Multi-agv path planning method based on path label identification and conflict search

Abstract

This invention discloses a multi-AGV path planning method based on path labeling and conflict search. The method includes: a central controller of a flexible manufacturing system plans production tasks based on user orders, generating a navigation start point and navigation end point for each AGV. If the task exists in the path trajectory label database, the path is parsed. Otherwise, a conflict search-based (CBS) multi-AGV path planning solver is used to plan a specific path for each AGV. Each AGV senses local production workshop environmental information and executes actions according to the planned path. The path trajectory information is labeled using the attribute information of each AGV when performing the path navigation task in the actual workshop. The path trajectory label database is then updated. This invention solves the problem of existing AGV path planning methods consuming significant time and computing power. By gradually updating and learning, the path trajectory label database plans more realistic and reliable paths, effectively improving the accuracy and efficiency of the path planning system.

Description

Technical Field

[0001] This invention relates to the technical field of multi-AGV path planning, and more specifically, to a multi-AGV path planning method based on path trajectory data tags in a flexible manufacturing system. Background Technology

[0002] Traditional manufacturing plants need to transform into smart factories, and the primary task in this process is to achieve unmanned operation of the workshop. Automated Guided Vehicles (AGVs), as a transportation tool, are widely used in manufacturing facilities, warehouses, and distribution to move raw materials or tools. They are an important part of realizing unmanned workshops and are gradually becoming an effective means of automating modern logistics and processing.

[0003] Multiple AGVs play a crucial role in improving the operational stability and efficiency of rail transit, and reducing production costs through collaborative task scheduling. It is an essential path for enterprises to achieve equipment automation, management informatization, and improved profitability. However, with the expansion of task volume and the increase in the number of available AGVs, the complexity of tasks such as task allocation and multi-AGV scheduling also increases. Planning collision-free optimal task scheduling for multiple AGVs in a fixed scenario remains a challenging problem for AGVs.

[0004] Multi-AGV collaborative transportation, based on learning map creation and environmental features, combines machine vision with reinforcement learning and scene understanding technologies to analyze and allocate tasks, accurately plan the paths of each AGV, and achieve optimal scheduling to complete material transportation tasks at the lowest cost. Multi-AGV collaborative task scheduling, multi-AGV path planning and autonomous navigation, intelligent obstacle avoidance, and simulation verification are the key research areas for coordinating multiple AGVs to complete automated material and tool transportation. Although some progress has been made, issues such as task conflicts between multiple AGVs and the lack of consideration for dynamic environments still exist.

[0005] Existing technologies for handling obstacle avoidance and path planning problems of multiple AGVs mostly adopt grid-based map modeling, utilizing prior knowledge of the map to handle collisions offline based on experience. However, due to the uncertainty and non-stationary dynamic characteristics of flexible manufacturing workshops, these methods are insufficient to adapt to the AGV operating environment in smart warehouses: warehouse managers and multiple AGVs act as dynamic obstacles, and without prior knowledge, it is difficult to predict their movement routes and thus avoid collisions. Summary of the Invention

[0006] The purpose of this invention is to provide a multi-AGV scheduling method that combines path labeling technology with scheduling in a flexible manufacturing workshop environment. It proposes a multi-AGV scheduling method and system that is easy to implement in engineering, fast and efficient, avoids congestion between AGVs, reduces the total AGV transportation time, and improves the material transportation efficiency in the workshop.

[0007] The technical solution adopted by this invention to achieve the above objectives is: a multi-AGV path planning method based on path label identification and conflict search, comprising the following steps:

[0008] Step S1: Plan production tasks according to user orders and generate navigation start point and navigation end point for each AGV;

[0009] Step S2: Retrieve the path trajectory label database; if the task exists in the path trajectory label database, parse the path; otherwise, use multi-AGV path planning based on conflict search to plan a path for each AGV;

[0010] Step S3: Each AGV senses environmental information and executes actions according to the planned path;

[0011] Step S4: Use the attribute information of each AGV when performing the path task in the actual workshop to add data tags to the path trajectory information;

[0012] Step S5: Update the path trajectory label database based on the data labels.

[0013] The multi-AGV path planning based on conflict search in step S2 involves planning a path for each AGV, including the following steps:

[0014] Step S21: Using the grid method, the real production workshop environment is constructed into a computer-recognizable grid map, where each grid cell serves as a path node;

[0015] Step S22: Construct a conflict search tree; where each node in the conflict tree stores three pieces of information, including: a constraint set storing conflict constraint triples (agv) i ,v,t), indicating that AGV i cannot occupy path node v at time t; the solution set stores the path set of each AGV; the cost value stores the cost value of all AGV paths;

[0016] The A* path planning algorithm is used to generate the shortest path for each AGV, the path is saved in the solution set of the root node of the conflict search tree, the path cost of the root node is calculated, and the root node is added to the Open table.

