Scheduling optimization method and system for one-way loop multi-load automated guided vehicle material distribution

Abstract

The present application belongs to the technical field of production logistics scheduling, and relates to a scheduling optimization method and system for one-way loop multi-load automated guided vehicle material distribution. For the one-way loop material distribution scene, the scheduling method constructs a mixed integer linear programming scheduling model coupling entity flow and load flow, unifies the loop track passing limit, workstation stopping rule and inventory upper and lower limit constraint into standardized modeling, adopts a just-in-time driven greedy random adaptive search algorithm for iterative optimization, generates a globally optimal scheduling scheme with the least total departure frequency under the boundary conditions of strictly meeting conflict-free vehicle operation, zero material supply interruption at the workstation and inventory not exceeding the upper limit, and issues the scheme to the loop material distribution system to complete the corresponding scheduling. The present application can effectively resolve the scheduling contradiction between capacity improvement and traffic safety under the physical constraints of one-way loop, and improve the distribution accuracy and capacity utilization efficiency.

Description

Technical Field

[0001] This invention belongs to the field of production logistics scheduling technology, specifically a scheduling optimization method and a scheduling optimization system for material distribution of unidirectional loop multi-load automated guided vehicles. Background Technology

[0002] Circular track material delivery systems, with their compact layout, fixed paths, and controllable flow, are widely used in industrial scenarios such as flexible manufacturing workshops, fiberglass production drawing areas, and automated warehousing. They are core logistics equipment for achieving automated material transfer between multiple workstations and supporting continuous production. Among them, unidirectional circular tracks are the most commonly used track type in the industry due to their simple structure and low control costs.

[0003] To adapt to the increased demand for material delivery brought about by increased production capacity, a common optimization approach in existing technologies is to increase the number of shuttle cars or automated guided vehicles (AGVs) running on the circular track, thereby shortening the delivery cycle and increasing the frequency of replenishment by increasing vehicle density. However, unidirectional circular tracks have inherent physical characteristics such as the inability to overtake and vehicles blocking the flow of vehicles along the entire track when stopping at workstations. As the number of vehicles increases to a certain extent, mutual interference between vehicles and queuing congestion caused by stopping at workstations will gradually intensify, leading to a decrease in the overall system efficiency, an increase in delivery delays, and in severe cases, even track deadlock. To address this issue, patent application CN202211211927.0 proposes a dual-station shuttle transport system and scheduling method. By setting two loading stations on a single shuttle, along with a rotary loading and unloading area and turntable around the circular track, the system completes the loading and unloading of materials by coordinating the timing of the loading and unloading stations. This improves the single-trip carrying capacity and loading and unloading efficiency without increasing the number of vehicles, and plays a certain role in alleviating the congestion problem caused by the increase in vehicles.

[0004] However, the scheduling logic of the above scheme is triggered by real-time loading and unloading demands and planned on a single transport cycle basis. It responds to the needs of each workstation one by one before arranging the stopping order and loading and unloading sequence of the shuttle for that trip. The optimization scope is limited to the loading and unloading efficiency of a single transport process. It does not take into account the material requirements of all workstations in the complete production cycle, the operating rules of vehicle round-trip cycles, and the hard constraints of loop track passage and workstation stopping. Therefore, it is difficult to coordinate and match the departure timing, the combination of workstations along the route, and the replenishment batch of each workstation under the physical constraints of a unidirectional loop track, and thus it is difficult to achieve the optimal overall transportation capacity efficiency of the system. At the same time, the timing coordination of the preceding and following workstations in this scheme can only be achieved by relying on dedicated supporting hardware such as rotary loading and unloading areas, rotary tables, and temporary storage areas. The scheduling logic is designed around the exchange of empty and full vehicles in a specific scenario, and the scenario adaptability is relatively limited. It is difficult to directly apply to the general flexible manufacturing material replenishment scenario without dedicated rotary devices and with continuous production of multiple workstations. More notably, most existing ring track scheduling schemes focus on improving the efficiency of vehicle passage and loading / unloading operations themselves, rarely combining the material consumption patterns of each workstation, the production requirements of inventory limits, and ring track passage constraints to design a unified scheduling strategy. The scheduling process is difficult to accurately match the material consumption rhythm of each workstation, making it difficult to stably guarantee the material supply required for continuous production at each workstation, and also difficult to control workstation inventory within a reasonable capacity range. It is impossible to control the total number of vehicle departures at the most economical level while simultaneously meeting traffic safety and production requirements. The scheduling contradiction between capacity improvement and traffic safety under the physical constraints of a unidirectional ring track has not yet been effectively resolved. Summary of the Invention

[0005] To address the inherent physical constraints of unidirectional circular track material delivery systems, such as the inability to overtake or stop and disrupt traffic flow, which make it difficult for existing scheduling methods to simultaneously achieve conflict-free operation and timely material supply, and hinder the coordinated improvement of delivery safety and efficiency, this invention provides a scheduling optimization method and a scheduling optimization system for unidirectional circular track multi-load automated guided vehicles (AGVs) material delivery.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A scheduling optimization method for material delivery of unidirectional loop multi-load automated guided vehicles (AGVs) is applied to a flexible manufacturing material delivery system consisting of a single replenishment point, multiple workstations, and a unidirectional circular track. The method includes the following steps: The system adopts a rack-based operation mechanism with scheduling constraints and optimization objectives. The scheduling constraints include: the number of goods loaded by a single automated guided vehicle (AGV) at one time does not exceed the preset rack slot limit; the departure time interval between any two AGVs is not less than the minimum safety interval; only one AGV is allowed to stop or pass through any workstation at any given time; and the material inventory of each workstation is always greater than or equal to zero and does not exceed the maximum inventory capacity of the corresponding workstation. The optimization objective is to minimize the total number of AGV departures within the planning period. Based on scheduling constraints and optimization objectives, and taking the minimization of the total number of departures under each constraint as the objective function, a mixed-integer linear programming scheduling model coupling entity flow and load flow is constructed. A just-in-time driven greedy stochastic adaptive search algorithm is used to solve the mixed-integer linear programming scheduling model. The latest departure time is derived in reverse from the global supply interruption warning. The remaining capacity of the automated guided vehicles is filled by a greedy stochastic mechanism of a restricted candidate set. After multiple rounds of iterative screening, the optimal automated guided vehicle scheduling scheme that satisfies all constraints is output and sent to the flexible manufacturing material distribution system for the allocation of automated guided vehicles.

