Multi-AGV path planning method and device based on time window and computer equipment

Abstract

The invention relates to a multi-AGV path planning method and device based on a time window and computer equipment. The method comprises the following steps: constructing a working environment map of AGVs by adopting a grid method; according to the working environment map, performing path planning on each AGV by adopting an improved Floyd bidirectional search JPS algorithm to obtain an initial global optimal path when each AGV executes a respective delivery task, constructing a time window model of the AGV, and according to the time window model and the initial global optimal paths when all the AGVs execute respective delivery tasks, constructing a time window model of the AGV, and according to the time window model and the initial global optimal paths when all the AGVs execute the respective delivery tasks, constructing a time window model of the AGV; and local path planning is carried out by adopting a conflict resolution strategy in which a translation time window algorithm and a re-planning strategy are parallel, and a conflict-free path of each AGV is obtained. According to the method, aiming at possible conflicts among multiple AGV paths, local path planning is carried out by adopting a time window and replanning mixed conflict resolution method, conflicts among the AGVs are avoided, and more reliable and efficient gains are achieved.

Description

Technical Field

[0001] The present application relates to the technical field of path planning, and in particular to a method, device and computer equipment for multi-AGV path planning based on a time window. Background Art

[0002] With the automation and intelligence of manufacturing, AGV is becoming more and more important in the current automation laboratory, and path planning technology is an important part of it. The core of AGV path planning is the classic path finding problem, that is, to plan the optimal path from the starting point to the end point for each AGV. The so-called "optimal" can be the shortest distance, the shortest driving time, the least turning nodes, etc. The essence of the path planning problem is to search for the shortest path between two specified points, and then develop into finding a collision-free shortest path from the starting point to the target point under specific constraints.

[0003] Existing path planning can be divided into global planning and local planning according to the scope of environmental information acquisition. Global planning, also known as static planning, refers to planning an obstacle-free driving path from the starting point to the end point for each AGV in a static working environment. The route design and driving execution are relatively simple, but the ability to handle sudden problems during operation is weak, and there are problems such as deadlock and system paralysis. Local path planning, also known as dynamic planning and obstacle avoidance path planning, is to obtain the surrounding environment information in real time through sensors after the AGV obtains the optimal path, and consider how to quickly respond to dynamic events such as temporary insertion, AGV failure or changes in local road conditions, so as to achieve timely adjustment of subsequent paths during the operation of the AGV. Local path planning has a strong ability to handle conflicts, but it is not easy to obtain the global optimal solution, and the algorithm is relatively complex. Static planning problems are relatively mature. Commonly used algorithms include Dijkstra algorithm, A* algorithm, genetic algorithm, etc., but this static planning has poor responsiveness to unexpected situations and cannot meet actual transportation needs. In addition to avoiding collisions with static obstacles in the workshop, conflicts between AGVs should also be avoided during path planning. The existing technology does not have strong flexibility and reliability in avoiding the combined problem of conflicts between static obstacles and AGVs in the same environment. Summary of the invention

[0004] Based on this, it is necessary to provide a multi-AGV path planning method, device and computer equipment based on a time window to address the above technical problems.

[0005] A multi-AGV path planning method based on a time window, the method comprising: The grid method is used to construct the working environment map of AGV.

[0006] According to the working environment map, the improved Floyd two-way search JPS algorithm is used to plan the path of each AGV to obtain the initial global optimal path of each AGV when performing its respective delivery tasks; the improved Floyd two-way search JPS algorithm is used to plan the path of each AGV using the JPS algorithm that introduces a two-way search mechanism, and then the Floyd algorithm is used to optimize the path planning results to obtain the initial global optimal path of each AGV when performing its respective delivery tasks.

[0007] Construct a time window model for AGV.

[0008] According to the time window model and the initial global optimal path of all AGVs when performing their respective delivery tasks, a conflict resolution strategy that combines a shifting time window algorithm with a replanning strategy is used to perform local path planning to obtain a conflict-free path for each AGV; the conflict resolution strategy refers to using a shifting time window method to resolve path conflicts under normal driving conditions of AGVs, and using a path replanning method to resolve path conflicts under abnormal driving conditions of AGVs.

[0009] A multi-AGV path planning device based on a time window, the device comprising: The working environment map construction module is used to construct the working environment map of the AGV using a grid method.

[0010] The initial global optimal path planning module is used to plan the path of each AGV using the improved Floyd two-way search JPS algorithm according to the work environment map, so as to obtain the initial global optimal path of each AGV when performing its respective delivery tasks; the improved Floyd two-way search JPS algorithm is used to plan the path of each AGV using the JPS algorithm that introduces a two-way search mechanism, and then the Floyd algorithm is used to optimize the path planning results, so as to obtain the initial global optimal path of each AGV when performing its respective delivery tasks.

[0011] The time window model building module is used to build the time window model of AGV.

[0012] The local path planning module based on time windows is used to perform local path planning based on the time window model and the initial global optimal path of all AGVs when performing their respective delivery tasks, using a conflict resolution strategy that combines a shifting time window algorithm with a replanning strategy to obtain a conflict-free path for each AGV; the conflict resolution strategy refers to using a shifting time window method to resolve path conflicts under normal driving conditions of the AGV, and using a path replanning method to resolve path conflicts under abnormal driving conditions of the AGV.

[0013] The above-mentioned multi-AGV path planning method, device and computer equipment based on time window, the method includes: using the grid method to construct the working environment map of AGV; according to the working environment map, using the improved Floyd two-way search JPS algorithm to plan the path of each AGV, to obtain the initial global optimal path of each AGV when performing its respective delivery task, and constructing the time window model of AGV; according to the time window model and the initial global optimal path of all AGVs when performing their respective delivery tasks, using the conflict resolution strategy of the translation time window algorithm and the replanning strategy in parallel to perform local path planning, to obtain the conflict-free path of each AGV. This method uses a conflict resolution method mixed with time window and replanning to perform local path planning in view of the possible conflicts between the paths of multiple AGVs, avoids conflicts between AGVs, and achieves more reliable and efficient gains. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a flowchart of a multi-AGV path planning method based on a time window in one embodiment; Figure 2 An actual working environment map established by using a grid method in one embodiment; Figure 3 A schematic diagram of a forward and reverse alternating search in one embodiment; Figure 4 A schematic diagram of a time window in which an AGV passes through a node in one embodiment; Figure 5 A schematic diagram of reducing turning points using the Floyd algorithm in another embodiment; Figure 6 It is a Floyd optimization path map in another embodiment, (a) is the initial path map, (b) is the path map after Floyd optimization; Figure 7 A schematic diagram of the improved Floyd bidirectional search JPS algorithm flow in another embodiment; Figure 8 is a comparison diagram of algorithm operation results in another embodiment, wherein (a) is a diagram of traditional A* operation results, (b) is a diagram of Floyd improved A* operation results, (c) is a diagram of improved evaluation function operation results, (d) is a diagram of Floyd+ improved evaluation function operation results, (e) is a diagram of bidirectional JPS operation results, and (f) is a diagram of Floyd improved bidirectional JPS operation results; Fig. 9 is the time window model of the kth AGV in another embodiment; Fig.10 A schematic diagram of the process flow of a translation time window algorithm and a replanning hybrid obstacle avoidance strategy in another embodiment; Fig.11Schematic diagram of conflict types in another embodiment, wherein (a) is a schematic diagram of an approaching conflict, (b) is a schematic diagram of an encounter conflict, (c) is a schematic diagram of an overtaking conflict, and (d) is a schematic diagram of a stopping conflict; Fig.12 It is an experimental simulation map in another embodiment; Fig.13 is an initial path planning diagram of AGVS in another embodiment; Fig.14 is a path graph after the first path replanning in another embodiment; Fig.15 is a path diagram after the second path replanning in another embodiment; Fig.16 It is a structural block diagram of a multi-AGV path planning device based on a time window in one embodiment; Fig.17 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0015] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0016] In one embodiment, Figure 1 As shown, a multi-AGV path planning method based on a time window is provided, and the method comprises the following steps: Step 100: Use the grid method to construct the working environment map of the AGV.

