Multi-AGV scheduling method and system fused with emergency order insertion strategy

Abstract

The invention relates to the technical field of AGV scheduling, and particularly discloses a multi-AGV task scheduling system and method for solving an emergency order insertion problem, and the system comprises a task scheduling module, a path planning module, a real-time monitoring module and a map management module. The task scheduling module realizes multi-objective optimization task allocation based on an improved NSGA-II algorithm, optimizes an initialized population through an NNC (Neighbor Construction) strategy, enhances the global search capability in combination with a 2-opt mutation method, and designs an emergency order insertion strategy to dynamically respond to an emergency task; the real-time monitoring module integrates a MySQL database and a Visual Studio interactive interface, and supports electronic map construction (based on ArcMap10.4), AGV state real-time feedback and task dynamic adjustment; the map management module defines workshop layout and road traffic rules. According to the invention, by fusing the multi-objective optimization algorithm and the dynamic path planning technology, the problem of low efficiency of multi-AGV task allocation is solved, and the logistics efficiency and the system robustness of the track equipment assembly workshop are significantly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of AGV scheduling, and specifically discloses a multi-AGV task scheduling system and method for solving the problem of emergency order insertion. Background Art

[0002] Automated Guided Vehicles (AGVs) are core equipment for material handling in smart factories. The efficiency and robustness of their scheduling systems directly determine the efficiency of workshop logistics. Traditional AGV scheduling systems often utilize a centralized control architecture, relying on a single master computer to perform task allocation, path planning, and conflict detection. Traditional multi-AGV scheduling strategies often require re-dispatching AGVs when urgent orders require AGV completion. This scheduling approach, where a single AGV matches a single task, wastes resources and suffers from significant drawbacks: insufficient dynamic responsiveness; centralized systems struggle to cope with sudden task insertions and real-time path adjustments; task allocation algorithms are often based on static optimization, with weak global search capabilities and a tendency to fall into local optima, resulting in low task completion rates and high costs; and low system integration: Traditional solutions separate task scheduling, path planning, and monitoring modules, resulting in high data exchange latency, making it difficult to achieve real-time updates of electronic maps and accurate AGV status feedback, hindering the need for intelligent workshop upgrades. While some existing research has attempted optimization through partitioned scheduling or priority strategies, problems persist, such as poor task allocation coupling and weak adaptability to dynamic environments, making it difficult to meet the stringent requirements of intelligent manufacturing for efficient logistics.

[0003] Therefore, there is an urgent need for an AGV collaborative scheduling system that integrates multi-objective optimization algorithms and emergency order insertion strategies to solve core problems such as low task allocation efficiency and difficulty in solving emergency order insertion problems. Summary of the Invention

[0004] (1) Purpose of the present invention

[0005] A multi-AGV scheduling system and method based on multi-objective optimization and emergency order insertion strategy is provided to solve the following technical problems: the traditional multi-AGV scheduling system has low task allocation efficiency, wastes resources in emergency order insertion, and lacks the system's dynamic adjustment capability.

[0006] (2) Technical solution

[0007] A multi-AGV scheduling system based on multi-objective optimization and dynamic path planning includes a presentation layer, a business logic layer, a data layer, and a device layer. The following are the detailed steps of the invention:

[0008] Step 1: Algorithm operation process planning

[0009] (1) Presentation layer:

[0010] The presentation layer is responsible for data interaction, including real-time monitoring module and map management module.

[0011] (2) Device layer:

[0012] Directly interacts with AGVs and field devices, including device management module.

[0013] (3) Data layer:

[0014] Responsible for storing and managing the data of the system.

[0015] (4) Business logic layer:

[0016] The business logic layer undertakes the core algorithms of the system, responsible for processing data verification, executing calculations, applying rules and strategies, etc., including task allocation module and path planning module.