[0007] As a further improvement to the above scheme, the process of constructing a mixed-integer linear programming scheduling model is as follows: Define the set parameters, automated guided vehicle (AGV) driving time parameters, and material delivery business parameters corresponding to the model as the input basis for the model; Three types of decision variables are set up: vehicle spatiotemporal trajectory type 0-1 discrete variables used to track the spatiotemporal operation trajectory of automated guided vehicles; resource system state type continuous variables used to characterize system resources and inventory status; and shelf load flow type continuous variables used to track changes in the quantity of goods carried by automated guided vehicles. The objective function is determined to minimize the total number of automated guided vehicle departures within the planning period. A constraint system is constructed that covers vehicle resource cycle constraints, online traffic conservation constraints, shelf load flow constraints, single-track implicit anti-conflict constraints, and dynamic inventory balance constraints. The logical coupling of entity flow and load flow is achieved by binding the upper bound of the vehicle spatiotemporal trajectory class 0-1 discrete variables with the shelf load flow class continuous variables.

[0008] As a further improvement to the above scheme: the set parameters include: the discretized time step set T={1,2,...,T} max}, where T max The planning cycle length is defined by the following: the set of workbench nodes is n={1,...,i,...,N}, where i is the workbench number and N is the total number of workbench nodes; the replenishment point is denoted as node 0; the set of unloaded goods in a single operation is K={1,2,3}. The parameters for the Automated Guided Vehicle (AGV) operation time include: the total physical number V of AGVs in the flexible manufacturing material delivery system. total Offline preparation time t after the automated guided vehicle returns to the replenishment point refill Minimum safe time interval H between departures, the time it takes for the vehicle to travel from the replenishment point to the first workbench. The number of time steps required to travel at a constant speed from the i-th workbench to the (i+1)-th workbench. The number of steps required for the vehicle to travel at a constant speed from workbench N back to the replenishment point. The total fixed time steps t required for an automated guided vehicle to stop and unload k goods u (k), where k is the number of fully loaded shelves in a single unloading, and the value is 1, 2, or 3; Material distribution operation parameters include: the material consumption D per unit time for the i-th workstation. i The precise quantity Q of materials contained in a single fully loaded shelf corresponding to the i-th workbench. i The initial inventory I of the i-th workbench i,0 The maximum material inventory capacity of the i-th workbench .

[0009] As a further improvement to the above scheme: the vehicle spatiotemporal trajectory class of 0-1 discrete variables includes: the departure state variable u at time t. t This indicates whether an automated guided vehicle (AGV) carrying goods has been dispatched from the replenishment point at time t; 1 indicates yes, 0 indicates no. The state variable w of the replenishment point is returned at time t. t The variable p indicates whether an automated guided vehicle (AGV) that has completed its task returns to the replenishment point and begins preparation at time t; 1 indicates yes, 0 indicates no. The state variable p represents the vehicle traveling straight past the workbench at time t. i,t The variable t represents whether an automated guided vehicle (AGV) arrives at the i-th workbench and proceeds straight through at time t; 1 indicates yes, 0 indicates no. The variable t also represents the state variables for unloading k goods at time t. , indicates whether at time t there is an automated guided vehicle that arrives at the i-th workbench and starts to dock and unload k goods, 1 for yes and 0 for no; Continuous variables in the resource system state class include: the number of automated guided vehicles (AGVs) available at the replenishment point at time t (A). t , represents the number of automated guided vehicles (AGVs) in an available standby state at the replenishment point at time t; I represents the material inventory at time t. i,t , represents the remaining inventory of materials at the i-th workbench at time t; Continuous variables in the shelf load flow class include: the number of remaining goods when they arrive at the workbench. , representing the remaining fully loaded cargo carried by the automated guided vehicle (AGV) that arrives at the i-th workbench at time t; and the remaining cargo after passing the workbench. , representing the remaining cargo carried by the automated guided vehicle (AGV) that travels straight past the i-th workbench at time t; and the remaining cargo after stopping and unloading. , represents the number of goods remaining on the automated guided vehicle after stopping at the i-th workbench and unloading k goods at time t.

[0010] As a further improvement to the above scheme, the specific contents of various constraints in the constraint system are as follows: Vehicle resource cycle constraints are used to regulate the resource conservation relationship throughout the entire lifecycle of automated guided vehicles (AGVs), from departure and delivery to return to the replenishment point for preparation. Specifically, this includes: The vehicle return constraint is expressed as follows: ; In the formula, express At any given moment, if an automated guided vehicle arrives at workbench N and moves straight through, 1 indicates yes, and 0 indicates no. This indicates whether an automated guided vehicle (AGV) arrives at workbench N at time t and begins to dock and unload k goods; 1 indicates yes and 0 indicates no. Indicates any one item; Vehicle balance constraints can be used, expressed as: ; In the formula, A t-1 This indicates the number of automated guided vehicles (AGVs) that are available and ready to go at the replenishment point at time t-1. express At any given moment, if an automated guided vehicle (AGV) that has completed its mission returns to the replenishment point and begins preparation, the value is 1; otherwise, it is 0. The departure quantity limit constraint is expressed as follows: ; The initial condition constraints are expressed as follows: ; In the formula, A0 represents the number of automated guided vehicles (AGVs) that are in an available standby state at the initial replenishment point; Online traffic conservation constraints are used to ensure the continuity and uniqueness of the spatiotemporal trajectory of the automated guided vehicle on the circular track, specifically including: The first node enters the conservation constraint, expressed as: ; In the formula, p 1,t This indicates whether an automated guided vehicle (AGV) arrives at the first workbench and passes through it at time t; 1 indicates yes, and 0 indicates no. This indicates whether at time t, an automated guided vehicle (AGV) has arrived at the first workbench and started to dock and unload k goods; 1 indicates yes, 0 indicates no. express Whether an automated guided vehicle (AGV) carrying goods has been dispatched from the replenishment point at any given time is 1 if yes and 0 if no. The conservation constraint passed down through cascading workbenches is expressed as follows: ; In the formula, p i,t This indicates whether an automated guided vehicle (AGV) arrives at the i-th workbench and passes straight through at time t; 1 indicates yes, and 0 indicates no. This indicates whether at time t an automated guided vehicle (AGV) arrives at the i-th workbench and begins to dock and unload k goods; 1 indicates yes, 0 indicates no. The time steps required to travel at a constant speed from the (i-1)th workbench to the ith workbench; express At any given moment, if an automated guided vehicle arrives at the (i-1)th workstation and proceeds straight through, the value is 1; otherwise, it is 0. express At any given time, if an automated guided vehicle arrives at the (i-1)th workbench and begins to dock and unload k goods, 1 indicates yes and 0 indicates no. Shelf load flow constraints are used to track changes in the quantity of goods carried during the movement of automated guided vehicles (AGVs) and to match the correspondence between load flow and physical operation actions. Specifically, they include: The initial departure capacity constraint is expressed as: ; In the formula, Let t be the remaining fully loaded cargo carried by the automated guided vehicle that arrives at the first workbench at time t; The constraint for load propagation across nodes is expressed as follows: ; In the formula, Let t be the remaining fully loaded cargo carried by the automated guided vehicle that arrives at the i-th workstation at time t; for The number of remaining goods carried by the automated guided vehicle that passes through the (i-1)th workbench at any given time; for The number of remaining goods on the vehicle is automatically guided after the (i-1)th workbench stops and k goods are unloaded. The expression for unloading the deduction constraint is: ; In the formula, Let t be the number of remaining goods carried by the automated guided vehicle that passes the i-th workbench in a straight line at time t; Let t be the number of goods remaining on the autoguide vehicle after the i-th workbench stops and k goods are unloaded; Physical capacity binding constraints include straight-through scenario binding constraints and docking / unloading scenario binding constraints, with the following expressions: ; ; Implicit anti-collision constraints on monorails are used to avoid track collisions and platform congestion from a planning perspective. Specifically, they include: The departure safety interval constraint is expressed as: ; In the formula, For the time step summation variable; for Whether an automated guided vehicle (AGV) carrying goods has been dispatched from the replenishment point at any given time is 1 if yes and 0 if no. The physical mutual exclusion constraint of the workbench is expressed as follows: ; In the formula, express At any given moment, 1 indicates whether an automated guided vehicle (AGV) has arrived at the i-th workbench and started to dock and unload k goods; 0 indicates otherwise. Dynamic inventory balancing constraints are used to maintain a reasonable range of material inventory at each workstation, ensuring zero supply disruptions and no inventory overflows. Specifically, this includes: The dynamic inventory update constraint is expressed as follows: ; In the formula, I i,t-1 This represents the remaining inventory of materials at the i-th workbench at time t-1; express At any given moment, 1 indicates whether an automated guided vehicle (AGV) has arrived at the i-th workbench and started to dock and unload k goods; 0 indicates otherwise. Inventory boundary constraints, expressed as: .