[0017] Step S23: Check if the Open table is empty. If it is empty, there is no solution to the problem; if it is not empty, proceed to step S24.

[0018] Step S24: Select the node with the lowest generation value in the Open table for conflict detection; if there are multiple nodes with the lowest generation value, arbitrarily select one of them for conflict detection.

[0019] In this node, check if there is a conflict between the paths generated by any two AGVs; if there is a conflict, add different conflict constraints to the two AGVs respectively, and this node generates two child nodes.

[0020] Step S25: For the generated child nodes, consider the conflict constraints, re-apply the A* path planning algorithm to plan the path and generate a solution set, calculate the cost value, and update the generated child nodes; add the updated generated child nodes to the Open table; and delete the nodes in the Open table that are used for conflict detection.

[0021] Step S26: Repeat steps S23 to S25 until a solution set in a certain node is found to be free of conflict, then the solution set is obtained.

[0022] The size of the grid is larger than the size of the AGV.

[0023] Step S4, which uses the attribute information of each AGV when performing the path task in the actual workshop to tag the path trajectory information with data, includes the following steps:

[0024] Step S41: Obtain attribute information of multiple AGVs during task execution;

[0025] Step S42: Based on the pre-created identifier encoding specification and path running attribute information, generate the path identifier encoding of this task as a data label; so that the corresponding data can be obtained by parsing the task identifier encoding.

[0026] Step S5: Update the path trajectory label database according to the data labels, including the following steps:

[0027] Step S51: If the data label for this task does not exist in the path trajectory label database, then store the data label from step S4 into the database;

[0028] Step S52: If the task data tag exists in the path trajectory database, retrieve the task data tag from the database, parse the actual completion time and path length, and compare it with the path length and running time of the task being executed this time;

[0029] Step S53: If the actual completion time and path length of the current task are at least one less than the actual completion time and path length of the same task in the path trajectory label database, then update the data label of the task in the database; otherwise, discard it.

[0030] The attribute information includes: task type, origin and destination of each AGV, path nodes, running time and path length.

[0031] The data tags include: workshop number, task number, number of AGVs, running time, path length, and path trajectory.

[0032] The present invention has the following beneficial effects and advantages:

[0033] 1. In existing multi-AGV path planning methods, the path is regenerated each time through the path planning algorithm, and the generated path trajectory is discarded after execution. The method proposed in this invention stores the path trajectory data calculated by the algorithm each time in the multi-AGV path planning system, and stores the actual running time of the AGV when executing the path trajectory. It also classifies the path trajectory according to factors such as the start and end points of each task, task type, time, and number of workpieces, and assigns data tags. After the path trajectory tag database executes the task multiple times, it gradually fits the actual environment of the real production workshop and generates more accurate path trajectories.

[0034] 2. In addition, the path trajectory label provides the scheduling information of AGV in the production workshop at other times, and provides the operation life cycle data of AGV to perform tasks, which facilitates the retrospective of scheduling information of processed workpieces and AGVs in the later stage, and realizes data reuse and rapid collection and reporting. Attached Figure Description

[0035] Figure 1 A flowchart illustrating a multi-AGV path planning method based on path trajectory labeling and conflict search according to the present invention;

[0036] Figure 2 A flowchart illustrating the multi-AGV path planning solver based on conflict search in this invention;

[0037] Figure 3 The conflict-based algorithm of this invention is applied to the path planning process in specific scenarios.

[0038] Figure 4 The flowchart of the path trajectory information data tagging process of the present invention.

[0039] Figure 5 A schematic diagram of the predefined, pre-created identifier encoding specification of the present invention.

[0040] Figure 6 The flowchart of the updated path trajectory label database of the present invention. Detailed Implementation

[0041] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0042] This invention discloses a multi-AGV path planning method based on path trajectory labeling and conflict search. The method includes: a central controller of a flexible manufacturing system plans production tasks based on user orders, generating navigation start and end points for each AGV. If the task exists in the path trajectory label database, the path is parsed. Otherwise, a conflict search-based (CBS) multi-AGV path planning solver is used to plan a specific path for each AGV. Each AGV senses local production workshop environment information and executes actions according to the planned path. Using the attribute information of each AGV when performing the path navigation task in the actual workshop, including: task type, start and end points of each AGV, path nodes, running time, and path length, data labels are added to the path trajectory information based on a pre-created label coding standard and path running attribute information. The path trajectory label database is then updated. This invention solves the problem of existing AGV path planning methods consuming significant time and computing power. The path trajectory label database, through gradual learning and updating, plans more realistic and reliable paths, effectively improving the accuracy and efficiency of the path planning system.