[0017] Step 2: Multi-AGV scheduling method

[0018] (1) Order splitting

[0019] The upper system splits the order into specific tasks, and the AGV scheduling system obtains the task information for the day. The system selects available AGVs into the working sequence and counts the maximum number of available AGVs; all AGVs are concentrated in the parking area under the initial condition to wait for instructions, and the current callable AGVs are recorded as S = {s1, s2, s3, …}; a set of feasible multi-AGV material handling path solutions is generated, there are N task points, the number of AGVs that can participate in scheduling is V, and the number of fitness functions is M; use numbers from 1 to N to represent the task point numbers in a single chromosome, use N+1 to N+V-1 to segment the task points, called "AGV separator identifier", to generate the running path of multiple AGVs.

[0020] (2) Task scheduling

[0021] The task scheduling module uses the assigned task data as raw input and sets the number of iterations, maximum distance, and maximum number of vehicles based on the optimization target. The correspondence between AGVs and transport tasks, as well as the order of tasks within the task sequence, is determined based on task priority and process. The first available AGV is selected, departing from the AGV parking point and starting the material delivery task at the scheduled time. Based on the principle of minimizing transportation costs, the candidate workstation with the lowest comprehensive transportation cost from the previous delivery workstation is selected and checked to see if it meets the constraints. If so, the workstation is added to the current path. Repeat the above selection process until the current AGV capacity limit is reached; if the current AGV transport route can no longer add new task nodes, the system will automatically schedule the new AGV transport and rebuild the transport route from the remaining unassigned workstations based on the same selection criteria; continue to iterate this scheduling mechanism until the access tasks of all workstations are completed; assign task priorities to different tasks; calculate the objective function value of the scheduling scheme to determine whether it meets the minimum value of the objective function; the objective function is recorded as: C = α1*C1+α2*C2, where C is the total cost, C1 is the fixed loss cost, C2 is the logistics cost, a1 is the fixed loss cost correction coefficient, and a2 is the transportation cost correction coefficient; the fixed loss cost is recorded as Where s is the total number of AGVs, i is the number of an AGV, and gi is the fixed cost of the i-th AGV; the logistics cost is recorded as: Where ei is the empty unit distance cost of AGV, n is the number of task points, j is the number of a task point, is the empty travel distance of AGV number i, d ij is the load travel distance of AGV number i to task point j, x is the load weight of the AGV, ω1 is the AGV no-load correction factor, and ω2 is the AGV load correction factor. If the objective function conditions and constraints are not met, a crossover and mutation strategy is used to generate a new scheduling plan and verify its objective function value.

[0022] (3) Path planning

[0023] The path planning module activates the D*Lite algorithm for steering cost optimization and synchronously loads the time window model to predict node occupancy conflicts.

[0024] (4) Emergency order insertion

[0025] When the emergency order is triggered, the system immediately starts the dynamic preemption mechanism: ① Real-time monitoring module scans the AGV task progress and marks the interruptible low priority tasks; ② Path planning module inserts virtual obstacles in the existing path, forcing the target AGV to perform local re-planning; ③ Task allocation module releases the occupied loading and unloading node resources; A, when the system receives an emergency order, it first evaluates the task completion degree and current location of all AGVs currently running. The system calculates the adaptability of each AGV to the emergency task according to the emergency task evaluation function, which takes into account the task completion progress of the AGV and its distance to the starting point of the emergency task. The AGV with the highest task completion degree and the closest distance to the starting point of the emergency task is selected to execute the new task, thereby achieving efficient use of resources. If the most suitable AGV can insert and execute the emergency task midway, its original path planning is adjusted and the emergency task is added to its task list. On the premise of ensuring that the original task completion time is not seriously affected, the scheduling system optimizes the travel arrangement of the selected AGV to include the new emergency task. The scheduling system monitors the running state and task execution of all AGVs in real time to ensure the stability and efficiency of the system throughout the process. For extreme emergency tasks that cannot be completed in time through existing AGV adjustments, the system will start the standby AGV mechanism to ensure that the emergency task can be handled quickly.