[0011] As a further improvement to the above scheme: the logical coupling between the entity flow and the load flow is achieved through physical capacity binding constraints; the physical capacity binding constraints use the vehicle spatiotemporal trajectory class 0-1 discrete variables as the upper bound control factor of the shelf load flow class continuous variables, binding the entity existence state of the automated guided vehicle with the quantity of goods carried, including the binding constraints of the straight-through scenario and the binding constraints of the docking and unloading scenario. Through the two binding constraints, when no automated guided vehicle arrives at the corresponding workbench, the value of the corresponding shelf load flow variable is limited to zero.

[0012] As a further improvement to the above scheme, the specific process of solving the problem using a just-in-time driven greedy stochastic adaptive search algorithm is as follows: Initialize the algorithm parameters, set the maximum number of iterations, and initialize the optimal scheduling scheme and the optimal end-of-period inventory balance score; The process of constructing a heuristic process to generate a feasible scheduling scheme is as follows: the first workstation in the entire field with a risk of supply interruption is located by global supply interruption early warning monitoring and the corresponding time of supply interruption. Based on the time of supply interruption, the latest departure time is deduced in reverse. The legal departure window is searched forward and a fire-fighting slot is reserved for the workstation with supply interruption. The remaining shelf slots of the automated guided vehicle are filled by a greedy random mechanism of a limited candidate set. After the feasibility is verified, the corresponding spatiotemporal resources are locked and the system status is updated. The global supply interruption early warning monitoring and slot filling process is executed cyclically until all workstations have no risk of supply interruption within the planning period, and a feasible scheduling scheme is obtained. Perform a dual-objective evaluation on the generated feasible scheduling schemes, and update the current optimal scheduling scheme based on the evaluation results; Repeat the heuristic process of constructing the optimal solution and the optimal solution update steps until the preset maximum number of iterations is reached, and output the final optimal automated guided vehicle scheduling scheme.

[0013] As a further improvement to the above scheme, the specific steps for filling the remaining capacity of automated guided vehicles using a greedy random mechanism with a limited candidate set are as follows: Construct a candidate workbench set: Iterate through all workbench, tentatively add an unloading slot to each workbench and verify the validity of the process, add the workbench that passes the verification to the candidate set, and calculate the supply interruption urgency score of the corresponding workbench. The earlier the supply interruption occurs, the higher the urgency. Candidate set sorting: Sort the workbenches in the candidate set from highest to lowest according to the urgency of the supply disruption; Random selection filling: Randomly select one workbench from the top three workbench of the sorted candidate set and add an unloading slot to it; Repeat the steps of building, sorting and randomly selecting to fill the candidate set until all remaining shelf slots of the automated guided vehicle are filled, or there are no valid candidate workbenches.

[0014] As a further improvement to the above scheme, the rules for dual-objective evaluation are as follows: The first priority evaluation target is the total number of departures; the fewer the total number of departures, the higher the priority of the scheduling scheme. The second priority evaluation target is the ending inventory balance score, which is the sum of the ending inventory of all workstations. When the total number of departures is the same for two scheduling schemes, the scheduling scheme with the higher end-of-period inventory balance score is judged to be better. The conditions for updating the optimal scheduling scheme are: the total number of departures of the current feasible scheduling scheme is less than that of the current optimal scheme, or the total number of departures is the same but the ending inventory balance score is higher than that of the current optimal scheme.

[0015] A scheduling optimization system for material delivery of unidirectional loop multi-load automated guided vehicles (AGVs) is applied to a flexible manufacturing material delivery system consisting of a single replenishment point, multiple workstations, and a unidirectional circular track. The system executes a scheduling optimization method for material delivery of unidirectional loop multi-load AGVs. The system includes: The scenario configuration and constraint definition module is used to configure scheduling constraints and optimization objectives when adopting the rack-on-vehicle operation mechanism. The mixed-integer linear programming model building module is used to construct a mixed-integer linear programming scheduling model that couples entity flow and load flow. The just-in-time driven greedy stochastic adaptive search solution module is used to solve the mixed-integer linear programming scheduling model, output the optimal automated guided vehicle scheduling scheme that satisfies all constraints, and send it to the flexible manufacturing material distribution system.

