Intelligent warehouse multi-agv scheduling method and system

Abstract

The application discloses a kind of intelligent warehousing multi-AGV scheduling method and system, it is related to warehousing scheduling technical field, task path is solved using graph optimization algorithm joint multi-objective scheduling trade-off factor model, output task allocation and path planning scheme, task allocation and path planning scheme are loaded to time axis, path is time sequence simulation, potential congestion point or resource conflict area in path is identified, and according to identification result, part of AGV path is optimized, time traffic pressure index and actual AGV operating state are used as input, scheduling verification is carried out through multiple rounds of offline simulation, and multi-AGV scheduling holographic dataset is integrated, and task state driven multi-AGV closed-loop scheduling scheme is constructed.The scheduling method is by constructing multi-dimensional scheduling model that fuses task, path and resource information, and combining graph optimization, dynamic simulation and closed-loop control, the efficient cooperation and real-time response of multi-AGV system are realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of warehouse scheduling, in particular to an intelligent warehouse multi-AGV scheduling method and system. BACKGROUND

[0002] With the development of Industry 4.0 and smart logistics, traditional labor-intensive warehouse operations are gradually evolving towards automation and intelligence. Automatic guided vehicles (AGVs) have been widely used in modern intelligent warehouse factories due to their high efficiency, safety, and sustainable operation. Compared with single AGV systems, multi-AGV systems can significantly improve task throughput and warehouse efficiency, and are particularly suitable for practical application requirements such as multi-task concurrency and high dynamic environment.

[0003] The prior art has the following defects:

[0004] The existing scheduling method relies on static path planning and ignores dynamic resource conflicts and space-time coupling problems during task execution, which cannot effectively deal with scheduling bottlenecks caused by task-intensive and multi-vehicle coordination in complex warehouse environments. In addition, the step-by-step task allocation and path planning lack multi-dimensional fusion modeling between task states, path resources, and vehicle behaviors, resulting in insufficient global optimization capability of the scheduling strategy.

[0005] Therefore, the present application provides an intelligent warehouse multi-AGV scheduling method and system, which builds a multi-dimensional scheduling model integrating task, path, and resource information, and combines graph optimization, dynamic simulation, and closed-loop control to achieve efficient coordination and real-time response of the multi-AGV system. SUMMARY

[0006] The purpose of the present application is to provide an intelligent warehouse multi-AGV scheduling method and system to solve the problems in the background art.

[0007] To achieve the above purpose, the present application provides the following technical solution: an intelligent warehouse multi-AGV scheduling method, the scheduling method comprising the following steps:

[0008] Step S1: The acquisition end obtains the structured layout information of the warehouse factory, and initializes the task state atlas in combination with the current running state of the AGV cluster;

[0009] Step S2: The intersection and resource occupation of AGV in spatial path and time resource are marked, and a multi-AGV collaborative scheduling atlas is generated;

[0010] Step S3: Assign a comprehensive scheduling weight to each feasible path and task allocation scheme, and build a multi-objective scheduling trade-off factor model;

[0011] Step S4: The task path is solved by using a graph optimization algorithm combined with a multi-objective scheduling weighting factor model, and a task allocation and path planning scheme is output.

[0012] Step S5: The task allocation and path planning scheme is loaded to a time axis, the path is time-sequenced simulated, potential congestion points or resource conflict areas in the path are identified, and part of the AGV path is optimized according to the identification result.

[0013] Step S6: The time traffic pressure index of each path in the scheduling period is counted, and the time traffic pressure index and the actual AGV running state are taken as inputs to verify the scheduling through multiple rounds of offline simulation.

[0014] Step S7: The multi-AGV scheduling holographic data set is integrated, and a task state driven multi-AGV closed loop scheduling scheme is constructed.

[0015] Preferably, the path is time-sequenced simulated, potential congestion points or resource conflict areas in the path are identified, and part of the AGV path is optimized according to the identification result, including the following steps:

[0016] In the task allocation and path planning scheme, the time occupation interval of each path is constructed .

[0017] The potential congestion point or resource conflict area judgment expression is: , wherein represents the actual occupation time interval of the i-th AGV on the path segment , and represents the actual occupation time interval of the j-th AGV on the path segment .

[0018] All paths satisfying the judgment expression are marked as conflict sensitive segments, a time traffic pressure index is introduced to describe the average AGV occupation intensity on the path segment per unit time, all conflict sensitive segments are sorted according to the traffic pressure level, and path conflict optimization strategies are executed according to the simulation result, including peak shifting, path reselection, and soft start control.

[0019] Preferably, the time traffic pressure index of each path in the scheduling period is counted, and the time traffic pressure index and the actual AGV running state are taken as inputs to verify the scheduling through multiple rounds of offline simulation, including the following steps:

[0020] The average AGV traffic frequency and the task density are weighted to obtain the time traffic pressure index.

[0021] Introducing dynamic feedback parameters for task execution, extracting each AGV's transit time and scheduling delay feedback from actual operation data;

[0022] The time-travel pressure indicators of all paths and the actual operating status of the AGV are input into multiple rounds of offline simulation programs to simulate the operating performance under different scheduling parameters.

[0023] Preferably, a comprehensive scheduling weight is assigned to each feasible path and task allocation scheme, and a multi-objective scheduling trade-off factor model is constructed, which includes the following steps:

[0024] Based on path distance and resource conflict costs Construct a comprehensive scheduling weight function of paths and tasks to evaluate AGVs Execute the task The overall dispatching cost .

[0025] Preferably, the resource conflict cost Defined as: ,in, Indicates AGV On the path Execute tasks on The time interval, n is the number of paths, Indicates the task time interval of other AGVs on the same path, function is an indicator function that outputs 1 when the two time intervals intersect, otherwise it outputs 0. Indicates AGV On the path Execute tasks on time interval, represents the intersection of two time intervals, If the intersection of the two time intervals is not empty, that is, there is a time overlap, the path is defined as a potential conflict segment.

[0026] Preferably, the intersection and resource occupation of AGVs on spatial paths and time resources are marked to generate a multi-AGV collaborative scheduling map, including the following steps:

[0027] Map the collected transport task queue to the constructed task state graph. Transport tasks include pickup, handling, and delivery.