[0026] (5) Conflict resolution

[0027] The conflict resolution module is based on spatiotemporal joint detection technology. When the AGV predicted arrival time difference is detected, the priority arbitration strategy is automatically executed: ① The emergency task AGV obtains the right of road priority; ② The downgraded AGV receives the waiting instruction and activates the countdown reminder.

[0028] (6) Map management

[0029] The map management module implements dynamic topology update, and the path planning module reconstructs the local road network topology to ensure that the AGV bypass path is updated synchronously with the production rhythm.

[0030] (7) Real-time monitoring

[0031] During the task execution phase, the real-time monitoring module synchronously tracks the acceleration curve and steering angle changes of the AGV.

[0032] (8) System maintenance

[0033] In the system maintenance phase, the device management module automatically assigns charging tasks to AGVs with less than 15% power, and migrates the original assigned tasks to adjacent AGVs through dynamic load balancing algorithm to ensure the continuous operation availability of the system.

[0034] The beneficial effects of this invention: Compared with existing technologies, it can effectively solve the problems of insufficient dynamic adjustment capabilities of traditional multi-AGV scheduling systems, low efficiency in path conflict detection, and delayed emergency task response. By integrating the improved NSGA-II multi-objective optimization algorithm with the D*Lite algorithm for steering cost optimization, the task scheduling cost is reduced by 40.94%. The layered system architecture design supports modular collaboration of task management, path planning, and equipment monitoring, while retaining the AGV's autonomous decision-making capabilities, and increases the completion rate of emergency insertion tasks by 39.92%. This invention combines algorithmic innovation with engineering practicality, providing a highly scalable and robust multi-AGV collaborative solution for intelligent manufacturing workshops. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive effort.

[0036] Figure 1 This is a system architecture diagram of the present invention;

[0037] Figure 2 This is a flow chart of the emergency order insertion strategy; DETAILED DESCRIPTION

[0038] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly described below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0039] like Figure One The system architecture diagram is shown as follows. The task scheduling process of the present invention is as follows:

[0040] Step 1: The upper-level system divides the order into specific tasks, and the AGV scheduling system obtains the task information of the day;

[0041] Step 2: The task scheduling module takes the task data as the original input and sets the number of iterations, maximum distance, and maximum number of vehicles based on the optimization target. The corresponding relationship between AGVs and transportation tasks, as well as the handling order of the task sequence, is determined based on the task priority and process.

[0042] Step 3: After the path planning module obtains the task scheduling results, if it detects dynamic changes in the environment, it generates a new path based on the improved D*Lite algorithm. It introduces a turning cost factor into the heuristic function calculation and triggers path replanning when the turning cost exceeds the threshold.

[0043] Step 4: When multiple AGVs are operating in collaboration, the time window detection module obtains the current number of path conflicts. If a path conflict occurs, the priority adjustment strategy is activated to eliminate the conflict by stopping and waiting or reallocating the path.

[0044] Step 5: The dynamic path planning layer updates the status according to the real-time map. If the obstacle change rate Δ>5%, the improved D*Lite algorithm is automatically called to generate a smooth path in combination with the steering cost optimization.

[0045] Step 6: When the conflict resolution module detects that the AGV intersection conflict probability Pc>80%, it will prioritize adjusting the low-priority AGV path. If the adjustment fails, the emergency stop protocol will be activated;

[0046] Step 7: After receiving the planning instruction, if the path execution layer detects that the actual number of turns exceeds the planned value by more than 2 times, it automatically feeds back to the scheduling system to trigger path re-optimization;

[0047] like Figure Two For the emergency insertion strategy flow chart, the system generates a trajectory with a time dimension for each AGV based on its planned path. This includes recording the specific start and end time of each node passed by the AGV. This information is used to construct a four-tuple time window data model, which contains key information such as node location, start time, and end time. The time window set of all AGVs is traversed and analyzed. The system pays special attention to those AGV combinations that overlap in space (i.e., pass through the same node). If the time intervals of these AGVs on the same node intersect, the node is marked as a potential conflict point.