[0017] Specifically, the working environment of the AGV in this application is a chemical laboratory, and the grid method is used to construct the digital map. In order to better approach the actual environment, the following settings are made: (1) Each grid is a square with a side length of l and has a unique ID; (2) Each grid can accommodate at most one AGV at a time; (3) Each grid is bidirectional at different times and unidirectional at the same time; The working environment map of AGV is established based on the grid method. The size of each grid is determined by the size of the obstacle and the size of the AGV. The actual working environment map established by the grid method is as follows: Figure 2 As shown. In general, the unit grid size is set to the AGV size.

[0018] Step 102: According to the work environment map, the improved Floyd two-way search JPS algorithm is used to plan the path of each AGV to obtain the initial global optimal path of each AGV when performing its respective delivery tasks; the improved Floyd two-way search JPS algorithm is used to plan the path of each AGV using the JPS algorithm that introduces a two-way search mechanism, and then the Floyd algorithm is used to optimize the path planning results to obtain the initial global optimal path of each AGV when performing its respective delivery tasks.

[0019] Specifically, single AGV path planning is an important link in realizing material transportation, and it is also the basis for multi-AGV path planning. The single AGV path planning problem can be simply described as: finding a collision-free shortest path from the starting point to the target point under the premise that the environment is known or partially known.

[0020] In the prior art, the JPS algorithm based on the grid map has the advantages of short path, small amount of calculation, low resource consumption, and will not fall into the local optimum and cause the target point to be unreachable, compared with the sampling-based algorithm and various intelligent optimization algorithms. However, there are still problems such as high memory consumption, poor path smoothness, and inability to adapt to dynamic environments. When planning the path of the JPS algorithm, energy consumption and efficiency are usually not considered, and only the shortest path is used as the main indicator. In the actual AGV operation scenario, the operating environment is usually simple and standardized, and the requirement for the shortest path is less than the requirement for high efficiency of the algorithm. Therefore, it is often necessary to reduce the search quality and increase the search rate under the premise of ensuring a certain search quality to meet the algorithm's requirements for system real-time performance. At the same time, in the path planning of multiple AGVs, the JPS algorithm may face problems such as a large number of turning points, long calculation time, long waiting time, and too many rescheduling times, and it is impossible to dynamically adjust the search efficiency according to the environment. Bidirectional search is a strategy for searching from the starting node and the target node at the same time. The basic idea is to search from the starting node and the target node separately. The usual termination condition is to search for the same expansion node in both directions. However, it is also possible that the same expansion node does not appear in the two directions during the search process, resulting in path search failure. Therefore, this method adopts a bidirectional search mechanism of alternating forward and reverse optimization, which can effectively prevent path search failure. The subnode of the previous optimization is used as the target node in the latter optimization, and the search is stopped when the same expansion node appears. Figure 3 It is a schematic diagram of the positive and negative alternating search mechanism. S represents the starting node, G Indicates the target node. S Perform a forward search to find the optimal expansion node S 1; then start the reverse search, G Point is the starting position, S Start searching as the target node and get the expanded node G1. Continue searching back and forth until the same expansion node appears in both the forward and reverse directions, then stop searching. The path obtained at this time is the optimal path.

[0021] To address the problem of too many turning points, the Floyd shortest path idea is introduced to optimize the path, effectively reducing the number of turns in the path planning process. To address the problem of long algorithm running time, a two-way search mechanism is proposed to effectively reduce the algorithm operation time. To address the problem of too long waiting time, a waiting time penalty coefficient is added to reduce the long waiting time caused by conflicts during path planning. To address the problem of too many rescheduling times, a rescheduling efficiency penalty is added to reduce the number of rescheduling times during path planning. To address the problem of fixed search efficiency, an adaptive evaluation function is introduced to adaptively adjust the evaluation function according to the number of obstacles within the area from the node to the target point. .

[0022] Step 104: Construct a time window model of the AGV.

[0023] Specifically, the time window method assumes that there is no external interference and that the AGV runs stably and without errors. It is an ideal predictive obstacle avoidance algorithm. It describes the entry and exit time of each AGV at each node in the map model. When generating each path, the time window occupied by the AGV passing through each path node is generated at the same time. By comparing with the node time windows of all previously planned paths, it is determined whether the time window of the node is repeatedly occupied to determine whether there are conflicts and collisions between different AGVs. Subsequently, appropriate solutions are adopted, such as stretching, compression or translation, to resolve the conflict. In order to better describe the time window model, the following settings are made: (1) The AGV travels at a constant speed during the straight-line operation, and the driving speed is ; (2) During the acceleration and deceleration process of AGV, the acceleration is remains unchanged; (3) The turning speed of the AGV is fixed at ; (4) The loading and unloading time of the AGV in the execution of the task is not considered; (5) The length of the grid is ; AGV models are the same, body length is .

[0024] AGV K Rows enter and leave the grid The moment is the occupation time window of the grid point, The AGV passes through the node time window as follows. Figure 4 shown.

[0025] By calculating the time windows of the grid nodes that all AGVs pass through within a period of time, the occupancy of the global nodes on the time axis can be obtained. By finding the overlapping positions of each AGV time window, the conflicting nodes, paths, and AGVs can be determined, and then corresponding adjustments can be made according to the type of conflict to solve congestion and conflict problems. At the same time, the time window can provide path and time information for subsequent AGV path planning to avoid conflicts.