[0016] Compared with the prior art, the beneficial effects of the present invention are: 1. In the scheduling optimization method, a three-layer technical path is adopted to systematically resolve the contradiction between operational safety and delivery efficiency under the physical constraints of a unidirectional circular track. This involves pre-fixing of constraint boundaries, global coordination through dual-flow coupling modeling, and efficient optimization using heuristic algorithms. The specific solution logic is as follows: First, the inherent physical constraints and production and delivery requirements of the unidirectional circular track are all fixed into inviolable scheduling boundary rules: the minimum safe interval between departures and the rule that only one vehicle can stop / pass through a single workstation at a time are used to strictly control the safety bottom line of the circular track from the source of scheduling design, completely avoiding the risks of vehicle rear-end collisions and workstation congestion, and eliminating the safety hazards caused by blindly increasing transport capacity. At the same time, the rigid production constraint of "workstation inventory is always greater than or equal to zero and does not exceed the maximum capacity" is used to pre-fix the requirement of timely material supply as a mandatory boundary of scheduling, rather than a reactive response afterward, ensuring the material replenishment needs of continuous production from the rule level. On this basis, a mixed-integer linear programming scheduling model coupling entity flow and load flow is constructed to unify the originally scattered vehicle operation, material replenishment, and inventory change processes into a unified standardized mathematical framework for global coordination. The model uses discrete variables (0-1) representing vehicle spatiotemporal trajectories to depict the physical trajectory of vehicles, and continuous variables representing shelf load flow to track changes in cargo volume. The 0-1 variables serve as upper bound control factors for the continuous variables to achieve dual-flow binding, ensuring physical consistency that "no physical vehicle means no corresponding cargo load" and avoiding logical loopholes. Simultaneously, all constraints, including vehicle resource circulation, loop flow conservation, workstation conflict prevention, and inventory boundaries, are incorporated into the model. With minimizing the total number of departures as the objective function, the actual scheduling problem is transformed into a solvable mathematical programming problem. Compared to existing single-trip, responsive local scheduling methods, this model covers the complete planning cycle and all workstations, enabling global coordination and matching of departure timing, workstation combinations along the route, and unloading batches at each workstation. It seeks the globally optimal solution within the common boundaries of all constraints, rather than merely pursuing local efficiency improvements for a single transport. Finally, a just-in-time driven greedy stochastic adaptive search algorithm is used to efficiently solve the model, balancing solution speed and solution quality. The algorithm uses a global supply disruption warning as the scheduling trigger, working backwards to deduce the latest departure time and prioritizing material replenishment for the most urgent workstations, aligning with the core requirements of just-in-time delivery. It fills remaining capacity with a greedy random mechanism using a restricted candidate set, prioritizing delivery timeliness based on the urgency of the supply disruption while avoiding local optima problems common in ordinary greedy algorithms through random selection. This allows it to continuously search for scheduling schemes with fewer total departures while satisfying all constraints. All candidate routes undergo constraint verification before resource locking, ensuring the final output scheme fully meets the requirements for loopway safety and workstation inventory. The final optimal scheduling scheme is directly deployed to the delivery system for execution, achieving a closed loop from planning to implementation. Under the premise of strictly ensuring conflict-free operation and timely material supply, it completes all delivery tasks with the fewest departures, effectively improving the overall system capacity efficiency.

[0017] 2. This invention clarifies the specific expressions and physical meanings of five types of constraints: vehicle resource circulation, online flow conservation, shelf load flow, implicit anti-collision on a single track, and dynamic inventory balance. It fully embeds the traffic safety constraints of the unidirectional ring track and the production business constraints of workstation inventory into the mathematical model. This not only strictly limits the departure interval of the ring track and the docking rules of the workstations, avoiding the risks of vehicle rear-end collisions and track congestion from the planning source, but also accurately matches the material consumption rhythm of each workstation, ensuring timely material replenishment and preventing inventory from exceeding limits. At the same time, the constraint system achieves a strict correspondence between the physical flow of vehicles and the flow of goods through physical capacity binding rules, eliminating logical loopholes in the model, ensuring the physical consistency and engineering feasibility of the scheduling scheme, and providing a rigorous and complete constraint foundation for solving the globally optimal scheduling scheme.

[0018] 3. After obtaining the optimal scheduling plan, it is sent to the vehicle control terminal, replenishment point operation terminal and production terminal of each workbench of the flexible manufacturing material distribution system. It is used to directly drive all automated guided vehicles to circulate and deliver materials along the unidirectional circular track according to the preset time sequence, and at the same time guide the replenishment end to complete the pre-loading of materials according to the shift and the production end to collect materials according to the rhythm. This solution pre-plans the spatiotemporal trajectories of all vehicles and arranges the parking sequence at workstations. During operation, it strictly adheres to departure intervals and workstation mutual exclusion rules, eliminating the need for complex real-time dynamic scheduling and fundamentally avoiding operational risks such as vehicle rear-end collisions and track congestion. Simultaneously, the solution precisely matches the material consumption rhythm of each workstation, stably maintaining material inventory within a reasonable range. This ensures timely replenishment of materials for continuous production, avoiding losses due to supply disruptions and production stoppages, while also preventing excessive inventory buildup at workstations that occupies workspace. Furthermore, the solution is generated with the goal of minimizing the total number of departures, maximizing the utilization rate of single-vehicle capacity while fully meeting delivery needs. This reduces ineffective empty runs and redundant departures, lowers vehicle operating energy consumption and replenishment workload, and achieves a synergistic improvement in operational safety, delivery timeliness, and transportation economy for unidirectional circular rail material delivery. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of a flexible manufacturing material delivery system.

[0020] Figure 2 The flowchart shows the scheduling optimization method. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] I. Flexible Manufacturing Material Distribution System

[0023] The scheduling optimization method for material delivery of unidirectional loop multi-load automated guided vehicles (AGVs) of the present invention is applied to a flexible manufacturing material delivery system consisting of a single replenishment point, multiple workbenches, and a unidirectional circular track. This system is designed for continuous production line material delivery scenarios in flexible manufacturing workshops. The system adopts a single-loop track architecture with unidirectional clockwise movement and relies on multi-load AGVs to complete the just-in-time replenishment of materials. The system architecture, operation mechanism, and core rules are described in detail below.

[0024] 1. System Topology and Basic Components

[0025] like Figure 1 As shown, the flexible manufacturing material distribution system consists of one replenishment point, N workstations (6 in this embodiment), and a unidirectional clockwise circular track. Each workstation strictly corresponds to the assembly task of a unique part. During the production process, the corresponding material is continuously consumed. The material is loaded from the replenishment point by an automated guided vehicle and transported along the circular track to the corresponding workstation for unloading and replenishment.

[0026] The system fleet consists of multiple homogeneous automated guided vehicles (AGVs). Each AGV adopts a "1 tractor-mounted vehicle + 2 trailers" structure, with a total of 3 fixed rack slots on the vehicle body and trailers. The maximum capacity of a single vehicle is 3 fully loaded racks.

[0027] 2. Shelving operation mechanism that allows the racks to remain on the vehicle

[0028] The system employs a core operating mechanism where the racks remain attached to the automated guided vehicle (AGV). During system operation, only batch transfers of materials occur; the racks themselves do not detach from the vehicle during material loading and unloading. Specific rules are as follows: The automated guided vehicle only completes the loading operation at the replenishment point. It can load 1 to 3 full racks of goods as needed, and empty racks are always kept in the vehicle slots. When the automated guided vehicle travels along the track to the corresponding workbench, it unloads only the corresponding number of goods. The empty shelves after unloading continue to travel with the automated guided vehicle and eventually return to the replenishment point. After the automated guided vehicle returns to the replenishment point, it reloads the goods on the empty shelves and begins the next round of delivery.