[0043] like Figure 1 As shown, an embodiment of the present invention provides a multi-AGV path planning method based on path trajectory label identification and conflict search, comprising the following steps:

[0044] Step S1: The central controller of the flexible manufacturing system plans production tasks based on user orders and generates the navigation start point and navigation end point for each AGV.

[0045] Step S2: Retrieve the path trajectory label database. If the task exists in the database, parse the path. Otherwise, use a conflict search-based (CBS) multi-AGV path planning solver to plan a specific path for each AGV.

[0046] Step S3: Each AGV senses the local production workshop environment information and executes actions according to the planned path.

[0047] Step S4: Use the attribute information of each AGV when performing the path navigation task in the actual workshop, including: the task type of this task, the start and end points of each AGV, path nodes, running time and path length, etc., to add data tags to the path trajectory information.

[0048] Step S5: Update the path trajectory label database.

[0049] like Figure 2 As shown, the flowchart of the multi-AGV path planning solver algorithm based on conflict search in step S2 includes:

[0050] Step S21: Using the grid method, the size of the grid is based on the size of the AGV itself, and the real production workshop environment is constructed into a grid map that can be recognized by the computer. The grid that does not contain obstacles such as workstations within its range is called a free grid; otherwise, it is called an obstacle grid.

[0051] Step S22: Construct the conflict search tree. Each node in the conflict tree stores three pieces of information: 1. The constraint set (constraints) stores the conflict constraint triples (agv...). i 1. The path set is defined as follows: A path i cannot occupy path node v at time t. 2. The solution set stores the path set for each AGV. 3. The cost set stores the cost of all AGV paths, for example, using path length to represent the cost. The root node constraint set is empty. The A* path planning algorithm is used to generate the shortest path for each AGV, and the path is saved to the solution set of the root node of the conflict search tree. The cost of the root node path is calculated. The root node is then added to the Open table.

[0052] Step S23: Check if the Open table is empty. If it is empty, the problem has no solution. If it is not empty, proceed to step S24.

[0053] Step S24: Select the node with the lowest value in the Open table for conflict detection, and check whether the generated path conflicts with other paths in pairs. If a conflict exists, add different conflict constraints to the two AGVs respectively, and generate two child nodes in the conflict search tree to ensure that subsequent planning will resolve the conflict.

[0054] Step S25: Consider conflict constraints in the child nodes, and reapply the A* path planning algorithm to plan the path and generate a solution set. This planning will consider the added constraints, resolve detected conflicts, and calculate the cost. Add the generated child nodes to the Open list, and pop the nodes that were just checked for conflicts from the Open list.

[0055] Step S26: Repeat steps S23 to S25 until a path solution set in a certain node is found to be free of conflict, then the solution set is obtained.

[0056] like Figure 3 As shown, the path planning process based on the conflict algorithm is fully described using two AGVs in a specific map scenario, considering, for example... Figure 3 The left image shows a 3×3 grid map scene, where AGV1 starts at [C,1] and targets at [A,3]; AGV2 starts at [C,3] and targets at [A,3]; [B,1] and [B,3] are obstacles that the AGVs cannot pass through.

[0057] First, construct the root node of the conflict search tree, such as... Figure 3 As shown in the right figure, the conflict set `constraints` of the root node is empty. Using the A* algorithm, shortest path solutions from the starting point to the target position are planned for AGV1 and AGV2 respectively. The shortest path for AGV1 is {[C,1],[C,2],[B,2],[A,2],[A,3]}, and the shortest path for AGV2 is {[C,3],[C,2],[B,2],[A,2],[A,1]}, which are added to the solution set `solution`. The cost of the root node is calculated to be 8.

[0058] Then, the root node is added to the Open table, and conflict detection is performed on the solution set of the root node. AGV1 and AGV2 first conflict at the [C,2] path node at time t=2. Constraints are added to AGV1: constraints = {(agv1,[C,2],2)}, meaning that AGV1 cannot occupy the [C,2] node at time t=2. Constraints are added to AGV2: constraints = {(agv2,[C,2],2)}, meaning that AGV2 cannot occupy the [C,2] node at time t=2. Two child nodes are generated.

[0059] The left node is solved using A* with constraints {(agv1,[C,2],2)}, yielding a solution set with a cost of 9. The right node is also solved using A* with constraints {(agv2,[C,2],2)}, yielding a solution set with a cost of 9. Both nodes are added to the Open list, and the root node is removed from the Open list.

[0060] Take the node with the smallest value in the Open table. If two nodes have the same value, randomly select one of them. Taking the left node as an example, continue to check for conflicts in each pair of the solution set. If there are no conflicts, then this node is the target node and the solution set is the optimal solution.