[0028] Decompose the transportation task process into task units ,in For the transport missions, is the i-th subtask step, and s is the number of subtask steps;

[0029] spatially extracting the paths involved in task execution , and mapping the node dimension into the spatial topology of the scheduling graph;

[0030] time-slicing the task flow according to the task time window in the time dimension to form a three-dimensional scheduling graph data body , constituting a three-dimensional mapping structure of task-path-node;

[0031] defining a time occupancy interval on each path , and introducing a resource occupancy tensor to describe the time domain competition relationship of AGV on system resources;

[0032] Based on the scheduling graph and the resource occupancy tensor, the AGV behavior data and the task path timing are fused to generate a multi-AGV collaborative scheduling graph.

[0033] Preferably, a three-dimensional scheduling graph data body is formed, which is represented as: , wherein, represents a set of spatial nodes in the three-dimensional scheduling graph, is a set of path edges, is a time layer structure for describing the time occupancy interval of each task unit on the path.

[0034] Preferably, the acquisition end obtains the structured layout information of the warehouse factory, including the spatial coordinates and capacity parameters of the storage sites, the topological structure and passing constraint conditions of the conveying paths, the resource scheduling priority and reachability determination of the transfer nodes, and the work capacity parameters of the work stations.

[0035] The dynamic running state data of the AGV cluster is obtained, including the current position coordinates, remaining power, current load weight of each AGV, and task queue state.

[0036] Preferably, the acquisition end obtains the structured layout information of the warehouse factory, and initializes the task state graph combining the current running state of the AGV cluster, including the following steps:

[0037] The structured layout information and the running state data are used to construct the task state graph, and the multiple relationships of AGV, task, path and resource are represented by a graph structure;

[0038] Each node in the task state graph represents a scheduling element, and each edge represents the transfer path or resource occupancy relationship of the task;

[0039] The task state graph is represented as: , wherein, represents a set of nodes in the graph, including AGV nodes, task nodes and resource nodes, representing a set of edges between nodes, is a weight matrix.

[0040] The application also provides an intelligent warehouse multi-AGV scheduling system, comprising a graph generation module, a scheme output module, a simulation verification module and a scheduling scheme construction module.

[0041] The graph generation module: obtains the structured layout information of the warehouse factory, initializes the task state graph in combination with the current running state of the AGV cluster, labels the intersection and resource occupation of the AGV on the spatial path and time resource, and generates a multi-AGV collaborative scheduling graph.

[0042] The scheme output module: gives a comprehensive scheduling weight to each feasible path and task allocation scheme, constructs a multi-objective scheduling trade-off factor model, solves the task path by using a graph optimization algorithm in combination with the multi-objective scheduling trade-off factor model, and outputs a task allocation and path planning scheme.

[0043] The simulation verification module: loads the task allocation and path planning scheme to a time axis, performs time sequence simulation on the path, identifies potential congestion points or resource conflict areas in the path, optimizes part of the AGV path according to the identification result, and calculates the time traffic pressure index of each path in the scheduling period.

[0044] The scheduling scheme construction module: integrates a multi-AGV scheduling holographic data set, and constructs a multi-AGV closed-loop scheduling scheme driven by a task state.

[0045] In the above technical solution, the application provides the following technical effects and advantages:

[0046] The application gives a comprehensive scheduling weight to each feasible path and task allocation scheme, constructs a multi-objective scheduling trade-off factor model, solves the task path by using a graph optimization algorithm in combination with the multi-objective scheduling trade-off factor model, outputs a task allocation and path planning scheme, loads the task allocation and path planning scheme to a time axis, performs time sequence simulation on the path, identifies potential congestion points or resource conflict areas in the path, optimizes part of the AGV path according to the identification result, calculates the time traffic pressure index of each path in the scheduling period, takes the time traffic pressure index and the actual AGV running state as inputs, performs scheduling verification through multiple rounds of offline simulation, integrates a multi-AGV scheduling holographic data set, and constructs a multi-AGV closed-loop scheduling scheme driven by a task state. The scheduling method constructs a multi-dimensional scheduling model integrating task, path and resource information, and combines graph optimization, dynamic simulation and closed-loop control to realize efficient collaboration and real-time response of the multi-AGV system, and significantly improves the scheduling accuracy and running stability. BRIEF DESCRIPTION OF DRAWINGS

[0047] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the accompanying drawings needed in the embodiments will be briefly introduced as follows. Obviously, the accompanying drawings in the following description only represent some embodiments of the present application, and all other drawings obtained by those of ordinary skill in the art based on these drawings belong to the scope of the present application.

[0048] Figure 1 The flow chart of the scheduling method of the present application.

[0049] Figure 2 The timing chart of the scheduling method of the present application.

[0050] Figure 3 The architecture diagram of the scheduling system of the present application. DETAILED DESCRIPTION

[0051] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of the present application.

[0052] Embodiment 1: Please refer to Figures 1-2 As shown in the figure, the present embodiment provides an intelligent warehouse multi-AGV scheduling method, which comprises the following steps:

[0053] Step S1: The acquisition end obtains the structured layout information of the warehouse factory, which includes storage sites, conveying paths, transfer nodes, workstations and the like. The current running state (position information, load state, power level, task queue and the like) of the AGV cluster is combined to initialize a task state atlas, which is used as the basic input for scheduling analysis.

[0054] Step S2: The current to-be-executed transportation task queue is mapped to the task state atlas. The task-path-node three-dimensional scheduling atlas is established by decomposing the task flow (such as picking, carrying and delivery), and the potential intersection and resource occupation of AGV in the spatial path and time resource are marked to generate a multi-AGV collaborative scheduling atlas.

[0055] Step S3: The path planning criteria (such as load balancing) and the resource competition control mechanism (such as avoidance priority) are introduced. The comprehensive scheduling weight is given to each feasible path and task allocation scheme, and a multi-objective scheduling trade-off factor model is constructed.

[0056] Step S4: Use a graph optimization algorithm (such as improved A*) combined with a multi-objective scheduling trade-off factor model to solve the task path, output a set of task allocation and path planning scheme, task allocation and path planning scheme record the task sequence, path node, expected execution time and resource occupation details of each AGV.

[0057] Step S5: Load the task allocation and path planning scheme to the timeline, perform time sequence simulation on the path, identify potential congestion points or resource conflict areas in the path; and optimize part of the AGV path (such as staggered driving, path reselection, slow start) according to the identification results, to minimize path conflicts.

[0058] Step S6: Statistic the path load characteristics of each path in the scheduling period, the path load characteristics include average AGV passing frequency and task density, construct the time passing pressure index of the path; combine the task execution dynamic feedback parameters, the task execution dynamic feedback parameters include AGV passing time, scheduling delay feedback and other data, identify risk bottleneck section, provide target area positioning for subsequent scheduling optimization, take the time passing pressure index and the actual AGV running state as input, verify the scheduling stability and response efficiency through multiple rounds of offline simulation, and calibrate the scheduling parameters.