[0048] For all nodes marked as potential conflicts, a detailed topological analysis is performed to determine the conflict type, including but not limited to cross-path conflicts, overtaking conflicts at the start and end of a shared one-way lane, and conflict between vehicles moving in opposite directions within the lane.

[0049] Using a rolling time window mechanism, after completing conflict detection at each node, the system dynamically adjusts the time window parameters of subsequent nodes based on the actual operating speed of the AGV. This helps to predict and compensate for potential conflicts in advance, improving overall scheduling efficiency.

[0050] When a conflict is detected, the system initiates a priority arbitration mechanism. This process considers factors such as task urgency, AGV loading status, and path completion progress to calculate the priority weight of each AGV. High-priority AGVs are granted the right of way and occupy the conflicting node's resources through a locking mechanism. Low-priority AGVs are required to insert waiting time windows into their paths to avoid conflicts.

[0051] The above is a detailed description of an embodiment of the present invention. However, the content described is only a preferred embodiment of the present invention and should not be considered to limit the scope of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. A multi-objective optimization and emergency order scheduling method, characterized by: Task scheduling technology: supports real-time reception of urgent tasks; dynamically evaluates the completion rate and location adaptability of existing AGV tasks; selects the optimal AGV to insert new tasks, and activates the backup AGV mechanism when necessary. Multi-objective optimization algorithm: Comprehensively considers multiple optimization objectives such as transportation cost, energy consumption, and task completion time; supports multi-AGV path collaborative planning to avoid local optimality. Emergency order processing mechanism: supports dynamic preemptive scheduling; achieves rapid response through virtual obstacles and path replanning; introduces priority arbitration and backup AGV mechanisms to improve emergency response capabilities.

2. The task scheduling technology according to claim 1, wherein: (1) The upper-level system divides the order into specific tasks, and the AGV scheduling system obtains the task information of the day; (2) The task scheduling module takes the task data as the original input and sets the number of iterations, the maximum distance, and the maximum number of vehicles according to the optimization target direction. The corresponding relationship between AGV and transportation tasks, as well as the handling order of the task sequence, is determined based on the task priority and process. (3) The path planning module activates the D*Lite algorithm for steering cost optimization and simultaneously loads the time window model to predict node occupancy conflicts; (4) When an emergency insertion instruction is triggered, the system immediately activates the dynamic preemption mechanism: ① The real-time monitoring module scans the progress of each AGV task and marks the interruptible low-priority tasks; ② The path planning module inserts virtual obstacles into the existing path, forcing the target AGV to perform local replanning; ③ The task allocation module releases the occupied loading and unloading node resources; (5) The conflict resolution module is based on spatiotemporal joint detection technology. When the difference in the estimated arrival time of the AGV is detected, the priority arbitration strategy is automatically executed: ① The emergency task AGV obtains the right of way on the road; ② The downgraded AGV receives the waiting instruction and activates the countdown reminder. (6) The map management module implements dynamic topology updates, and the path planning module reconstructs the local road network topology to ensure that the AGV detour path is updated synchronously with the production rhythm. (7) During the task execution phase, the real-time monitoring module synchronously tracks the acceleration curve and steering angle changes of the AGV; (8) During the system maintenance phase, the equipment management module automatically assigns charging tasks to AGVs with a battery level below 15%, and migrates the original assigned tasks to adjacent AGVs through a dynamic load balancing algorithm to ensure the continuous operation availability of the system.

