. In other words, any search within the fourth level of the constraint tree will not find a feasible solution. Therefore, prioritizing the expansion of nodes with the most constraints can speed up advanced searches, quickly expand the lower bound, obtain the target node faster, and improve the efficiency of problem solving.
[0061] Specifically, the main difference between Case 3 and Case 1 and Case 2 is that the best node in the FOCAL list is selected for expansion in step 6. In this constraint tree, all non-target nodes have only one cardinality conflict, so when selecting the node to be expanded, the number of constraints is selected as the basis for sorting. In this case, since the nodes are sorted according to the number of constraints, the directionality of the constraint tree during construction is guaranteed, that is, only one agent is added with constraints, and it is asked to wait for another agent to cross the corridor. The number of nodes generated by using the present invention to solve the corridor conflict of length 4 is 9, while the CBS algorithm needs to generate 64 nodes.
[0062] Prioritizing the expansion of nodes with a large number of constraints can effectively accelerate the expansion of the lower bound. Since these nodes contain more constraint information, unnecessary path selection can be eliminated more quickly by expanding them, thereby reducing the number of backtracking and improving the efficiency of the entire algorithm. Therefore, this strategy shows higher efficiency in resolving corridor conflicts and can find feasible solutions more quickly. Figure 7 This is a schematic diagram of partial constraint tree generation, which uses the simplified constraint tree generated in Case 3. The numbers on the left side of the box represent the cost value of agent a1, and the numbers on the right side of the box represent the cost value of agent a2.
[0063] Example 5 A bounded suboptimal agent path planning system based on conflict search, comprising: A data acquisition module is configured to acquire multi-agent road information; The path generation module is configured to initialize the root node and generate the shortest path according to the multi-agent road information; The conflict detection module is configured to detect whether there is a conflict in the path. If there is no conflict, directly output the shortest path as the final solution; A conflict resolution module, configured to, if a conflict exists, resolve the conflict by inserting a node and generate a new child node; The path updating module is configured to update the path of the generated new child node; The update iteration module,is configured to,check new child nodes and repeat the conflict detection,update until a feasible solution is found.
[0064] A computer-readable storage medium stores a plurality of instructions, wherein the instructions are suitable for being loaded and executed by a processor of a terminal device, for a bounded suboptimal agent path planning method based on conflict search.
[0065] A terminal device includes a processor and a computer-readable storage medium, wherein the processor is used to implement various instructions; the computer-readable storage medium is used to store multiple instructions, wherein the instructions are suitable for being loaded by the processor and executing the bounded suboptimal intelligent agent path planning method based on conflict search.
[0066] The above are all preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention. Therefore, any equivalent changes made based on the structure, shape, and principle of the present invention should be included in the protection scope of the present invention.
Claims
1. A bounded suboptimal agent path planning method based on conflict search, characterized in that: include: Obtain multi-agent road information; Initialize the root node and generate the shortest path based on the multi-agent road information; Check whether there is a conflict in the path. If there is no conflict, directly output the shortest path as the final solution; If there is a conflict, resolve the conflict by inserting the node and generate a new child node; Update the path of the generated new child node; Check new child nodes and repeat conflict detection and update until a feasible solution is found.
2. A bounded suboptimal agent path planning method based on conflict search according to claim 1, characterized in that: The initialization of the root node includes setting the constraint set Root.constraints of the root node to an empty set, which is used to indicate that the current node has no constraints, thereby ensuring that subsequent searches are performed under unconstrained conditions.
3. A bounded suboptimal agent path planning method based on conflict search according to claim 2, characterized in that: The method of generating the shortest path based on multi-agent road information includes generating an initial path for each agent through a low-level search, calling a low-level search A^* algorithm, calculating the shortest path from the starting point to the end point for k agents in turn, and storing the search results in Root.solution.
4. The method for bounded suboptimal agent path planning based on conflict search according to claim 3, characterized in that: The detection path is conflicting, including determining whether two agents and Are they at the same time step? Occupies the same vertex or the same side , if the conflict occurs at a vertex, the conflict is represented as , if the conflict occurs on an edge, the conflict is represented as .
5. A bounded suboptimal agent path planning method based on conflict search according to claim 4, characterized in that: The method of resolving conflicts and generating new child nodes by inserting nodes includes inserting the Root node into the OPEN list when a conflict exists, checking whether the node meets the suboptimal limit and inserting it into the FOCAL list, and selecting the best node in the FOCAL list for expansion, and generating new child nodes for the conflicting agents.
6. A bounded suboptimal agent path planning method based on conflict search according to claim 5, characterized in that: The path updating of the generated child nodes includes adding the constraints of the conflicting agent a_i to the constraint set A.constraints of the new node, and updating its solution A.solution, and generating a new path for the conflicting agent by calling the low-level search again.
7. A bounded suboptimal agent path planning method based on conflict search according to claim 6, characterized in that: The checking of new child nodes and repeated conflict detection and updating until a feasible solution is found includes checking the cost value of the newly generated child node. If the cost is less than infinity, the new child node is inserted into the OPEN list, and conflict detection and path updating are repeated until a conflict-free solution is found.
8. A bounded suboptimal agent path planning system based on conflict search, characterized in that: include: A data acquisition module is configured to acquire multi-agent road information; The path generation module is configured to initialize the root node and generate the shortest path according to the multi-agent road information; The conflict detection module is configured to detect whether there is a conflict in the path. If there is no conflict, directly output the shortest path as the final solution; A conflict resolution module, configured to, if a conflict exists, resolve the conflict by inserting a node and generate a new child node; The path updating module is configured to update the path of the generated new child node; The update iteration module,is configured to,check new child nodes and repeat the conflict detection,update until a feasible solution is found.
Citation Information
Patent Citations
Multi-robot path planning method based on conflict classification search
CN114967712A
Multi-agent path planning method and device in dynamic environment, equipment and medium
CN115981322A
Multi-vehicle cooperative path planning method in unstructured closed environment
CN117029818A
Multi-robot path planning method based on improved conflict search method
CN117270534A
Improved method based on solution space cutting
CN118780453A
Cited By
Conflict search path optimization method based on warehouse automatic guiding robot
CN119714298A
AGV network path planning method and system of multi-agent path planning algorithm based on lazy constraint addition
CN119717805A
Robot cluster scheduling method, device and equipment
CN119902446A