[0026] Step 106: Based on the time window model and the initial global optimal paths of all AGVs when performing their respective delivery tasks, a conflict resolution strategy that uses a shifting time window algorithm in parallel with a replanning strategy is used to perform local path planning to obtain a conflict-free path for each AGV; the conflict resolution strategy refers to using a shifting time window method to resolve path conflicts under normal driving conditions of the AGV, and using a path replanning method to resolve path conflicts under abnormal driving conditions of the AGV.

[0027] In the above-mentioned multi-AGV path planning method based on time window, the method includes: using the grid method to construct the working environment map of AGV; according to the working environment map, using the improved Floyd bidirectional search JPS algorithm to plan the path of each AGV, obtaining the initial global optimal path of each AGV when performing its respective delivery task, constructing the time window model of AGV, and according to the time window model and the initial global optimal path of all AGVs when performing their respective delivery tasks, using the conflict resolution strategy of parallel translation time window algorithm and replanning strategy to perform local path planning, and obtaining the conflict-free path of each AGV. This method uses the conflict resolution method of hybrid time window and replanning to perform local path planning in view of the possible conflicts between multiple AGV paths, avoids conflicts between AGVs, and achieves more reliable and efficient gains.

[0028] In one embodiment, step 102 includes: using the JPS algorithm to perform path planning for each AGV to obtain an initial global optimal path; using the starting node and the target node of the initial global optimal path of each AGV as the forward search starting node and the reverse search starting node respectively; using the JPS algorithm to search the forward search starting node and the reverse search starting node for each AGV at the same time, and when the same extension node is searched in two directions, the initial global path of each AGV is obtained; and using the Floyd algorithm to optimize the initial global path of each AGV to obtain the initial global optimal path of each AGV when performing its respective delivery tasks.

[0029] In one of the embodiments, for each AGV, the JPS algorithm is used to search from the forward search start node and the reverse search start node at the same time. When the same extension node is searched in two directions, the initial global path of each AGV is obtained, including: taking the forward search start node as the forward parent node, taking the reverse search start node as the reverse parent node, and adding the forward parent node and the reverse parent node to the forward search node list and the reverse search node list respectively; the forward parent node and the reverse parent node take each other as the target node and perform hop search respectively; adding the extension node of the forward parent node to the forward search node list, and adding the extension node of the reverse parent node to the reverse search node list; calculating the cost valuation of the extension node; determining whether the extension node with the smallest cost valuation obtained by the forward and reverse searches is the same node; if not, taking the extension node obtained by the forward search as the new parent node of the forward search and putting it into the forward search node list, taking the extension node with the smallest cost valuation obtained by the reverse search as the new parent node of the reverse search and putting it into the reverse search node list, and continuing the hop search; if yes, the initial global path of the AGV is obtained.

[0030] In one embodiment, the cost estimate of calculating the expansion node is: (1) (2) in, is the evaluation function, is the environmental barrier rate, h ( n ) is the heuristic function, To add the estimated waiting time, is the actual cost function, It is the starting point of AGV. is the end point of the AGV. The number of obstacles that form a rectangular area for the start and end points.

[0031] Specifically, the derivation principle of the cost estimation of the expansion node is as follows: (1) Consider the waiting time penalty coefficient Go to grid from current grid In the process of Conflicts occur, leading to Need to wait in the current grid. However, if Go to another grid , will not conflict with any other AGV. Pausing and waiting will reduce the efficiency of the road network and increase the probability of detours and rescheduling. When estimating the cost from the current location to the next location, add the estimated waiting time , which can effectively reduce the waiting time.

[0032] (2) Integrating the Floyd algorithm to reduce the number of turning points Due to the limitations of the JPS search algorithm, a large number of turning points will be generated, which will also increase the path length. This patent uses the Floyd algorithm to optimize the generated path, remove redundant points, and reduce turning points. The Floyd algorithm, also known as the insertion point method, is an algorithm used to solve the shortest distance between two points. Its principle is as follows Figure 5 shown.

[0033] definition Represents the distance between two points. Indicate point Directly to the point without passing through other points If there is an obstacle between two points, the distance between the two points is set to infinity. .

[0034] yes , If the points between the expanded , then let ,use As , The waypoints between Change to .

[0035] if , There is a middle point between ,and , then let , discard point B as the waypoint and replace it with point C, Change to The path diagram before and after Floyd algorithm optimization is as follows Figure 6 shown.

[0036] (3) Introducing obstacle rate to improve the evaluation function The evaluation function of the traditional A* algorithm is composed of the actual cost function and the heuristic function Among them, the heuristic function It plays a leading role in the search performance of the traditional A* algorithm. However, in practical applications, there are often obstacles, which leads to a larger search space, reduced efficiency, and more redundant nodes. When there are obstacles between the starting point and the end point, the path estimated by the Euclidean distance may be different from the actual path, making the algorithm slower and less accurate. The presence of obstacles leads to multiple optional paths, and the shortest path may also be greater than the Euclidean distance between the starting point and the end point. In order to more accurately estimate the shortest path, the environmental obstacle rate is introduced Abstract response environment complexity. Assume that the starting point of the mobile robot is , the end point is , the number of obstacles in the rectangular area formed by the starting point and the end point is , then the ratio of obstacles As shown in formula (2).

[0037] Ratio of environmental obstacles It is determined by the number of obstacles and the starting and ending points, and its size changes with the change of the starting and ending points. Introducing evaluation function , and change the heuristic function h ( n ) weights to realize the evaluation function Adaptive adjustment of . Heuristic function h ( n ) Select Euclidean distance, the improved evaluation function As shown in formula (1). From formula (1), we can see that when there are many obstacles in the environment, the heuristic function h ( n )coefficient The algorithm shrinks the search space and improves the search efficiency. On the contrary, when there are fewer obstacles in the environment, the heuristic function h ( n ) The algorithm expands the search space, reduces the search speed, and avoids falling into the local optimum.

[0038] In a specific embodiment, the improved Floyd bidirectional search JPS algorithm process is as follows: Figure 7 As shown in the figure, this method reduces the number of traversed nodes during the path search process, thereby speeding up the algorithm's operation speed. The bidirectional alternating search strategy reduces the algorithm's search time to a certain extent. The introduction of the Floyd algorithm further reduces redundant points and optimizes the smoothness of the final path. The specific algorithm implementation steps are as follows: Step 1: Algorithm initialization. Add the starting node S to the SP list as the starting node for forward search, and add the target node G to the GP list as the starting node for reverse search; Step 2: Jump point search. Take S and G nodes as parent nodes and each other as target nodes, and perform jump point search respectively. Add the coordinates of the two parent nodes to the SC list and GC list; Step 3: Add expansion nodes. Add the expansion nodes of the positive parent node to the SP list, and add the expansion nodes of the negative parent node to the GP list. Calculate the cost estimates of the expansion nodes according to formula (1) respectively, and select the node with the smallest cost estimate among the expansion nodes as the parent node for the next expansion; Step 4: Determine and update. Determine whether the expansion node with the smallest cost estimate found in the forward and reverse searches is the same node. If not, put the expansion node found in the forward search into the SC list as the new parent node for the forward search. Put the expansion node with the smallest estimate found in the reverse search into the GC list as the new parent node for the reverse search. Then return to step (2) to continue the algorithm. If they are the same, it means that the algorithm has found the optimal path. According to the parent node of the final node in the SC list and GClist, the path traversed in reverse is the optimal path found, and each node is added to the optimal path.