[0059] Step S7: Integrate the multi-AGV scheduling holographic dataset, the multi-AGV scheduling holographic dataset includes task state atlas, multi-AGV collaborative scheduling atlas, AGV path optimization result and scheduling simulation verification result, construct the multi-AGV closed-loop scheduling scheme driven by task state, realize the cycle control mode of task loading-path generation-state feedback-path reconstruction, so as to realize the efficient, robust and adaptive collaborative scheduling of the multi-AGV system in complex dynamic warehouse scenarios.

[0060] The present application assigns a comprehensive scheduling weight to each feasible path and task allocation scheme, constructs a multi-objective scheduling trade-off factor model, uses a graph optimization algorithm combined with a multi-objective scheduling trade-off factor model to solve the task path, outputs a task allocation and path planning scheme, loads the task allocation and path planning scheme to the timeline, performs time sequence simulation on the path, identifies potential congestion points or resource conflict areas in the path, and optimizes part of the AGV path according to the identification results, statistics the time passing pressure index of each path in the scheduling period, takes the time passing pressure index and the actual AGV running state as input, performs scheduling verification through multiple rounds of offline simulation, integrates the multi-AGV scheduling holographic dataset, and constructs the multi-AGV closed-loop scheduling scheme driven by task state. The scheduling method realizes efficient collaboration and real-time response of the multi-AGV system by constructing a multi-dimensional scheduling model integrating task, path and resource information, and combining graph optimization, dynamic simulation and closed-loop control, significantly improves the scheduling accuracy and operation stability.

[0061] Please refer toFigure 3 As shown, the embodiment provides an intelligent warehouse multi-AGV scheduling system, which comprises a graph generation module, a scheme output module, a simulation verification module and a scheduling scheme construction module.

[0062] The graph generation module: acquires the structured layout information of the warehouse factory, initializes the task state graph in combination with the current running state of the AGV cluster, labels the intersection and resource occupation of AGV on the spatial path and time resource, generates a multi-AGV collaborative scheduling graph, and sends the task state graph and the multi-AGV collaborative scheduling graph to the scheduling scheme construction module. The task state graph is sent to the scheme output module.

[0063] The scheme output module: gives a comprehensive scheduling weight to each feasible path and task allocation scheme, constructs a multi-objective scheduling trade-off factor model, solves the task path by using a graph optimization algorithm combined with the multi-objective scheduling trade-off factor model, outputs the task allocation and path planning scheme, and sends the task allocation and path planning scheme to the simulation verification module.

[0064] The simulation verification module: loads the task allocation and path planning scheme to the time axis, performs time sequence simulation on the path, identifies potential congestion points or resource conflict areas in the path, optimizes part of the AGV path according to the identification result, and calculates the time traffic pressure index of each path in the scheduling period. The time traffic pressure index and the actual AGV running state are used as inputs, and the scheduling is verified through multiple rounds of offline simulation. The AGV path optimization result and the scheduling simulation verification result are sent to the scheduling scheme construction module.

[0065] The scheduling scheme construction module: integrates the multi-AGV scheduling holographic data set, and constructs a task state driven multi-AGV closed loop scheduling scheme.

[0066] Embodiment 2: Step S1: The acquisition end acquires the structured layout information of the warehouse factory, which includes storage sites, conveying paths, transfer nodes and workstations, etc. In combination with the current running state of the AGV cluster (position information, load state, power level, task queue, etc.), the task state graph is initialized as the basic input for scheduling analysis.

[0067] The acquisition end first acquires the structured layout information of the warehouse factory through the industrial Internet of Things device, the warehouse management system (WMS) and the manufacturing execution system (MES). The information covers multiple dimensions, including the spatial coordinates and capacity parameters of storage locations, the topological structure and passing constraint conditions of transport routes, the resource scheduling priority and reachability determination of transfer nodes, and the work capacity parameters (such as processing tact, loading and unloading efficiency, and work time window) of workstations. These information collectively constitute the basis of digital expression of the warehouse physical space, and are the core support for subsequent AGV scheduling modeling.

[0068] On this basis, the system acquires the dynamic running state data of the AGV cluster, specifically including the current position coordinates, remaining power and current load weight of each AGV. To further support intelligent scheduling analysis, a task state mapping graph (Task-State-Mapping-Graph) needs to be constructed to represent the multiple relationships of AGV, task, path and resource in a graph structure. Each node in the graph represents a scheduling element (such as an AGV or a task site), and each edge represents the transfer path or resource occupation relationship of the task. The edge weight can be weighted according to the time cost, energy consumption, passing priority, etc. This graph can be formalized as: wherein, represents the node set in the graph, including AGV nodes, task nodes and resource nodes, represents the edge set between nodes, is a weight matrix reflecting the scheduling cost function (such as path length, task priority, etc.). The element in the weight matrix is defined as: wherein, represents the scheduling cost of node i to j; is the path distance from node i to j, is the total path length.

[0069] The establishment of the above task state mapping graph not only provides the scheduling system with comprehensive cognitive ability of the existing resource and task state, but also provides a unified data basis and modeling context for subsequent path optimization, task allocation and simulation prediction. It is the starting point and core input of the entire multi-AGV scheduling method.

[0070] Step S2: Map the current transportation task queue to be executed to the task state mapping graph, establish a task-path-node three-dimensional scheduling graph by decomposing the task flow (such as picking, carrying and delivering), and mark the potential intersection and resource occupation of AGV in the space path and time resource to generate a multi-AGV collaborative scheduling graph.

[0071] The collected transport task queue is mapped to the constructed task state map, and detailed modeling is carried out based on the operational process of the task life cycle. Transport tasks usually include several operational units, such as pickup, transport, and drop-off, and their processes have obvious spatial continuity and time constraints. Therefore, in order to more accurately depict the interactive relationship between tasks, paths, and resources, the transport task process needs to be decomposed into task units. ,in is the kth transportation task, is the i-th subtask step, and s is the number of subtask steps.

[0072] In space, the system extracts the paths involved in task execution , and mapped to the spatial topology of the scheduling graph in the node dimension; in the time dimension, the task process is time-sliced ​​according to the system beat or task time window to form a three-dimensional scheduling graph data body , recorded as: ,in, Represents the set of spatial nodes in a three-dimensional scheduling graph, is the set of path edges, The time-layer structure describes the time interval of each task unit on the path, forming a three-dimensional mapping structure of "task-path-node". To achieve collaborative scheduling, the system needs to further identify conflict areas such as path overlap, time intersection, and resource competition between AGVs on the three-dimensional scheduling diagram.