[0029] 3. Complete operation process of automated guided vehicles

[0030] The complete workflow for a single delivery task by a single automated guided vehicle is as follows: (1) After the automated guided vehicle completes the loading of goods and equipment preparation at the replenishment point, it departs from the replenishment point and enters the circular track; (2) Travel clockwise at a constant speed along the circular track, passing each workbench in turn; (3) When it arrives at the workbench where it needs to be unloaded, stop at the workbench and unload the corresponding number of goods. After unloading, continue to travel along the track. (4) After completing all the preset unloading tasks, continue to travel along the circular track and eventually return to the replenishment point; (5) Carry out offline preparation at the replenishment point, including reloading of goods, equipment inspection and maintenance, etc. After the preparation is completed, enter the standby state and prepare for the next departure.

[0031] 4. Implicit physical anti-collision mechanism

[0032] Because the circular track is a unidirectional monorail structure, any automated guided vehicle (AGV) stopping at the work platform will block the entire main road, preventing subsequent AGVs from overtaking. This system achieves absolute conflict prevention at the source of scheduling and planning through two implicit rules, eliminating the need for real-time communication and dynamic path adjustments during operation: (1) Departure safety interval rule: The departure time interval between any two automated guided vehicles shall not be less than the minimum safety interval H to prevent rear-end collisions of automated guided vehicles during track travel; (2) Absolute space mutual exclusion rule of work platform: At any given time, only one automated guided vehicle can stop or pass through any work platform to prevent multiple automated guided vehicles from colliding, congesting or deadlocking at the work platform.

[0033] 5. Define the scheduling problem

[0034] The one-way loop automated guided vehicle (AGV) scheduling problem solved by this invention can be described as follows: Given the total physical number of AGVs in the system, the material consumption cycle of each workbench, the material capacity of a single shelf, the initial inventory level, and the AGV travel time and unloading time, under the premise of satisfying all physical and production constraints, determine the departure time of each AGV and the unloading quantity of each workbench, with the optimization objective of minimizing the total number of departures, while ensuring zero supply interruption and no warehouse overload on the production line.

[0035] The core constraints that need to be met include: The number of goods loaded by a single automated guided vehicle at one time shall not exceed the preset upper limit of rack slots (i.e., no more than 3 slots). The time interval between any two automated guided vehicles shall not be less than the minimum safe interval. At any given time, only one automated guided vehicle is allowed to stop or pass through any workstation. The material inventory at each workbench is always greater than or equal to zero and does not exceed the maximum inventory capacity of the corresponding workbench.

[0036] II. Scheduling Optimization Method for Material Delivery of Multi-Load Automated Guided Vehicles on One-Way Loop Lines

[0037] like Figure 2As shown, this scheduling optimization method consists of three core steps: determining scheduling constraints and optimization objectives, constructing a mixed-integer linear programming scheduling model that couples entity flow and load flow, and solving and outputting a scheduling scheme using a just-in-time driven greedy stochastic adaptive search algorithm. The following provides a detailed explanation of each step.

[0038] (a) Determine scheduling constraints and optimization objectives

[0039] For the aforementioned flexible manufacturing material distribution system, based on the shelf-based operation mechanism, all scheduling constraints and core optimization objectives of the system are defined: Scheduling constraints (1) Load capacity constraint: The number of goods loaded by a single automated guided vehicle in a single trip shall not exceed 3 full-load racks; (2) Departure safety constraints: The departure time interval between any two automated guided vehicles shall not be less than the minimum safety interval H; (3) Workstation mutual exclusion constraint: At any given time, only one automated guided vehicle is allowed to stop or move straight through any workstation; (4) Inventory boundary constraints: The material inventory of each workbench is always greater than or equal to zero (zero supply disruption) and does not exceed the maximum inventory capacity of the corresponding workbench (no overstocking). (5) Vehicle resource constraints: The number of automated guided vehicles that can depart at any time shall not exceed the number of available vehicles at the current replenishment point.

[0040] 2. Optimization Objectives

[0041] The core optimization objective is to minimize the total number of automated guided vehicles (AGVs) departures within the planning period, which is equivalent to minimizing the total system mileage, reducing the risk of track congestion, and improving the utilization rate of individual vehicle capacity.

[0042] (II) Constructing a mixed-integer linear programming scheduling model that couples entity flow and load flow

[0043] Based on the aforementioned constraints and optimization objectives, a mixed-integer linear programming scheduling model coupling entity flow and load flow is constructed. Continuous time is modeled using discrete time steps to fully characterize the entire pattern of vehicle operation, cargo flow, and inventory changes. The model construction process involves defining input parameters, setting decision variables, determining the objective function, constructing the constraint system, and achieving dual-flow coupling.

[0044] 1. Define the set of models and input parameters.

[0045] The model's input parameters are divided into three categories: set parameters, automated guided vehicle (AGV) operation time parameters, and material delivery business parameters, which are defined as follows: (1) Set parameters The discretized time step set T = {1, 2, ..., T} max}, where T max The total number of time steps in the planning period is given, and all scheduling decisions are executed on discrete time steps. The set of workbench nodes is n={1,...,i,...,N}, where i is the workbench number, N is the total number of workbench, and the replenishment point is denoted as node 0. The set of unloaded goods in a single operation, K={1,2,3}, corresponds to the number of fully loaded shelves that a single automated guided vehicle can unload at a single workbench.

[0046] (2) Parameters of the travel time of the automated guided vehicle

[0047] The total quantity V of materials in the automated guided vehicle in the flexible manufacturing material distribution system total ; After the automated guided vehicle completes its mission and returns to the replenishment point, it will be reloaded with goods and ready for offline preparation time (t) before going back online. refill ; The minimum safe time interval H for departure is the minimum time step interval between the departure of two automated guided vehicles. The number of steps required to travel at a constant speed from the replenishment point to the first workbench ; The number of time steps required to travel at a constant speed from the i-th workbench to the (i+1)-th workbench. where i∈{1,2,...,N-1}; The number of time steps required to travel at a constant speed from the Nth workbench back to the replenishment point ; The total fixed time steps t required for an automated guided vehicle to stop and unload k goods u (k), where k is the number of fully loaded shelves in a single unloading, with values ​​of 1, 2, and 3.