[0039] Step 5: Use Floyd algorithm to optimize the path and reduce turning points.

[0040] In a verification example, in order to verify that the improved JPS algorithm proposed in this application has stronger practicality and feasibility, a 31*31 grid map model is built in MATLAB 2020 for comparative simulation. While ensuring that the starting point, end point and obstacle information are the same, the A* algorithm, Floyd improved A* algorithm, improved evaluation function A* algorithm, Floyd improved evaluation function A* algorithm, JPS jump point algorithm, and Floyd-based improved bidirectional JPS algorithm are compared. The circle represents the starting point and the triangle represents the end point. The running results of each algorithm are shown in the figure below. Figure 8 As shown, (a) is the traditional A* running result graph, (b) is the Floyd improved A* running result graph, (c) is the improved evaluation function running result graph, (d) is the Floyd + improved evaluation function running result graph, (e) is the bidirectional JPS running result graph, and (f) is the Floyd improved bidirectional JPS running result graph. Table 1 lists the optimal path length, search time, number of expanded nodes, and number of turning points of the algorithm.

[0041] Table 1 Path planning of each algorithm

[0042] Analysis of the data in Table 1 shows that, by comparing the traditional A* algorithm and Floyd's improved A* algorithm, the improved evaluation function A* algorithm and Floyd's improved evaluation function A* algorithm, the bidirectional JPS algorithm and Floyd's improved bidirectional JPS algorithm, it can be seen that although the Floyd algorithm slightly prolongs the path planning time, it can shorten the final path and reduce a large number of turning points. Compared with the traditional A algorithm and the bidirectional JPS algorithm, although the path length planned by the bidirectional JPS algorithm is slightly longer than that of the A algorithm, it has achieved significant improvements in planning time, expansion nodes, and path turning points. Compared with the traditional A* algorithm, the Floyd's improved bidirectional JPS algorithm reduces planning time by 13.45%, expansion nodes by 70%, and path turning points by 60%.

[0043] In view of the fact that this application studies the path planning problem of multiple AGVs, it may be necessary to re-plan the path multiple times. If the A* algorithm is selected, not only is the real-time performance poor and the time consumption long, but the searched path is close to obstacles and the path turns many times, which is not conducive to reducing the total energy consumption of the system. Therefore, Floyd's improved two-way JPS is selected as the path planning algorithm for a single AGV. The experimental simulation results show that compared with the traditional A Compared with the algorithm, JPS algorithm and Floyd's improved evaluation function A* algorithm, this algorithm significantly shortens the algorithm running time, reduces the number of node expansions, and reduces the number of turns.

[0044] In one embodiment, step 104 includes: k AGVs enter and leave the grid i The moment is defined as the time window occupied by the grid point; the time windows of all grid points passed by AGVs within a period of time are calculated to obtain the time window model of each AGV; the time window set of all AGVs on all sections is. The time window set of all AGVs on all sections is: (3) In the formula, Indicates the total number of path nodes; Indicates the total number of AGVs, For the k The time window set of all path points of the AGV, Representation Node Relative to the k The time window of an AGV, where i =1,2,3,……, n , k =1,2,3,……, m .

[0045] Specifically, in a multi-AGV system, the mathematical description of the time window is: Assume that in the time period Inside AGVs need to perform transportation tasks, using a collection express; The path node set is ;node Relative to The time window is , The path point time set is . The time window model is as follows Fig. 9 Shown: The time window set of all AGVS on all road sections is shown in formula (3). Each column in formula (3) represents the time window set of an AGV on its task section.

[0046] In one embodiment, the time window of all grid points passed by the AGV within a period of time is calculated, including: dividing the grid points passed by the AGV during operation into start nodes, stop nodes, turning nodes, and straight-line motion passing nodes; when the grid point is a start node, the first k AGV in the first s The relationship satisfied by the time window of each grid point is: (4) (5) (6) in, Respectively represent k AGV in the s The start and end time of the time window of each grid point, represents the Euclidean distance from the AGV starting point to the task starting point, Indicates the starting point of AGV, represents the starting point of the AGV mission, L represents the length of the vehicle body, l represents the grid length, Indicates the driving speed of the AGV.

[0047] When the grid point moves in a straight line through the node, k AGV in the first i The relationship satisfied by the time window of each grid point is: (7) (8) (9) in, Respectively represent k AGV in the iThe start and end time of the time window of each grid point, Grid points i To grid point i -1 Euclidean distance, Indicates that AGV is in the i The coordinate positions of the grid points, Indicates that AGV is in the i -1 grid point coordinate position.

[0048] When the grid point is a turning node, k AGV in the first i The relationship satisfied by the time window of each grid point is: (10) (11) (12) in, a represents acceleration, Indicates the steering speed, Indicates the k AGV in the i -1 grid point time window end time.

[0049] When the grid point is a stop node, k The relationship satisfied by the time window of an AGV at the e-th grid point is: (13) (14) (15) in, Respectively represent k AGV in the e The start and end time of the time window of each grid point, is the Euclidean distance from the stop point to the previous grid point, Indicates that AGV is in the e The coordinate positions of the grid points, Indicates that AGV is in the e -1 grid point coordinate position.

[0050] Specifically, the reliability and accuracy of the time window are affected by the accessibility of the path and the accuracy of time calculation, so it is necessary to accurately describe the operating status of the AGV. In order to obtain a more accurate time window of the vehicle path grid point, it is necessary to classify the motion status of the vehicle path node.

[0051] The grid points that the AGV passes through during operation are divided into start nodes, stop nodes, turn nodes, and linear motion nodes. The calculation method of the node time window is given below for different types of nodes.

[0052] (1) Start the node No. k The AGV is in the grid points, starting from rest with acceleration Accelerate to driving speed , time window The relationship shown in formula (4) to formula (6) is satisfied.

[0053] (2) Linear motion through a node When driving in a straight line, k Each AGV travels at a constant speed from the time it reaches the edge of the node until it completely leaves the node. The start time is related to the end time of the previous node. The time window The relationship shown in formula (7) to formula (9) is satisfied.