[0073] On each path, the time interval is defined as: ,in, For the task On the path The actual occupied time interval on is the time of entering the path, is the departure time. If there are two tasks , which are on the same path Satisfaction on: , For the task On the path The actual occupied time interval on represents the intersection of two time intervals, If the intersection of these two time intervals is not empty, that is, there is a time overlap, then the path is defined as a potential conflict segment. It is necessary to introduce a coordination mechanism to lock resources or stagger tasks. To describe the time domain competition relationship between AGVs for system resources, the resource occupancy tensor is introduced: ,in, A three-dimensional occupancy tensor is provided, which can be used to quickly calculate the AGV density and conflict frequency on any road segment at any time, as a basic data structure for subsequent scheduling optimization.

[0074] It is assumed that there is a path segment R12 in the intelligent warehouse system, which is a shared channel that must be passed through by a plurality of transportation tasks. The system needs to determine whether there is a time overlap of a plurality of tasks on the path segment, so as to trigger resource locking or task scheduling adjustment.

[0075] It is assumed that there are two tasks:

[0076] executed by AGV , which is planned to pass through the path ; executed by AGV , which also needs to pass through the path . The scheduling system records the following time occupancy:

[0077] seconds;

[0078] seconds;

[0079] According to the determination rule: , it can be seen that the intersection of the two time periods is [108, 110], that is, there is a 2-second time overlap, so: the path segment is determined as a potential conflict segment;

[0080] The system needs to introduce a coordination mechanism, such as setting AGV to enter 3 seconds later, that is, staggered passing, or setting this segment as a “mutually exclusive segment” to allow only one AGV to pass through at the same time; or enabling a priority strategy (such as passing through an urgent order first).

[0081] Finally, the system generates a multi-AGV cooperative scheduling graph (Multi_AGV_Cooperative-Scheduling-Graph) based on the scheduling graph and resource occupancy tensor, fuses AGV behavior data and task path timing, and generates a multi-AGV cooperative scheduling graph (Multi_AGV_Cooperative-Scheduling-Graph) for overall scheduling strategy design and conflict control mechanism derivation. The graph not only retains the task structure and dynamic state of the resource, but also embeds cooperative conflict labeling, which is an indispensable intermediate result for subsequent scheduling trade-off modeling and path optimization.

[0082] Step S3: Introduce path planning criteria (such as load balancing) and resource competition control mechanisms (such as avoidance priority), assign a comprehensive scheduling weight to each feasible path and task allocation scheme, and construct a multi-objective scheduling trade-off factor model.

[0083] To improve the adaptability of the scheduling system to complex task environments and the overall resource utilization efficiency, a scheduling weight factor model with multi-objective optimization capability needs to be built. First, path planning criteria such as load balancing, shortest path, and deadline-aware strategies are introduced to guide the preliminary direction of task allocation. At the same time, the system considers resource competition control mechanisms, including yielding priority, mutual exclusion of segments, and energy-aware constraints, to avoid conflicts or resource overlaps on the path of AGVs.

[0084] To quantify and integrate the above scheduling objectives, a comprehensive scheduling weight function of path and task is constructed based on path distance and resource conflict cost to evaluate the comprehensive scheduling cost of AGVs performing tasks. The weight function is defined as follows: wherein, is the path distance required by AGV to perform task , is the resource conflict cost of AGV performing task , is the scheduling weight factor, and the resource conflict cost can be further defined as: wherein, is the time interval of AGV performing task on path , n is the number of paths, is the time interval of other AGVs performing tasks on the same path, the function outputs 1 when there is an intersection between the two time intervals, otherwise it is 0, is the time interval of AGV performing task on path , is the intersection of the two time intervals,

[0085] The system periodically calculates the path conflict frequency​​ with path average delay , to dynamically adjust : wherein, represents the path conflict ratio in the current scheduling period; this setting improves weight, enhancing the scheduling tendency of conflict avoidance.

[0086] Through the above weight function, the system can establish a comparable scheduling score system for each AGV-task-path triple, supporting the scheduling strategy to achieve a controllable multi-objective trade-off between efficiency and safety. Finally, the constructed scheduling trade-off factor model (Multi-Objective-Scheduling-Trade-Off-Model) provides a clear cost measurement standard for subsequent path optimization and task assignment, which is the key foundation to achieve the global optimal scheduling scheme.

[0087] Step S4: Use a graph optimization algorithm (such as improved A*) combined with the multi-objective scheduling trade-off factor model to solve the task path, output a set of task allocation and path planning scheme, which records the task sequence, path node, expected execution time and resource occupation details of each AGV.

[0088] Based on the aforementioned multi-objective scheduling trade-off factor model, a graph optimization algorithm (such as heuristic enhanced A* algorithm, time expansion graph model, etc.) is used to globally solve the path of each transportation task to obtain a set of scheduling results that take into account path optimality and resource coordination. This process not only focuses on the traditional shortest path problem, but also includes scheduling cost, multi-vehicle interference and resource conflict into the optimization target, outputting a comprehensive task path scheme with executability and scheduling stability.

[0089] Specifically, the warehouse scheduling environment is modeled as a directed weighted graph wherein is a set of spatial nodes (including storage sites, transfer stations, delivery points, etc.), is a set of passable path edges. Each edge is assigned a scheduling value , representing the comprehensive scheduling cost that the AGV needs to bear when executing the task through edge , the calculation method is referred to the weight function in step S3.

[0090] An improved algorithm is used for path search, and the heuristic function is defined as follows:

[0091] wherein, Total cost estimate for the current search node g, Cumulative scheduling weight from the start point to the current node, Minimum remaining cost heuristic value from the current node to the target node, heuristic The minimum congestion guidance strategy can be used to dynamically guide the search away from the path hotspot area. For each AGV , the system performs path search and at the same time completes task assignment, forming its task sequence set and the corresponding path node sequence . K is the total number of tasks, and M is the number of path nodes.

[0092] Suppose there is an AGV (Automatic Guided Vehicle) A1 in the intelligent warehouse system, which needs to start from the current node S (i.e. the starting point) and go to the target node (i.e. the work point) to complete a carrying task. The path contains multiple candidate nodes, and the system uses the improved A* algorithm for path planning to obtain the optimal path with the minimum cost.