[0048] (3) Material distribution business parameters

[0049] The amount of material consumed by the i-th workstation at each time step is D. i ; The precise quantity Q of materials contained in a single fully loaded shelf corresponding to the i-th workbench. i ; The material inventory I of the i-th workstation at the initial time (t=0) i,0 ; Maximum material inventory capacity of the i-th workbench .

[0050] 2. Set up three types of decision variables

[0051] The model categorizes decision variables into three types, corresponding to vehicle physical trajectory, system resource status, and cargo load flow, as detailed below: (1) Vehicle spatiotemporal trajectory 0-1 discrete variables This type of variable is a 0-1 binary variable, used to track the spatial position and operation of the automated guided vehicle time-by-time: Departure state variable u t : A 0-1 variable, indicating whether an automated guided vehicle carrying goods is dispatched from the replenishment point at time t; 1 if yes, 0 otherwise. Return the replenishment point state variable w t : 0-1 variable, indicating whether at time t an automated guided vehicle that has completed its task returns to the replenishment point and begins preparation; 1 if yes, 0 otherwise. The state variable p of the workbench is passed straight through. i,t : A 0-1 variable, indicating whether an automated guided vehicle arrives at the i-th workstation and passes straight through at time t; 1 if yes, 0 otherwise. Docked and unloaded k cargo state variables : A 0-1 variable, indicating whether at time t an automated guided vehicle arrives at the i-th workbench and begins to dock and unload k goods; 1 if yes, 0 otherwise.

[0052] (2) Continuous variables of resource system state

[0053] The continuous variables in the resource system state class are non-negative continuous variables, used to characterize the dynamic state of global vehicle resources and workstation inventory: Number A of automated guided vehicles available at the replenishment point t : A continuous variable, representing the number of automated guided vehicles (AGVs) in an available standby state at the replenishment point at time t, satisfying A t ≥0; Material inventory I i,t : A continuous variable, representing the remaining inventory of materials at the i-th workstation at time t, satisfying I i,t ≥0.

[0054] (3) Continuous variables of shelf load flow

[0055] The continuous variables in the shelf load flow category are non-negative continuous variables, used to track changes in the amount of goods carried during the entire journey of the automated guided vehicle (AGV). Number of remaining goods upon arrival at the workbench : A continuous variable, representing the remaining fully loaded cargo carried by the automated guided vehicle arriving at the i-th workstation at time t, satisfying the following condition: ; Number of remaining goods after passing the workbench in a straight line : A continuous variable, representing the number of remaining goods carried by the automated guided vehicle (AGV) that travels straight past the i-th workstation at time t, satisfying the following condition: ; Number of remaining cargo after unloading : A continuous variable, representing the number of remaining goods on the automated guided vehicle after it stops at the i-th workbench and unloads k goods at time t, satisfying the following condition: .

[0056] 3. Determine the objective function

[0057] The model takes minimizing the total number of automated guided vehicle (AGV) departures within the planning period as its core objective function, expressed as: ; In the formula, Z represents the total number of departures of the automated guided vehicles within the planning period, and min represents the minimum value operation.

[0058] 4. Construct a five-category constraint system

[0059] The model construction covers a complete constraint system encompassing vehicle resource circulation, online traffic conservation, shelf load flow, single-track implicit conflict prevention, and dynamic inventory balancing. The specific content and expressions of various constraints are as follows: (1) Vehicle resource cycle constraint, used to regulate the resource conservation relationship of the automated guided vehicle throughout its entire life cycle from departure and delivery to return to the replenishment point for preparation, specifically includes four sub-constraints: ① Vehicle return constraint, the expression is: ; In the formula, express At any given moment, if an automated guided vehicle arrives at workbench N and moves straight through, 1 indicates yes, and 0 indicates no. This indicates whether an automated guided vehicle (AGV) arrives at workbench N at time t and begins to dock and unload k goods; 1 indicates yes and 0 indicates no. It indicates any one of the items.

[0060] The physical meaning of this constraint is: after the automated guided vehicle leaves the last workbench and travels for the corresponding duration, it must return to the replenishment point and begin preparation. The return status at time t corresponds one-to-one with the straight-ahead and parking status of the last workbench at the corresponding time.

[0061] ② Vehicle balance constraints can be used, and the expression is: ; In the formula, A t-1 This indicates the number of automated guided vehicles (AGVs) that are available and ready to go at the replenishment point at time t-1. express At any given time, 1 indicates whether an automated guided vehicle (AGV) that has completed its mission returns to the replenishment point and begins preparation; 0 indicates otherwise.

[0062] The physical meaning of this constraint is: the number of automated guided vehicles available at the replenishment point follows a dynamic update rule. The number available at the current moment is equal to the number available at the previous moment minus the number of vehicles dispatched this time, plus the number of vehicles that have been re-added to the standby pool after completing offline preparation.

[0063] ③ The departure quantity limit constraint is expressed as: ; The physical meaning of this constraint is: the number of vehicles dispatched at any given time must not exceed the number of available automated guided vehicles (AGVs) that are in standby status at the current replenishment point, so as to ensure that vehicle resources are not overused.

[0064] ④ Initial condition constraints, the expression is: ; In the formula, A0 represents the number of automated guided vehicles (AGVs) that are in an available standby state at the initial replenishment point.

[0065] The physical meaning of this constraint is: at the initial moment, all automated guided vehicles are in a standby state at the replenishment point.

[0066] (2) Online flow conservation constraint, used to ensure that the spatiotemporal trajectory of the automated guided vehicle on the circular track is continuous and unique, specifically including two sub-constraints: ① The first node enters the conservation constraint, the expression is: ; In the formula, p 1,t This indicates whether an automated guided vehicle (AGV) arrives at the first workbench and passes through it at time t; 1 indicates yes, and 0 indicates no. This indicates whether at time t, an automated guided vehicle (AGV) has arrived at the first workbench and started to dock and unload k goods; 1 indicates yes, 0 indicates no. express Whether an automated guided vehicle (AGV) carrying goods has been dispatched from the replenishment point at any given time is indicated by 1 if yes and 0 if no.

[0067] The physical meaning of this constraint is: the automated guided vehicle (AGV) dispatched from the replenishment point, after a certain travel time... Afterwards, it will inevitably reach the first workstation. At time t, the total number of arrivals (straight-through + stopping) at the first workstation is exactly equal to the number of departures at the corresponding time.

[0068] ② The conservation constraint of the cascaded workbench is expressed as: ; In the formula, p i,t This indicates whether an automated guided vehicle (AGV) arrives at the i-th workbench and passes straight through at time t; 1 indicates yes, and 0 indicates no. This indicates whether at time t an automated guided vehicle (AGV) arrives at the i-th workbench and begins to dock and unload k goods; 1 indicates yes, 0 indicates no. The time steps required to travel at a constant speed from the (i-1)th workbench to the ith workbench; express At any given moment, if an automated guided vehicle arrives at the (i-1)th workstation and proceeds straight through, the value is 1; otherwise, it is 0. express The value is 1 if an automated guided vehicle (AGV) arrives at the (i-1)th workbench and begins to dock and unload k goods at any given time, and 0 if not.