[0054] (3) Turning through a node No. k Before reaching the turning point, the AGV needs to change from the straight-line speed to Slow down to turning speed , and then pass the turning point at the turning speed. The turning stroke starts from the time when the front edge of the AGV enters the left edge (or upper edge) of the grid until the rear end of the AGV leaves the upper and lower edges (or left / right edges) of the grid. The start time is related to the end time of the previous node. The time window The relationship shown in formulas (10) to (12) is satisfied.

[0055] (4) Stop the node No. k The AGV is located in e grid points, based on the driving speed With deceleration Decelerate uniformly to stop, start time and end time are related, time window Satisfies the relationship shown in formula (13) to formula (15). Among them, For the k AGVs.

[0056] The time window algorithm performs well under ideal conditions, and effectively resolves potential conflicts by calculating the AGV running time. However, the complexity of the actual laboratory environment makes it subject to interference from many aspects, such as outbound time, AGV failure, load and power differences, and the AGV's inability to achieve uniform motion and human interference. These factors make it difficult to accurately predict the motion state of the AGV, causing system errors. Therefore, it is necessary to introduce real-time monitoring and conflict detection mechanisms to detect these errors in a timely and accurate manner, formulate treatment measures, and ensure that the errors of the time window are within a controllable range.

[0057] In one embodiment, step 106 includes: judging whether there are repeated path nodes according to the path nodes of each AGV, if not, obtaining a conflict-free path for each AGV; if yes, judging whether there are overlaps in the time windows of the path nodes, if not, obtaining a conflict-free path for each AGV; if there are overlaps, taking path replanning measurements to replan the path for the AGV; detecting whether there is a conflict, if there is no conflict, obtaining a conflict-free path for each AGV; if there is a conflict, adopting a shifting time window strategy to resolve the conflict, and then calculating the time window of the AGV passing through each node from the current moment, continuing to perform conflict detection and conflict resolution until all path nodes are traversed to obtain a conflict-free path for each AGV.

[0058] Specifically, when the planned path conflicts with the time window of the existing path, the conflict can be avoided by shifting, stretching, and compressing the time window. Time window shifting means avoiding conflicts by making AGVs with low task priority stop and wait. This method is simple to implement and works well for situations where there is less overlap in time windows, but it is prone to long waits and congestion. Time window stretching and compression is to avoid collisions between AGVs by accelerating and decelerating AGVs. This method is very suitable for solving the problem of AGV conflicts in the same direction, but it is difficult to resolve the problem of conflicts in opposite directions. Due to the shortcomings of the current AGV speed control technology, this method often causes AGV collisions due to inaccurate speed correction. At the same time, unstable AGV operation and complex environment will increase the error of time calculation.

[0059] Common obstacle avoidance strategies mainly include parking and waiting and path replanning. Parking and waiting means avoiding collisions by delaying the time to reach the conflicting node. However, it will increase the number of starts and stops of the AGV and the total energy consumption. Path replanning is to compare the priorities of the AGVs so that the AGVs with higher priorities continue to drive along the original path, while the AGVs with lower priorities replan their paths. When planning the path, the path weight of the conflicting section is increased to avoid conflicts and prevent collisions, but it will reduce the working efficiency of the AGVS system.

[0060] It is difficult for a single strategy to obtain the global optimal solution in complex operating scenarios. This application proposes a conflict resolution strategy that combines a time-shifting window algorithm with a replanning strategy to solve the conflict problem. The time-shifting window method, that is, the stop-and-wait strategy, is used to resolve the path conflict under normal driving conditions. The path replanning method is used to resolve the path conflict under abnormal driving conditions.

[0061] The specific process of the hybrid conflict resolution strategy of the shifting time window algorithm and replanning is as follows: Fig.10 shown.

[0062] In one embodiment, the specific steps of detecting whether there is a conflict include: performing offline conflict detection based on the real-time sensing of surrounding information by the AGV's own sensors; performing online conflict detection based on the actual position of the AGV and the path node time window and the conflict detection equation; wherein the conflict detection equation includes: an opposite conflict detection equation, an encounter conflict detection equation, a surpassing conflict detection equation and a stop conflict detection equation; the opposite conflict detection equation is: (16) in, Indicates a AGV passes through the path node time; Respectively represent b AGV passes through the path node q time; Indicates that when the path node and path nodes q When overlapped, a Car and b The safe interval time for two AGVs to pass through the path node; Indicates a AGV path nodes Location; Indicates m AGV path nodes The location of the previous path node; Indicates m AGV path nodes The location of the next path node; Indicates b AGV path nodes q Location; Indicates b AGV path nodes q The location of the previous path node; Indicates b AGV path nodes qThe location of the next path node.

[0063] The encounter conflict detection equation is: (17) The transcendental conflict detection equation is: (18) in, Indicates a The speed of the AGV; The stop conflict detection equation is: (19) in, Indicates b The speed of an AGV.

[0064] Specifically, since node and path resources are limited, when multiple AGVs are running, there is a high probability that they will occupy the same node at the same time, and a node can only be occupied by one AGV at a time, which will cause conflicts or deadlocks. In order to avoid conflicts between AGVs, it is necessary to perform safety checks on the initial path of each AGV and the path of each AGV during the process of traveling, find out the potential conflicts between the paths and eliminate them.

[0065] (1) Conflict Type The main types of conflicts between AGVS are: opposite conflicts, overtaking conflicts, node conflicts, and stop conflicts. They are introduced one by one below: 1) Opposing conflict like Fig.11 As shown in (a), a head-on conflict refers to two or more AGVs traveling in opposite directions on the same road. They will collide when they arrive at the same node when the time difference is less than the safety time. Head-on encounters are the most difficult type of conflicts to resolve. When such conflicts occur, they can easily cause system deadlock and traffic congestion, requiring manual handling to resolve.

[0066] 2) Encounter conflict like Fig.11 As shown in (b), the encounter conflict usually refers to the traffic conflict at the intersection. This conflict occurs when the AGVs are driving along different roads and the two AGVs drive to the intersection or approach the safe distance at almost the same time. This kind of conflict can be resolved by using the stop method and the time window method to stagger the time when the two AGVs pass through the node.

[0067] 3) Transcend conflict Fig.11(c) in the figure indicates an in-phase conflict, that is, two AGVs on the same path are traveling in the same direction. Since the rear car is faster than the front car, the rear AGV will overtake the front AGV after a long period of operation, causing the rear car to intermittently trigger safety obstacle avoidance, which leads to low stability of the AGV system.

[0068] 4) Stop the conflict like Fig.11 As shown in (d), a stop conflict refers to two AGVs traveling in the same direction or opposite directions on the same path, and one of the AGVs stops at a node, usually due to a malfunction, the AGV is out of power, or the AGV is loading or unloading goods, while the other AGV is about to pass through the node. If no measures are taken, an AGV collision will occur in the future.