[0093] Suppose the path is as follows:

[0094] S—A—B—E

[0095] \ /

[0096] C—D

[0097] The AGV currently starts from S, and the target is to reach E. The scheduling system has the following information when calculating the current candidate node B:

[0098] g(B)=10: cumulative cost from the starting point S to B (may include path length, congestion cost, etc.); h(B)=4: minimum estimated cost from B to target E (based on shortest remaining distance + traffic heat); then the total cost: f(B)=g(B)+h(B)=10+4=14, and the system also considers node D: g(D)=11, h(D)=2, then f(D)=13, at this time the system will prefer to expand node D because it has a smaller f(n) estimate, i.e. the current total cost path is better.

[0099] Suppose the AGV is assigned a task sequence: , the system performs an improved A* search on the graph for each task, obtaining the corresponding path node sequence:

[0100] , where represents the total number of tasks, is the total number of all path nodes. Finally, the complete execution path graph of the AGV is formed.

[0101] To realize system operation simulation and scheduling conflict detection, the system also needs to record the expected execution time of each AGV corresponding to the task path for subsequent scheduling beat control and timing analysis; resource occupation details, defined as the start and end time of AGV occupation of each path to support dynamic conflict detection. Among them, the expected execution time of AGV executing the task can be calculated by path cost normalization as follows: , wherein is the average running speed of AGV , is the physical distance of the path edge . The final output of the scheduling scheme is a set of all AGV path and task matching sets: , wherein represents the full set of scheduling solutions, represents the i-th AGV, is the task sequence set assigned to AGV , is the path node sequence of AGV , represents the expected task execution time of AGV , represents the resource occupation details of AGV , N is the number of AGVs, and each tuple in the set represents the complete scheduling path and task information of an AGV. Through this path optimization process, the system can search for the path scheme with the least conflict and optimal resource utilization in the scheduling graph, ensuring the coordinated operation and dynamic response of each AGV in complex task scenarios, and providing high-quality input basis for subsequent time domain simulation and traffic pressure evaluation.

[0102] Step S5: Load the task allocation and path planning scheme to the time axis, perform timing simulation on the path, identify potential congestion points or resource conflict areas in the path; and optimize part of the AGV path according to the identification results (such as staggered driving, path reselection, and slow start), to realize path conflict minimization.

[0103] Embed the task allocation and path planning scheme generated in the previous step into the time dimension to form a dynamic execution sequence of multi-AGV scheduling, and further build a scheduling timing simulation environment. This process takes path node sequence, task execution time, and resource occupation details as input, maps the path segmentation, beat information, and resource request of each AGV to a unified time axis, establishes a task allocation and path planning scheme (Path-Time-Graph) for simulation analysis, identifies potential congestion points and resource conflict areas, and realizes dynamic optimization of the scheduling scheme.

[0104] Firstly, the system constructs the time occupancy interval of each path in the task allocation and path planning scheme, defined as follows: wherein, is the occupancy time interval of AGV on path , is the entering time of AGV on the path, is the leaving time, both of which are determined by AGV speed, path length and task beat parameters. Subsequently, the system performs time interval overlap determination on all paths to identify whether the paths of any two AGVs have spatial and temporal overlap conflicts. The judgment criterion is: wherein, represents the actual occupancy time interval of the i-th AGV on path segment , represents the actual occupancy time interval of the j-th AGV on path segment , represents that the intersection is not empty, i.e., there is overlap in the time interval, i.e., the two AGVs have overlapping passing time intervals on the same path, indicating a potential conflict. The system marks all paths that meet the above conditions as Conflict-Sensitive-Segments and further optimizes them. To quantify the congestion degree of each path, the system introduces a time passing pressure index , which is used to describe the average occupancy intensity of AGVs on the path segment per unit time, and the calculation formula is: wherein, S is the total simulation period time, N is the number of AGVs, represents the occupancy time length of AGV on path

[0105] . This index can be used to sort the passing pressure levels of all conflict-sensitive segments, thereby locating the bottleneck segments and high-risk conflict points in the system. For the identified conflict paths and high-pressure passing areas, the system performs path conflict optimization strategies based on simulation results, including: Time-Shifted-Execution: setting a time offset for the conflict AGV

[0106] to delay its entering time to stagger the overlapping area;

[0107] Re-Routing: searching for alternative paths in the dispatching graph, and preferentially avoiding the current high-pressure segment;

[0108] Slow-Start-Control: reducing the running speed of AGV at the initial stage, and dynamically adjusting the arrival rhythm of the subsequent path.​For example, staggered driving can be achieved by modifying the entering time expression: where, is the adjusted entering time, is the time offset for the conflicting AGV. After optimization, the system re-evaluates the conflict in the task assignment and path planning scheme to verify whether the conflict-minimization goal is achieved, i.e., to reduce the resource overlap and path intersection frequency to below the set threshold without significantly increasing the total execution time or path length.

[0109] Suppose in an intelligent warehouse system, there are 3 AGVs (numbered AGV1, AGV2, and AGV3) and the dispatch system assigns them carrying tasks respectively. Suppose path segment R5 is one of the main paths in the warehouse and is shared by multiple tasks. The system needs to determine whether there is a traffic conflict on this path segment.

[0110] The dispatch system determines the actual traffic time of each AGV on path segment R5 as follows: The traffic time interval for AGV1 is [8, 12] seconds, the traffic time interval for AGV2 is [10, 14] seconds; the traffic time interval for AGV3 is [11, 15] seconds. According to the judgment formula: we find that: AGV1 and AGV2 overlap in time on path segment R5 from 12 to 14 seconds, so there is a conflict; AGV1 and AGV3 have no overlap in time; AGV2 and AGV3 have no overlap in time. Therefore, the system will mark path segment R5 as a conflict-sensitive segment and enter the optimization phase. For path segment R5, let the dispatch cycle be T seconds, and the traffic pressure indicator be: , which indicates that the traffic pressure on this path segment is high. The system will optimize the path according to the conflict area and pressure indicator, such as delaying the start time of AGV2 by 3 seconds to adjust it to [15, 19] to avoid overlapping with AGV1; or changing AGV2 to path segment R6 to use the alternate channel to reduce the pressure on R5; or setting AGV3 to a slow start to delay its entering time by reducing the initial speed.

[0111] ​​​​​​​​​​​​​​​​​The simulation-optimization-verification closed-loop process significantly enhances the robustness of the scheduling strategy in actual operation, improves the collaborative efficiency between AGVs, and lays a time sequence data foundation for subsequent traffic pressure modeling and feedback scheduling control.