[0061] like Figure 4 As shown, in step S4, the attribute information of each AGV when performing the path navigation task in the actual workshop is used to tag the path trajectory information with data, including:

[0062] Step S41: Obtain attribute information of multiple AGVs during task execution, including the start and end points of each AGV in the task, task type, path nodes, running time, path length, etc.

[0063] Step S42: Generate the path identifier code for this task based on the pre-created identifier encoding specification and path running attribute information.

[0064] Step S43: Write the task attribute information into the task identifier code of this task so that the corresponding data can be obtained by parsing the task identifier code.

[0065] like Figure 5 The diagram illustrates a pre-created identifier encoding specification according to an embodiment of the present invention. The number of bits in the tag generated by this encoding rule is not fixed, but is generated based on the number of AGVs in the order task.

[0066] like Figure 6 As shown, step S5: updating the path trajectory label database includes:

[0067] Step S51: If the task data label does not exist in the path trajectory label database, then store the path trajectory label from step S4 into the database.

[0068] Step S52: If the task data tag exists in the path trajectory database, parse and obtain the actual completion time and path length in the database, and compare it with the path length and running time of the task being executed.

[0069] Step S53: If it is better than the path trajectory label database, then update the database information. Otherwise, discard it.

Claims

1. A multi-AGV path planning method based on path label identification and conflict search, characterized in that, Includes the following steps: Step S1: Plan production tasks according to user orders and generate navigation start point and navigation end point for each AGV; Step S2: Retrieve the path trajectory label database; if the task exists in the path trajectory label database, parse the path; otherwise, use multi-AGV path planning based on conflict search to plan a path for each AGV; Step S3: Each AGV senses environmental information and executes actions according to the planned path; Step S4: Use the attribute information of each AGV when performing the path task in the actual workshop to add data tags to the path trajectory information; Step S5: Update the path trajectory label database based on the data labels; Step S4, which uses the attribute information of each AGV when performing the path task in the actual workshop to tag the path trajectory information with data, includes the following steps: Step S41: Obtain attribute information of multiple AGVs during task execution; Step S42: Based on the pre-created identifier encoding specification and path running attribute information, generate the path identifier encoding of the task as a data label; so that the corresponding data can be obtained by parsing the task identifier encoding; Step S5: Update the path trajectory label database according to the data labels, including the following steps: Step S51: If the data label for this task does not exist in the path trajectory label database, then store the data label from step S4 into the database; Step S52: If the task data tag exists in the path trajectory database, retrieve the task data tag from the database, parse the actual completion time and path length, and compare it with the path length and running time of the task being executed this time; Step S53: If the actual completion time and path length of the current task are at least one less than the actual completion time and path length of the same task in the path trajectory label database, then update the data label of the task in the database; otherwise, discard it. The attribute information includes: task type, origin and destination of each AGV, path nodes, running time and path length; The data tags include: workshop number, task number, number of AGVs, running time, path length, and path trajectory.

2. The multi-AGV path planning method based on path label identification and conflict search according to claim 1, characterized in that, The multi-AGV path planning based on conflict search in step S2 involves planning a path for each AGV, including the following steps: Step S21: Using the grid method, the real production workshop environment is constructed into a computer-recognizable grid map, where each grid cell serves as a path node; Step S22: Construct a conflict search tree; where each node in the conflict tree stores three pieces of information, including: a constraint set storing conflict constraint triples. , indicating that AGV i cannot occupy path node v at time t; the solution set stores the path set of each AGV; the cost value stores the cost value of all AGV paths; Among them, A is used The path planning algorithm generates the shortest path for each AGV, saves the path to the solution set of the root node of the conflict search tree, calculates the path cost of the root node, and adds the root node to the Open table. Step S23: Check if the Open table is empty. If it is empty, there is no solution to the problem; if it is not empty, proceed to step S24. Step S24: Select the node with the lowest generation value in the Open table for conflict detection; if there are multiple nodes with the lowest generation value, arbitrarily select one of them for conflict detection. In this node, check if there is a conflict between the paths generated by any two AGVs; if there is a conflict, add different conflict constraints to the two AGVs respectively, and this node generates two child nodes. Step S25: Consider conflict constraints for the generated child nodes and reuse A. The path planning algorithm plans a path, generates a solution set, calculates the cost, and updates the generated child nodes; it adds the updated child nodes to the Open table; and it deletes the nodes in the Open table that are used for conflict detection. Step S26: Repeat steps S23 to S25 until a solution set in a certain node is found to be free of conflict, then the solution set is obtained.

3. The multi-AGV path planning method based on path label identification and conflict search according to claim 2, characterized in that, The size of the grid is larger than the size of the AGV.

Images