[0069] (2) Conflict Detection Conflict detection is used to determine conflicts between AGVs and take corresponding obstacle avoidance strategies. Conflict detection is the first step in solving the problem of multi-AGV conflicts. Conflict detection is divided into online conflict detection and offline conflict detection. Online conflict detection predicts whether there is a conflict with other AGVs based on the location uploaded by the AGV and the time window of the path node, and has a certain degree of foresight. Offline conflict detection relies on the AGV's own sensors to sense the surrounding information in real time and determine whether to stop or accelerate or decelerate. The following is a detailed introduction to online conflict detection and offline conflict detection: 1) Online conflict detection Online conflict detection is based on the AGV's driving route and real-time location information, the nodes it passes through, and the arrival at the destination task point. It monitors and judges whether the AGV is operating normally on the pre-planned path and within the corresponding time window, and predicts whether the AGV will conflict in the future. Online conflict detection can effectively prevent conflicts between multiple AGVs. Since the occupancy time of each node and section of the system may have a certain error with the ideal time window when multiple AGVs perform actual tasks, in order to ensure the accuracy and robustness of the system, the time window must be corrected according to the operating status of the AGV, and abnormal situations must be handled in a timely manner. For the four types of conflict problems, corresponding conflict detection equations are established respectively, and the relevant parameters are set as follows: a) The conflict detection equation is shown in formula (16). If all the conditions of formula (16) are met, it can be judged that Moment a AGV and the b AGVs at the grid nodes Conflicts occurred at the place.

[0070] b) The encounter conflict detection equation is shown in formula (17). If formula (17) is satisfied, it can be judged that Moment a AGV and theb AGVs at the grid nodes Conflict occurred at the place.

[0071] c) The transcendence conflict detection equation is shown in formula (18). If formula (18) is satisfied, it can be determined that Moment a AGV and the b AGVs at the grid nodes Beyond conflict occurs.

[0072] d) Stop conflict detection equation is shown in formula (19). If formula (19) is satisfied, it can be judged that Moment a AGV and the b AGVs at the grid nodes Stop conflict.

[0073] 2) Offline conflict detection AGVs are usually equipped with various sensors, such as laser radar, infrared sensors, ultrasonic sensors, etc., to sense the surrounding environment and obstacles. Through real-time detection and monitoring, AGVs can promptly detect obstacles ahead and take corresponding measures, such as slowing down, avoiding or stopping. Offline conflict detection is used when there are moving obstacles or people who suddenly appear on the road, and is the last line of defense to avoid collisions.

[0074] It should be understood that although Figure 1 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 1 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0075] In a verification example, in order to verify the effectiveness of the improved JPS algorithm based on time window in multi-AGV conflict-free path planning, this experiment uses three AGVs as simulation examples to generate a 30*31 grid map. Fig.12 As shown, the black ones are obstacles, the white ones are the AGV drivable areas, the grids are of equal size, and the positions are express, Indicates the row where the grid is located. Indicates the column where the grid is located. The grid node spacing is 1m, and each AGV model is abbreviated to a 0.5*0.5m square. Its straight-line moving speed is 1m / s, the turning speed is 0.5m / s, and the acceleration and deceleration is 1m / s. If two adjacent grid points are the starting point (deceleration point) and the turning point, they will directly accelerate or decelerate to the turning speed to start (stop). The safety time for AGVS to pass the same grid point is set to 2s.

[0076] The simulation experiment is used to verify the effectiveness of resolving conflicts when three AGVs work simultaneously. The number of each AGV, the information of the transportation task, the starting point coordinates and the end point coordinates are shown in Table 2.

[0077] Table 2 AGVS information table

[0078] According to the grid map obstacles and the coordinates of the starting and ending points of each AGV, the improved JPS algorithm is used to calculate the initial paths and initial time windows of the three AGVs. The initial paths are as follows: Fig.13 The initial path nodes and time windows of AGV1 are shown in Table 3, the initial path nodes and time windows of AGV2 are shown in Table 4, and the initial path nodes and time windows of AGV3 are shown in Table 5.

[0079] Table 3 AGV1 initial path nodes and time windows

[0080] Table 4 AGV2 initial path nodes and time windows

[0081] Table 5 AGV3 initial path nodes and time windows

[0082] The initial paths and time window nodes of the three AGVs are conflicted. AGV1 and AGV2 have a node conflict at grid (16,13) at 31 seconds, and AGV2 and AGV3 have a node conflict at grid (25,20) at 11.5 seconds. Since the priority is AGV1>AGV2>AGV3, the path of AGV2 is replanned. Fig.14 The path diagram after replanning, the path nodes and time windows are shown in Table 6.

[0083] Table 6 Path nodes and time windows of AGV2 after the first path replanning

[0084] The path points of GV1 and AGV2 intersect at (10,18). The time point when AGV1 passes through (10,18) is [18.8,19.8], while the time window when AGV2 passes through point (10,18) is [26.5,27.5]. The time difference is 7.8s, which is the safe passing time and no conflict will occur. After conflict detection, it is determined that AGV2 and AGV3 will conflict at grid (25,26), and the path of AGV3 needs to be replanned. Fig.15 The path diagram after replanning, the path nodes and time windows are shown in Table 7.

[0085] Table 7 Path nodes and time windows of AGV3 after the second path replanning

[0086] Depend on Fig.15 It can be seen that the path of AGV3 intersects with the paths of AGV2 and AGV1 at grids (15,21) and (16,8) respectively. From the time window of the previous node (15,26) of (15,21) and formulas (7) to (9), it can be calculated that the time window of AGV3 passing through grid point (15,21) is [17.4,18.4], while the time window of AGV2 passing through point (15,21) is [19.5,20.5], the time window of AGV1 passing through point (16,8) is [32.61,34.21], and the time window of AGV3 passing through (16,8) is [30.4,32.0]. The time intervals are all longer than the safety time, so no conflict will occur, and conflict-free path planning is completed.

[0087] In one embodiment, Fig.16 As shown, a multi-AGV path planning device based on a time window is provided, comprising: a working environment map construction module, an initial global optimal path planning module, a time window model construction module and a local path planning module based on a time window, wherein: The working environment map construction module is used to construct the working environment map of the AGV using a grid method.

[0088] The initial global optimal path planning module is used to plan the path of each AGV using the improved Floyd two-way search JPS algorithm according to the work environment map, so as to obtain the initial global optimal path of each AGV when performing its respective delivery tasks; the improved Floyd two-way search JPS algorithm is used to plan the path of each AGV using the JPS algorithm that introduces a two-way search mechanism, and then the Floyd algorithm is used to optimize the path planning results, so as to obtain the initial global optimal path of each AGV when performing its respective delivery tasks.