[0112] Step S6: Statistics of the path load characteristics of each path in the scheduling period, including the average AGV traffic frequency and task density, construct the time traffic pressure index of the path; Combine the task execution dynamic feedback parameters, including AGV passing time, scheduling delay feedback and other data, identify the risk bottleneck section, provide target area positioning for subsequent scheduling optimization, take the time traffic pressure index and the actual AGV running state as input, verify the scheduling stability and response efficiency through multiple rounds of offline simulation, and calibrate the scheduling parameters.

[0113] In step S6, the system carries out path load analysis and scheduling verification around the actual operation performance of the scheduling scheme, in order to realize the stability evaluation and parameter calibration of the scheduling strategy. This step first carries out statistical analysis on the traffic behavior of each path in the scheduling period, extracts its path load characteristics, mainly including two key indicators: average AGV traffic frequency and task density distribution, so as to construct the time traffic pressure index which comprehensively represents the path running state, for identifying the high-risk bottleneck area in scheduling.

[0114] Among them, the average AGV traffic frequency represents the path The average number of times passed by AGV in the entire scheduling period can be expressed as: , wherein is the traffic frequency of path , T is the total time of the scheduling period, is the number of AGVs, is the indicator function, which takes the value of 1 when AGV passes path in the period, otherwise 0. The task density measures the task coverage degree on the path, defined as the total number of times that the unit length path is accessed by tasks in the scheduling period, and the specific calculation is as follows: , wherein is the task density of path , L is the physical length of the path, is the total number of tasks, indicates whether the task involves the path. Based on the above two characteristic quantities, the time traffic pressure index of path

[0115] is defined as: ​​

[0116] wherein, and are the weight coefficients of the passing frequency and the task density (usually determined according to the path importance) respectively, which are used to comprehensively measure the load intensity of the path, is the task density of the path , is the passing frequency of the path .

[0117] The system calculates the average passing frequency fluctuation coefficient of each path segment in multiple scheduling periods and the task density fluctuation coefficient , according to which the weight is dynamically adjusted: If the frequency fluctuation is larger, the system pays more attention to its traffic influence; otherwise, the task density dominates.

[0118] At the same time, the system introduces the task execution dynamic feedback parameter (Task-Execution-Feedback-Metrics), which extracts the following information of each AGV from the actual running data:

[0119] Passing time : represents the actual time taken by the AGV to pass through the path ;

[0120] Scheduling delay feedback : represents the delay time between the AGV executing the task and the scheduled plan.

[0121] The dynamic bottleneck pre-score of the path can be further defined , combining the static passing pressure and the dynamic feedback as follows: wherein, is the average value of the passing time of all AGVs of the path , is the average scheduling delay of all tasks associated with the path, is the feedback coefficient, used to adjust the proportion of dynamic performance in the score.

[0122] By counting the historical scheduling periods:

[0123] Calculate the average value and the standard deviation of the passing time of all AGVs of the path, and divide the standard deviation by the average value to obtain the coefficient of variation of the passing time of all AGVs of the path;

[0124] Calculate the average scheduling delay and the scheduling delay standard deviation of all tasks associated with the path, and divide the scheduling delay standard deviation by the average scheduling delay to obtain the scheduling delay coefficient of variation;

[0125] After normalization as feedback coefficient: , wherein, is the path All AGV passing time variation coefficient, is the path All AGV scheduling delay variation coefficient, when the path passing performance fluctuates greatly, the system automatically increases , pay more attention to the stability of passing; when the scheduling delay fluctuates greatly, increase , pay more attention to the scheduling efficiency.

[0126] High value path will be marked as risk bottleneck segment (Bottleneck-Segment), as the key intervention object of subsequent scheduling optimization. Finally, the system inputs all path time passing pressure indicators and AGV actual running state into the multi-round offline simulation program (Offline-Simulation-Engine) together, to simulate the running performance under different scheduling parameters. The simulation goals include:

[0127] Ensure that the system keeps the path conflict frequency and delay fluctuation within an acceptable range in multiple rounds of task execution; analyze the average task completion time, AGV turnover rate and resource utilization trend. Based on the simulation output results, the system automatically adjusts the scheduling parameters, realizes the calibration and adaptive optimization of the scheduling model parameters, and provides experiential input and strategy reinforcement basis for subsequent closed-loop scheduling control.

[0128] Step S7: integrate the multi-AGV scheduling holographic data set, which includes task state atlas, multi-AGV collaborative scheduling atlas, AGV path optimization result and scheduling simulation verification result, to construct the task state driven multi-AGV closed-loop scheduling scheme, realize the cycle control mode of task loading-path generation-state feedback-path reconstruction, and realize the efficient, robust and adaptive collaborative scheduling of the multi-AGV system in complex dynamic warehouse scenarios.

[0129] By integrating a multi-AGV holographic scheduling dataset (HSD), a closed-loop multi-AGV scheduling solution with task-state driven capabilities is constructed, implementing a closed-loop control mechanism for the entire process, from task input to scheduling output, and then to system feedback and path reconstruction. This dataset integrates core information from key aspects of the scheduling system, including: a task state map (representing task distribution and resource requirements), a multi-AGV collaborative scheduling map (representing the interaction between AGVs on paths and time), path optimization results (task allocation and path node sequence), and scheduling simulation verification results (reflecting scheduling execution efficiency and operational feedback data). Through the dynamic integration and updating of this dataset, the system can perceive state changes in real time during operation and quickly adapt scheduling strategies.

[0130] The closed-loop scheduling solution is centered around a task-state driven mechanism. Its essence lies in using task changes as triggers to link path selection and scheduling optimization, forming the following control loop:

[0131] Task loading → path generation → state feedback → path reconstruction, in which the task loading phase is composed of the task queue When a new task is triggered in the task state graph, the system obtains the spatial target, time constraint and execution priority of the current task from the task state graph, and calls the scheduling model accordingly.

[0132] The path generation phase combines the trade-off factor model with the collaborative scheduling graph to dynamically calculate the path scheduling cost: ,in, Indicates AGV At the moment The scheduling cost of the path selected for executing the task, Indicates AGV The path distance of the path selected to execute the task at time t, Indicates AGV The time traffic pressure of the path selected to execute the task at time t, It is a dynamically adjusted weighting coefficient that is updated in real time with system status feedback.