[0069] The physical meaning of this constraint is that the number of automated guided vehicles (AGVs) arriving at the i-th workstation is exactly equal to the number of AGVs leaving the (i-1)-th workstation, thus realizing the continuous node-by-node transfer of AGVs along the circular track.

[0070] (3) Shelf load flow constraint, used to track changes in the quantity of goods carried during the movement of the automated guided vehicle, and to match the correspondence between the load flow and the physical operation actions, specifically includes four sub-constraints: ① Initial departure capacity constraint, the expression is: ; In the formula, Let t be the remaining fully loaded cargo carried by the automated guided vehicle that arrives at the first workbench at time t.

[0071] The physical meaning of this constraint is: the number of fully loaded racks that a single automated guided vehicle can load in a single trip shall not exceed 3, and the maximum load capacity when it arrives at the first workbench is strictly bound to the departure status.

[0072] ② Load propagation constraints across nodes, the expression is: ; In the formula, Let t be the remaining fully loaded cargo carried by the automated guided vehicle that arrives at the i-th workstation at time t; for The number of remaining goods carried by the automated guided vehicle that passes through the (i-1)th workbench at any given time; for The number of remaining goods on the automated guided vehicle after stopping at the (i-1)th workbench and unloading k goods.

[0073] The physical meaning of this constraint is: the cargo load arriving at the i-th workstation is exactly equal to the cargo load leaving the (i-1)-th workstation, thus realizing the continuous transfer of cargo load node by node with the automated guided vehicle.

[0074] ③ Unload the deduction constraint, the expression is: ; In the formula, Let t be the number of remaining goods carried by the automated guided vehicle that passes the i-th workbench in a straight line at time t; Let t be the number of goods remaining on the automated guided vehicle after the i-th workbench is stopped and k goods are unloaded at time t.

[0075] The physical meaning of this constraint is: the remaining cargo capacity when the automated guided vehicle leaves the workbench is equal to the cargo capacity at arrival minus the amount of cargo unloaded during this stop, accurately depicting the change in cargo capacity during the unloading process.

[0076] ④ Physical capacity binding constraints include straight-through scenario binding constraints and docking / unloading scenario binding constraints, with the following expressions: ; ; This constraint binds the upper bound of the continuous load variable to a 0-1 state variable, realizing the logical association between the existence of the automated guided vehicle (AGV) entity and the load capacity: when no AGV arrives at the corresponding workstation, the load capacity of the corresponding workstation is forcibly limited to zero, eliminating the logical loophole of having a load of goods even when there is no physical vehicle.

[0077] (4) Implicit anti-collision constraints for monorails are used to avoid track rear-end collisions and platform collisions and congestion at the planning level. Specifically, they include two sub-constraints: ① Departure safety interval constraint, the expression is: ; In the formula, For the time step summation variable; for Whether an automated guided vehicle (AGV) carrying goods has been dispatched from the replenishment point at any given time is indicated by 1 if yes and 0 if no.

[0078] The physical meaning of this constraint is: at most one automated guided vehicle can be dispatched within any consecutive H time steps, thus avoiding rear-end collisions on the track from the source.

[0079] ② The physical mutual exclusion constraint of the workbench is expressed as: ; In the formula, express The value at any given time is 1 if an automated guided vehicle (AGV) arrives at the i-th workbench and begins to dock and unload k goods, and 0 if not.

[0080] The physical meaning of this constraint is: at any given time, only one automated guided vehicle can stop or pass through any workbench, and the unloading operation must be fully covered throughout the entire time period to avoid collisions and congestion at the workbench.

[0081] (5) Dynamic inventory balance constraints are used to maintain a reasonable range of material inventory at each workstation, ensuring zero supply disruptions and no inventory overflows. Specifically, they include: ① Inventory dynamic update constraint, specifically includes two sub-constraints: ; In the formula, I i,t-1 This represents the remaining inventory of materials at the i-th workbench at time t-1; express The value at any given time is 1 if an automated guided vehicle (AGV) arrives at the i-th workbench and begins to dock and unload k goods, and 0 if not.

[0082] The physical meaning of this constraint is: the workbench material inventory follows a dynamic update rule. The current inventory equals the previous inventory minus the consumption per unit time, plus the amount of material put into storage after unloading. The material is put into storage instantly after unloading.

[0083] ② Inventory boundary constraints, the expression is: ; The physical meaning of this constraint is: the material inventory of each workbench is always greater than or equal to zero and does not exceed the maximum inventory capacity, ensuring that the production line has no risk of supply interruption or production stoppage and no risk of material overstock.

[0084] 5. Logical coupling mechanism between entity flow and load flow

[0085] The logical coupling between the physical flow and the load flow is achieved through the above-mentioned physical capacity forced binding constraint: using the vehicle spatiotemporal trajectory class 0-1 discrete variable as the upper bound control factor of the shelf load flow class continuous variable, the physical existence state of the automated guided vehicle is strictly bound to the quantity of goods carried.

[0086] For the straight-through scenario: when the straight-through state variable is 0 (no vehicles going straight), the upper bound of the corresponding straight-through load variable is 0, and the load is forced to be 0; when the state variable is 1 (vehicles going straight), the load variable can be up to 3, which meets the single vehicle capacity limit.

[0087] For docking and unloading scenarios: when the docking and unloading status variable is 0 (no vehicle docking), the upper bound of the load variable after docking is 0, and the load is forced to be 0; when the status variable is 1 (a vehicle docks and unloads k goods), the load variable can be up to 3-k, which is consistent with the remaining capacity after unloading.

[0088] Through the above binding mechanism, when no automated guided vehicle arrives, the load of the corresponding workstation is forcibly limited to zero, eliminating the logical loophole of having cargo loads even without physical vehicles, and ensuring a strict one-to-one correspondence between physical flow and load flow.

[0089] (III) Solving the model using a just-in-time driven greedy stochastic adaptive search algorithm

[0090] To address the low efficiency of solving mixed-integer linear programming models in large-scale scenarios, a just-in-time driven greedy stochastic adaptive search algorithm is adopted for efficient solution. Guided by the principle of zero-disruption on-time supply, this algorithm employs a "firefighting scheduling" approach, using global supply disruption warnings as a starting point to deduce departure timing in reverse. A restricted candidate set mechanism balances the quality and diversity of solutions, and after multiple rounds of iterative selection, the optimal scheduling scheme is output.