[0089] The time window model building module is used to build the time window model of AGV.

[0090] The local path planning module based on time windows is used to perform local path planning based on the time window model and the initial global optimal path of all AGVs when performing their respective delivery tasks, using a conflict resolution strategy that combines a shifting time window algorithm with a replanning strategy to obtain a conflict-free path for each AGV; the conflict resolution strategy refers to using a shifting time window method to resolve path conflicts under normal driving conditions of the AGV, and using a path replanning method to resolve path conflicts under abnormal driving conditions of the AGV.

[0091] In one of the embodiments, the initial global optimal path planning module is also used to perform path planning for each AGV using the JPS algorithm to obtain an initial global optimal path; the starting node and the target node of the initial global optimal path of each AGV are used as the forward search starting node and the reverse search starting node respectively; for each AGV, the JPS algorithm is used to search from the forward search starting node and the reverse search starting node at the same time, and when the same extension node is searched in two directions, the initial global path of each AGV is obtained; the initial global path of each AGV is optimized using the Floyd algorithm to obtain the initial global optimal path of each AGV when performing its respective delivery tasks.

[0092] In one of the embodiments, the initial global optimal path planning module is also used to use the forward search start node as the forward parent node, the reverse search start node as the reverse parent node, and add the forward parent node and the reverse parent node to the forward search node list and the reverse search node list respectively; the forward parent node and the reverse parent node use each other as the target node to perform hop search respectively; add the extension node of the forward parent node to the forward search node list, and add the extension node of the reverse parent node to the reverse search node list; calculate the cost valuation of the extension node; determine whether the extension node with the smallest cost valuation obtained by the forward and reverse searches is the same node; if not, use the extension node obtained by the forward search as the new parent node of the forward search and put it into the forward search node list, use the extension node with the smallest cost valuation obtained by the reverse search as the new parent node of the reverse search and put it into the reverse search node list, and continue the hop search; if so, the initial global path of the AGV is obtained.

[0093] In one embodiment, the initial global optimal path planning module is further used to calculate the cost estimate of the expansion node using formula (1) to formula (2).

[0094] In one embodiment, the time window model building module is also used to k AGVs enter and leave the grid iThe moment is defined as the time window occupied by the grid point; the time windows of the grid points passed by all AGVs within a period of time are calculated to obtain the time window model of each AGV; the set of time windows of all AGVs on all road sections is shown in formula (3).

[0095] In one embodiment, the time window of all grid points passed by the AGV within a period of time is calculated, including: dividing the grid points passed by the AGV during operation into start nodes, stop nodes, turning nodes, and straight-line motion passing nodes; when the grid point is a start node, the first k The AGV is in the The relationship satisfied by the time window of each grid point is shown in formula (4) to formula (6); when the grid point moves in a straight line through the node, k The AGV is in the i The relationship satisfied by the time window of each grid point is shown in formula (7) to formula (9); when the grid point is a turning node, k The AGV is in the i The relationship satisfied by the time window of each grid point is shown in formula (10) to formula (12); when the grid point is a stop node, k The AGV is in the e The relationship satisfied by the time window of each grid point is shown in formulas (13) to (15).

[0096] In one of the embodiments, the local path planning module based on the time window is also used to determine whether there are repeated path nodes according to the path nodes of each AGV. If not, a conflict-free path for each AGV is obtained; if so, it is determined whether the path node time windows overlap. If not, a conflict-free path for each AGV is obtained; if there is overlap, path replanning measurements are taken to replan the path for the AGV; detect whether there is a conflict. If there is no conflict, a conflict-free path for each AGV is obtained; if there is a conflict, a translation time window strategy is used to resolve the conflict, and then the time window of the AGV passing through each node from the current moment is calculated, and conflict detection and conflict resolution are continued until all path nodes are traversed to obtain a conflict-free path for each AGV.

[0097] In one embodiment, the local path planning module based on the time window is also used to perform offline conflict detection according to the real-time sensing of surrounding information by the AGV's own sensors; and to perform online conflict detection according to the actual position of the AGV and the path node time window and the conflict detection equation; wherein the conflict detection equation includes: an opposite conflict detection equation, an encounter conflict detection equation, an overtaking conflict detection equation and a stop conflict detection equation; the opposite conflict detection equation is shown in formula (16), the encounter conflict detection equation is shown in formula (17), the overtaking conflict detection equation is shown in formula (18), and the stop conflict detection equation is shown in formula (19).

[0098] For the specific definition of the local path planning module device based on time windows, please refer to the definition of the local path planning module method based on time windows above, which will not be repeated here. Each module in the above-mentioned local path planning module device based on time windows can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0099] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Fig.17 As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a local path planning module method based on a time window is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a button, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0100] Those skilled in the art will understand that Fig.17 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0101] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps in the above method embodiment when executing the computer program.

[0102] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0103] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A multi-AGV path planning method based on time window, characterized in that: The method comprises: Use the grid method to construct the AGV's working environment map; According to the work environment map, the improved Floyd two-way search JPS algorithm is used to plan the path of each AGV, so as to obtain the initial global optimal path of each AGV when performing its respective delivery task; the improved Floyd two-way search JPS algorithm is used to plan the path of each AGV using the JPS algorithm that introduces a two-way search mechanism, and then the Floyd algorithm is used to optimize the path planning result, so as to obtain the initial global optimal path of each AGV when performing its respective delivery task; Construct a time window model for AGV; According to the time window model and the initial global optimal path of all AGVs when performing their respective delivery tasks, a conflict resolution strategy that combines a shifting time window algorithm with a replanning strategy is used to perform local path planning, so as to obtain a conflict-free path for each AGV; the conflict resolution strategy refers to using a shifting time window method to resolve path conflicts under normal driving conditions of the AGV, and using a path replanning method to resolve path conflicts under abnormal driving conditions of the AGV.

2. The multi-AGV path planning method based on time window according to claim 1 is characterized in that: According to the work environment map, the improved Floyd bidirectional search JPS algorithm is used to plan the path of each AGV to obtain the initial global optimal path of each AGV when performing its respective delivery task, including: Use the JPS algorithm to plan the path for each AGV and obtain the initial global optimal path; The starting node and target node of the initial global optimal path of each AGV are used as the forward search starting node and the reverse search starting node respectively; For each AGV, the JPS algorithm is used to search the starting node from the forward search direction and the starting node from the reverse search direction at the same time. When the same expansion node is found in the two directions, the initial global path of each AGV is obtained. The initial global path of each AGV is optimized using the Floyd algorithm to obtain the initial global optimal path of each AGV when performing its respective delivery tasks.