[0133] The system adjusts two weighted coefficients in real time based on the overall congestion level and traffic smoothness of the current path network: The current system average path traffic pressure: , let the ideal congestion tolerance upper limit be , the adjustment strategy is as follows: , when the overall traffic pressure is small, Higher weight (focus on efficiency); when the congestion level increases, Rise (prefer to avoid high-pressure path). The state feedback stage consists of running data and simulation predictions. Real-time running data includes AGV current node position, passing time, power change, waiting delay, etc., and simulation predictions are derived from bottleneck identification and efficiency analysis indicators output by the scheduling simulation verification module. The feedback data is input into the system to build a state offset function: wherein, represents the state offset of the i-th AGV at time t, is the ideal state trajectory, is the actual execution state. If the offset exceeds the set threshold, the path reconstruction mechanism is triggered. The scheduling cost of the current path is re-evaluated to determine whether it is still optimal. The system re-searches the path based on the new path scheduling cost and generates a new task path sequence to ensure that the scheduling system can quickly converge to an effective scheduling solution when path conflicts, task mutations, or resource bottlenecks occur. Through this closed-loop mechanism, the scheduling system can achieve:

[0134] Efficiency: dynamically optimize path and task allocation without stopping operation;

[0135] Robustness: rapid recovery ability to sudden events such as AGV failure and channel blockage;

[0136] Adaptability: continuously adjust scheduling parameters and priority strategies as the task flow changes.

[0137] Finally, the closed-loop scheduling strategy not only improves the overall throughput efficiency and resource utilization of the system, but also provides sustainable and scalable operation guarantees for multi-AGV scheduling in intelligent warehouse systems under multi-task, dynamic, and high-concurrency environments.

[0138] Example 3: The following is a typical application of the intelligent warehouse multi-AGV scheduling method in actual industrial scenarios, which helps to understand the role and value of each step in the real environment:

[0139] A nationwide e-commerce enterprise has set up a large daily consumer goods warehouse and distribution center in a logistics hub, with an area of 20,000 square meters and equipped with more than 60 AGVs (automatic guided vehicles), responsible for handling more than 100,000 order picking and distribution tasks per day during peak periods. The warehouse system adopts a dynamic scheduling mode to achieve "high concurrency, low latency, and zero congestion" in the whole process of material flow.

[0140] The warehouse center collects spatial data such as storage area, main / auxiliary transportation channel, entrance / exit, picking table, etc. through 3D modeling and Internet of Things sensing devices; at the same time, it collects the positions, loads, power, and task queues of 60 AGVs in real time, building a task state atlas to provide dynamic input for the scheduling system.

[0141] The order system generates thousands of sorting tasks in real time, each task containing "picking-transportation-delivery-return" operations. The system maps these tasks to the graph, builds a three-dimensional model of task-path-node, labels potential intersection points such as main channels and merging areas, and generates a multi-AGV collaborative scheduling graph for the current scheduling period.

[0142] The scheduling system considers factors such as path length, current traffic pressure, and AGV remaining power to assign weights to tasks. For example: heavy-load AGVs prefer smooth paths, low-power AGVs automatically avoid high-conflict areas, and urgent single tasks prefer to avoid busy areas. A comprehensive scheduling weight model is formed.

[0143] The system uses an improved A* algorithm combined with scheduling weights to plan paths for all tasks and match AGVs. It automatically outputs the task sequence, path nodes, estimated completion time, and resource occupation interval for each AGV. The results are generated within 2 seconds without human intervention.

[0144] Mapping the path scheme to the scheduling timeline reveals that the peak congestion period is from 9:30 to 10:00 in the morning. The dense traffic on the main channel segment R12 causes potential congestion. The system automatically implements staggered start, path detour, and slow start control for some AGVs to achieve avoidance optimization.

[0145] The system calculates the average traffic frequency and task density of all path segments during the scheduling period, and combines the actual passing time of AGVs and task delay information to identify 3 high-risk bottleneck segments (R12, R19, R27). Simulation analysis shows that these areas affect the overall throughput efficiency by 12%. The system adjusts the scheduling weight based on the simulation results to automatically adjust the scheduling parameters for the next period.

[0146] Integrating task status graph, scheduling graph, path results, and verification data, the scheduling system forms a multi-AGV scheduling holographic data set and implements a closed-loop control mechanism driven by task status. The system automatically updates the status every 5 seconds and reconstructs the path and task allocation in real time. During peak hours, it maintains a 99.3% non-delay execution rate, and the average empty running rate of AGVs is reduced by 22%. The application effect is summarized in the following table:

[0147]

[0148] This technical solution is particularly suitable for intelligent warehouse environments with multiple tasks, high density, and frequent dynamic changes. Whether it is e-commerce logistics, pharmaceutical distribution, or industrial product sorting, it can significantly improve the efficiency, stability, and intelligence level of the scheduling system.

[0149] In the description of the specification, reference to "one embodiment", "an example", "a specific example" or the like means that a particular feature, structure, material or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the application. The appearances of the phrases "in one embodiment", "an example", "a specific example" or the like in various places in the specification are not necessarily referring to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0150] The preferred embodiments of the application disclosed above are only to help explain the application. The preferred embodiments do not describe all the details of the application and limit the application to the specific embodiments. Obviously, many modifications and variations can be made in light of the contents of the specification. The specification selects and specifically describes these embodiments in order to better explain the principles and practical application of the application, so that those skilled in the art can well understand and utilize the application. The application is limited only by the claims and their full scope and equivalents.