3. The multi-objective optimization algorithm according to claim 1, wherein: (1) The system screens available AGVs into the work sequence and counts the maximum number of available AGVs; (2) All AGVs are initially concentrated in the docking area waiting for instructions. The currently available AGVs are denoted as S = {s1, s2, s3, ...}; (3) Generate a feasible multi-AGV material handling path plan, with N task points, V number of AGVs that can participate in scheduling, and M number of fitness functions; use numbers from 1 to N to represent the task point number in a single chromosome, and use N+1 to N+V-1 to split the task points, called "AGV separation identifiers", to generate the operation path of multiple AGVs; (4) Select the first available AGV, start from the AGV parking point, and start the material delivery task at the scheduled time. Based on the principle of minimizing transportation costs, select the candidate workstation with the lowest comprehensive transportation cost from the previous delivery workstation and verify whether it meets the constraints; (5) If the preset conditions are met, the workstation is added to the current path. Repeat the above selection process until the current AGV capacity limit is reached; (6) If the current AGV transport route cannot continue to add new task nodes, the system will automatically schedule the new AGV transport and reconstruct the transport route from the remaining unassigned workstations based on the same selection criteria; Continue to iterate this scheduling mechanism until all workstation visits are completed; (7) Assign task priorities to different tasks, and record the priority of each task as Where β is the correction coefficient, p is the number of reverse order pairs; (8) Calculate the objective function value of the scheduling plan and determine whether it meets the minimum value of the objective function; the objective function is recorded as: C = α1*C1+α2*C2, where C is the total cost, C1 is the fixed loss cost, C2 is the logistics cost, a1 is the fixed loss cost correction coefficient, and a2 is the transportation cost correction coefficient; the fixed loss cost is recorded as Where s is the total number of AGVs, i is the number of an AGV, g i is the fixed cost of the i-th AGV; the logistics cost is recorded as where e i is the empty unit distance cost of AGV, n is the number of task points, j is the number of a task point, d0 i is the empty travel distance of AGV numbered i, d ij is the load travel distance of AGV numbered i to task point j, x is the load weight of AGV, ω1 is the AGV no-load correction coefficient, and ω2 is the AGV load correction coefficient. (9) If the objective function conditions and constraints are not met, a crossover and mutation strategy is adopted to generate a new scheduling plan and verify its objective function value.

4. The emergency order insertion method according to claim 1, wherein: (1) When the system receives an urgent order, it first evaluates the task completion and current location of all currently running AGVs. (2) The system calculates the suitability of each AGV for the emergency task based on the emergency task evaluation function. This step takes into account the AGV's task completion progress and its distance to the starting point of the emergency task. (3) Select an AGV with a high degree of task completion and closest to the starting point of the urgent task to perform this new task, thereby achieving effective utilization of resources. (4) If the most suitable AGV can be inserted midway and perform the emergency task, its original path planning is adjusted and the emergency task is added to its task list. (5) Under the premise of ensuring that the original task completion time will not be seriously affected, the scheduling system re-optimizes the selected AGV's itinerary to include the new urgent task. (6) The dispatching system monitors the operating status and task execution of all AGVs in real time to ensure the stability and efficiency of the system throughout the entire process. (7) For extremely urgent tasks that cannot be completed in time through the adjustment of existing AGVs, the system will activate the backup AGV mechanism to ensure that the urgent tasks can be handled quickly.

5. An AGV scheduling system based on multi-objective optimization and dynamic path planning is characterized by It includes presentation layer, business logic layer, data layer and device layer.

6. As described in claim 5, the presentation layer is the interface layer for user interaction with the system, responsible for visual display and user operation input; providing visualization of information such as AGV operating status, task progress, path trajectory, etc.; and displaying spatial information such as factory maps, AGV paths, obstacle distribution, and loading and unloading point locations.

7. As described in claim 5, the business logic layer is the core control center of the system, responsible for handling all intelligent calculations and policy execution related to scheduling and path planning; reasonably allocating tasks to each AGV, considering factors such as priority, distance, and load; dynamically responding to emergency tasks, and performing preemptive scheduling and path replanning.

8. As described in claim 5, the data layer is the core of the system and is responsible for storing, managing and accessing all data required during the operation of the system.

9. As described in claim 5, the device layer is the physical execution layer of the system, which directly communicates with the AGV body and on-site equipment to complete task execution and status feedback.

Images