3. The multi-AGV path planning method based on time window according to claim 2 is characterized in that: For each AGV, the JPS algorithm is used to search the starting node from the forward search direction and the starting node from the reverse search direction. When the same extended node is found in the two directions, the initial global path of each AGV is obtained, including: Using the forward search start node as the forward parent node, using the reverse search start node as the reverse parent node, and adding the forward parent node and the reverse parent node to the forward search node list and the reverse search node list respectively; The forward parent node and the reverse parent node use each other as the target node and perform hop search respectively; Add the extended node of the forward parent node to the forward search node list, and add the extended node of the reverse parent node to the reverse search node list, Calculate the cost estimate of expanding the node; Determine whether the expansion node with the smallest cost estimate found in the forward and reverse searches is the same node; If not, the extended node found in the forward search is used as the new parent node of the forward search and is put into the forward search node list. The extended node with the smallest cost estimate found in the reverse search is used as the new parent node of the reverse search and is put into the reverse search node list to continue the hop search. If yes, the initial global path of the AGV is obtained.

4. The multi-AGV path planning method based on time window according to claim 3 is characterized in that: The cost estimate for calculating the expansion node is: ； ； in, is the evaluation function, is the environmental barrier rate, h ( n ) is the heuristic function, To add the estimated waiting time, is the actual cost function, It is the starting point of AGV. is the end point of AGV. The number of obstacles that form a rectangular area for the start and end points.

5. The multi-AGV path planning method based on time window according to claim 1 is characterized in that: Construct the time window model of AGV, including: The first k AGVs enter and leave the grid i The moment is defined as the time window occupied by the grid point; Calculate the time windows of the grid points passed by all AGVs within a period of time to obtain the time window model of each AGV; The time window set of all AGVs on all road segments is: ； In the formula, Indicates the total number of path nodes; Indicates the total number of AGVs, For the k The time window set of all path points of the AGV, Representation Node Relative to the k The time window of an AGV, where i =1,2,3,……, n , k =1,2,3,……, m .

6. The multi-AGV path planning method based on time window according to claim 1 is characterized in that: Calculate the time window of all grid points passed by AGVs within a period of time, including: The grid points that the AGV passes through during operation are divided into starting nodes, stopping nodes, turning nodes, and linear motion passing nodes; When the grid point is the start node, k AGV in the first s The relationship satisfied by the time window of each grid point is: ； ； ； in, Respectively represent k AGV in the s The start and end time of the time window of each grid point, represents the Euclidean distance from the AGV starting point to the task starting point, Indicates the starting point of AGV, represents the starting point of the AGV mission, L represents the length of the vehicle body, l represents the grid length, Indicates the driving speed of the AGV; When the grid point moves in a straight line through the node, k AGV in the first i The relationship satisfied by the time window of each grid point is: ； ； ； in, Respectively represent k AGV in the i The start and end time of the time window of each grid point, Grid points To grid point The Euclidean distance of Indicates that AGV is in the i The coordinate position of the grid points, Indicates that AGV is in the i -1 grid point coordinate position; When the grid point is a turning node, k AGV in the first i The relationship satisfied by the time window of each grid point is: ； ； ； in, a represents acceleration, Indicates the steering speed, Indicates the k AGV in the i -1 grid point time window end time; When the grid point is a stop node, k The relationship satisfied by the time window of an AGV at the e-th grid point is: ； ； ； in, Respectively represent k AGV in the e The start and end time of the time window of each grid point, is the Euclidean distance from the stop point to the previous grid point, Indicates that AGV is in the e The coordinate position of the grid points, Indicates that AGV is in the e -1 grid point coordinate position.

7. The multi-AGV path planning method based on time window according to claim 1 is characterized in that: According to the time window model and the initial global optimal path of all AGVs when performing their respective delivery tasks, a conflict resolution strategy in which the translation time window algorithm is paralleled with the replanning strategy is used for local path planning to obtain a conflict-free path for each AGV, including: According to the path nodes of each AGV, determine whether there are repeated path nodes. If not, obtain a conflict-free path for each AGV; If there is, determine whether the path node time windows overlap. If there is no overlap, a conflict-free path for each AGV is obtained; if there is an overlap, take path replanning measurement to replan the path for the AGV; detect whether there is a conflict. If there is no conflict, a conflict-free path for each AGV is obtained; if there is a conflict, use the translation time window strategy to resolve the conflict, and then calculate the time window of the AGV passing through each node from the current moment, continue to perform conflict detection and conflict resolution, until all path nodes are traversed, and a conflict-free path for each AGV is obtained.

8. The multi-AGV path planning method based on time window according to claim 7 is characterized in that: The specific steps to detect whether there is a conflict include: Offline conflict detection is performed based on the real-time sensing of surrounding information by the AGV’s own sensors; According to the actual position of the AGV, the time window of the path node and the conflict detection equation, online conflict detection is performed; wherein the conflict detection equation includes: the opposite conflict detection equation, the encounter conflict detection equation, the overtaking conflict detection equation and the stop conflict detection equation; The opposite conflict detection equation is: ； in, Indicates a AGV passes through the path node time; Respectively represent b AGV passes through the path node q time; Indicates that when the path node and path nodes q When overlapped, a Car and b The safe interval time for two AGVs to pass through the path node; Indicates a AGV path nodes Location; Indicates m AGV path nodes The location of the previous path node; Indicates m AGV path nodes The location of the next path node; Indicates b AGV path nodes q Location; Indicates b AGV path nodes q The location of the previous path node; Indicates b AGV path nodes q The location of the next path node; The encounter conflict detection equation is: ； The transcendental conflict detection equation is: ； in, Indicates a The speed of the AGV; The stop conflict detection equation is: ； in, Indicates b The speed of an AGV.

9. A multi-AGV path planning device based on time window, characterized in that: The device comprises: A working environment map construction module is used to construct the working environment map of the AGV using a grid method; The initial global optimal path planning module is used to plan the path of each AGV using the improved Floyd two-way search JPS algorithm according to the work environment map, so as to obtain the initial global optimal path of each AGV when performing its respective delivery tasks; the improved Floyd two-way search JPS algorithm is used to plan the path of each AGV using the JPS algorithm that introduces a two-way search mechanism, and then the Floyd algorithm is used to optimize the path planning results, so as to obtain the initial global optimal path of each AGV when performing its respective delivery tasks; A time window model building module is used to build a time window model for AGV; The local path planning module based on time windows is used to perform local path planning according to the time window model and the initial global optimal path of all AGVs when performing their respective delivery tasks, using a conflict resolution strategy that combines a shifting time window algorithm with a replanning strategy in parallel to obtain a conflict-free path for each AGV; the conflict resolution strategy refers to using a shifting time window method to resolve path conflicts under normal driving conditions of the AGV, and using a path replanning method to resolve path conflicts under abnormal driving conditions of the AGV.

10. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the multi-AGV path planning method based on time windows described in any one of claims 1 to 8 is implemented.

Images