Claims

1. A multi-AGV scheduling method for intelligent warehousing, characterized by: The scheduling method comprises the following steps: Step S1: The collection end obtains the structured layout information of the warehouse factory and initializes the task state map based on the current operating status of the AGV cluster; Step S2: Mark the intersection and resource occupancy of AGVs on spatial paths and time resources to generate a multi-AGV collaborative scheduling map; Step S3: Assign comprehensive scheduling weights to each feasible path and task allocation scheme, and construct a multi-objective scheduling trade-off factor model; Step S4: Using a graph optimization algorithm combined with a multi-objective scheduling trade-off factor model to solve the task path, and outputting a task allocation and path planning solution; Step S5: Load the task assignment and path planning scheme into the timeline, perform timing simulation on the path, identify potential congestion points or resource conflict areas in the path, and optimize some AGV paths based on the identification results; Step S6: Count the time-traffic pressure index of each path within the scheduling period, use the time-traffic pressure index and the actual AGV operation status as input, and perform scheduling verification through multiple rounds of offline simulation; Step S7: Integrate the multi-AGV scheduling holographic dataset and build a multi-AGV closed-loop scheduling solution driven by task status; Perform timing simulation on the path to identify potential congestion points or resource conflict areas in the path, and optimize some AGV paths based on the identification results. This includes the following steps: Construct each path in the task allocation and path planning scheme Time interval ; The potential congestion point or resource conflict area judgment expression is: , i≠j, where Indicates that the i-th AGV is in the path segment The actual occupied time interval on Indicates that the jth AGV is in the path segment The actual occupied time interval on All paths that meet the judgment expression are marked as conflict-sensitive segments, and the time traffic pressure index is introduced. , which is used to describe the average AGV occupancy intensity on the path segment per unit time, sort the traffic pressure levels of all conflict-sensitive segments according to the time traffic pressure index, and execute the path conflict optimization strategy based on the simulation results, including staggered driving, path reselection, and soft start control; The time-traffic pressure index of each path within the scheduling period is calculated. The time-traffic pressure index and the actual AGV operation status are used as inputs. The scheduling verification is performed through multiple rounds of offline simulation, including the following steps: The average AGV passage frequency and task density The time-based traffic pressure index is obtained by weighted calculation; Introducing dynamic feedback parameters for task execution, extracting each AGV's transit time and scheduling delay feedback from actual operation data; The time-travel pressure indicators of all paths and the actual operating status of the AGV are input into multiple rounds of offline simulation programs to simulate the operating performance under different scheduling parameters; Average AGV passage frequency Indicates the path The average number of times an AGV passes through during the entire scheduling cycle is expressed as: ,in, For path The frequency of passage, is the total scheduling cycle time, is the number of AGVs, is the indicator function, when AGV Passing the path within the cycle When , the value is 1, otherwise it is 0, the task density The task coverage on the path is measured, which is defined as the total number of times a unit-length path is accessed by tasks within the scheduling period. The calculation expression is: ,in, For path The task density, is the physical length of the path, is the total number of tasks, Represents task T j Whether the path is involved; Based on the above two characteristics, define the path Time traffic pressure index for: ,in, and are the weight coefficients of passage frequency and task density, For path The task density, For path Frequency of traffic; Calculate the average traffic frequency fluctuation coefficient of each path segment in multiple scheduling cycles and task density fluctuation coefficient , and dynamically adjust the weight accordingly: .

2. The intelligent warehousing multi-AGV scheduling method according to claim 1, characterized in that: Assign a comprehensive scheduling weight to each feasible path and task allocation scheme, and build a multi-objective scheduling trade-off factor model, which includes the following steps: Based on the path distance D ij and resource conflict cost C ij Construct a comprehensive scheduling weight function of paths and tasks to evaluate AGVs Execute Task T j The resulting comprehensive scheduling cost W ij .

3. The intelligent warehousing multi-AGV scheduling method according to claim 2, characterized in that: The resource conflict cost Cij is defined as: ,in, Indicates AGV On the path The time interval for executing task Tj, n is the number of paths, Indicates the task time interval of other AGVs on the same path, function is an indicator function that outputs 1 when the two time intervals intersect, otherwise it outputs 0. Indicates AGV On the path Execute task T j time interval, represents the intersection of two time intervals, If the intersection of the two time intervals is not empty, that is, there is a time overlap, the path is defined as a potential conflict segment.

4. The intelligent warehousing multi-AGV scheduling method according to claim 3 is characterized by: The intersection and resource occupancy of AGVs in spatial paths and time resources are marked to generate a multi-AGV collaborative scheduling graph, which includes the following steps: Map the collected transport task queue to the constructed task state graph. Transport tasks include pickup, handling, and delivery. Decompose the transportation task process into task units ,in For the transport missions, is the i-th subtask step, and s is the number of subtask steps; Extracting the paths involved in task execution in space , and mapped to the spatial topology of the scheduling graph in the node dimension; In the time dimension, the task process is time-sliced ​​according to the task time window to form a three-dimensional scheduling diagram data body. , forming a three-dimensional mapping structure of task-path-node; In each path The time occupancy interval is defined above, and the resource occupancy tensor is introduced to describe the time domain competition relationship of AGV on system resources; Based on the scheduling graph and resource occupancy tensor, the AGV behavior data and task path timing are integrated to generate a multi-AGV collaborative scheduling graph.

5. The intelligent warehousing multi-AGV scheduling method according to claim 4, characterized in that: Scheduling diagram data body forming a three-dimensional structure , expressed as: ,in, Represents the set of spatial nodes in a three-dimensional scheduling graph, is the set of path edges, It is a time layer structure used to describe the time interval occupied by each task unit on the path.

6. The intelligent warehousing multi-AGV scheduling method according to claim 5, characterized in that: The data collection end obtains the structured layout information of the storage plant, including the spatial coordinates and capacity parameters of the storage sites, the topological structure and access constraints of the transportation paths, the resource scheduling priority and accessibility determination of the transfer nodes, and the operating capacity parameters of the operating platforms. Obtain the dynamic operating status data of the AGV cluster, including the current location coordinates, remaining power, current load weight, and task queue status of each AGV.

7. The intelligent warehousing multi-AGV scheduling method according to claim 6, characterized in that: The collection end obtains the structured layout information of the warehouse factory and initializes the task state map based on the current operating status of the AGV cluster, including the following steps: Combine structured layout information and operation status data to construct a task status graph, using a graph structure to represent the multiple relationships among AGVs, tasks, paths, and resources. Each node in the task state graph represents a scheduling element, and each edge represents the task transfer path or resource occupancy relationship; Task status graph Expressed as: ,in, Represents the node set in the graph, including AGV nodes, task nodes and resource nodes. Represents the set of edges between nodes, is the weight matrix.

8. An intelligent warehousing multi-AGV scheduling system, used to implement the scheduling method according to any one of claims 1 to 7, characterized in that: It includes graph generation module, solution output module, simulation verification module and scheduling solution construction module; Graph generation module: This module obtains the structured layout information of the warehouse factory, initializes the task state graph based on the current operating status of the AGV cluster, marks the intersection and resource occupation of AGVs in spatial paths and time resources, and generates a multi-AGV collaborative scheduling graph; Solution output module: Assigns comprehensive scheduling weights to each feasible path and task allocation solution, constructs a multi-objective scheduling trade-off factor model, uses a graph optimization algorithm combined with the multi-objective scheduling trade-off factor model to solve the task path, and outputs the task allocation and path planning solution; Simulation Verification Module: This module loads the task assignment and path planning scheme into the timeline, performs timing simulation on the path, identifies potential congestion points or resource conflict areas in the path, optimizes some AGV paths based on the identification results, calculates the time traffic pressure index of each path within the scheduling cycle, and uses the time traffic pressure index and the actual AGV operating status as input to perform scheduling verification through multiple rounds of offline simulation. Scheduling solution construction module: Integrate multi-AGV scheduling holographic data sets to build a multi-AGV closed-loop scheduling solution driven by task status.

Images