[0091] 1. Overall Algorithm Framework

[0092] The algorithm adopts a multi-starting-point iterative architecture, and the overall execution flow is as follows: (1) Initialization: Set the maximum number of iterations of the algorithm, and initialize the optimal scheduling scheme and the optimal end-of-period inventory balance score; (2) Constructing a feasible solution: Execute the construction heuristic process to generate a feasible scheduling scheme that satisfies all constraints; (3) Solution evaluation: Perform a dual-objective evaluation on the generated feasible scheduling scheme and calculate its total number of departures and the balance score of ending inventory; (4) Optimal solution update: If the current feasible solution is better than the current optimal solution, then update the optimal scheduling scheme and the corresponding score; (5) Iteration termination judgment: If the number of iterations reaches the preset maximum number of iterations, the final optimal automated guided vehicle scheduling scheme is output and sent to the flexible manufacturing material distribution system; otherwise, return to step (2) to continue iterating.

[0093] 2. Spatiotemporal panoramic tracker

[0094] The spatiotemporal state panoramic tracker is the core data structure of the algorithm, responsible for tracking all state information of the system throughout its entire lifecycle in real time. It provides state query and state update interfaces for constructing heuristic processes and includes four types of states: (1) Inventory Status: Record the material inventory of each workstation at each time point. Initially, it is assumed that no materials are delivered, and the inventory is naturally depleted over time, i.e., the inventory of the i-th workstation at time t is inv[i][t] = I. i,0 - D i •t; When materials are unloaded, the inventory level is updated synchronously at the corresponding time and for all subsequent times.

[0095] (2) Workbench Mutual Exclusion State: Record whether each workbench is occupied at each time. The occupancy state of the i-th workbench at time t is mutex[i][t]=true. True means that the workbench is occupied at that time, and other automated guided vehicles cannot stop or pass through; otherwise, false means that the workbench is available at that time. Initially, all times are in the available state.

[0096] (3) Departure status: Record whether an automated guided vehicle (AGV) has departed at each time. The departure status at time t is dispatch[t]=true. True means that the vehicle has departed at that time, and other AGVs cannot depart at that time or within the minimum safe interval before and after. Initially, all times are in the non-departure state.

[0097] (4) Automated Guided Vehicle (AGV) Pool Status: Records the number of AGVs available in the system at each time point. Initially, the available number at all times is the total physical number of AGVs in the system, V. total When an automated guided vehicle (AGV) departs, deduct the available number of vehicles from the time of departure until the return and preparation are completed.

[0098] 3. Construct a heuristic core execution process

[0099] The heuristic process is constructed with "firefighting scheduling" as its core idea: continuously find the first workstation in the entire field that is about to run out of supply, reverse the calculation of the latest departure time, reserve a firefighting slot for the workstation within the legal departure window, and then fill the remaining slots through a restricted candidate set mechanism until all workstations are free from the risk of running out of supply throughout the entire scheduling cycle.

[0100] (1) The global supply interruption early warning monitoring traverses the inventory status trajectory of all workstations and finds the first time and workstation that meets the condition that the inventory is less than 0. This is the first time and workstation that will experience a supply interruption. If no supply interruption point is found after traversing the entire scheduling cycle, it means that the current scheduling scheme has met the on-time requirements of all workstations, and a feasible solution is output directly.

[0101] (2) The latest departure time is derived in reverse order to maximize the loading efficiency of the automated guided vehicles (AGVs) and to depart as late as possible to collect more urgent requests and fill more slots. Based on the moment of supply interruption, the latest departure time is derived in reverse order, and the calculation formula is as follows: ; In the formula, t d The latest departure time; stockout t The first detected supply interruption globally; stockout w s is the workstation number where the supply disruption occurred; TAU0 is the time step required for the replenishment point to travel to the first workstation; TAU I This represents the unit travel time steps between adjacent workstations.

[0102] Later than the latest departure time t d If the shipment is not made, the materials will not be delivered before the supply is cut off, and the production line will inevitably be disrupted.

[0103] (3) Search for legal departure windows

[0104] Starting from the latest departure time, search backwards to find the first departure time that satisfies all constraints. For each candidate departure time, first reserve one emergency response slot for the supply interruption workbench to ensure basic emergency response needs, then verify the validity of the single slot's travel. If valid, proceed to the slot filling stage; if invalid, advance the departure time by one time step and continue the search.

[0105] (4) Greedy random slot filling of restricted candidate set

[0106] The algorithm balances greediness and randomness through a restricted candidate set mechanism. Greediness ensures the quality of the solution, while randomness increases search diversity and avoids getting trapped in local optima. The specific filling steps are as follows: ① Construct a candidate workbench set: Traverse all workbench, tentatively add 1 unloading slot to each workbench and verify the validity of the process, add the workbench that passes the verification to the candidate set, and calculate the supply interruption urgency score of the corresponding workbench. The earlier the supply interruption occurs, the higher the urgency. ② Candidate set sorting: Sort the workbenches in the candidate set from highest to lowest according to the urgency of the supply disruption; ③ Random selection for filling: Randomly select one workbench from the top 3 workbench in the sorted candidate set and add 1 unloading slot to it; ④ Repeat the above steps until all remaining rack slots of the automated guided vehicle are filled, or there are no valid candidate workbenches.

[0107] (5) Spatiotemporal resource locking and status updating

[0108] Once a valid itinerary is identified, all spatiotemporal resources corresponding to that itinerary are locked, and the entire state of the spatiotemporal state panoramic tracker is updated, specifically including: ① Lock departure status: Mark the corresponding departure time as departed to prevent other automated guided vehicles from departing at that time and within the minimum safe interval before and after; ②Deduct automated guided vehicle (AGV) pool resources: Decrease the number of available AGVs for the AGV from the departure time to the return preparation completion time by 1; ③ Lock mutually exclusive resources at workstations: Mark the time period when the automated guided vehicle stops or passes through each workstation as occupied to prevent other automated guided vehicles from occupying the corresponding workstation during the same time period; ④ Update inventory status: Add the amount of material unloaded by the automated guided vehicle to the inventory of the corresponding workstation from the arrival time to the end of the planned cycle.

[0109] 4. Multi-dimensional feasibility verification mechanism

[0110] All candidate routes must pass five rigorous feasibility checks to ensure that the generated scheduling plan meets all constraints. The five check dimensions are as follows: (1) Anti-rear-end collision check: Check whether there are already automated guided vehicles departing within the minimum safe interval before and after the candidate departure time. If there is a departure record, the trip is deemed invalid. (2) Spatiotemporal mutual exclusion verification: Simulate the driving process of the automated guided vehicle and check whether the stopping time or straight passage time of each work platform conflicts with the planned route. If there is spatiotemporal overlap, the route is deemed invalid. (3) Lifeline interception verification: Check whether the arrival time of the automated guided vehicle is later than the time of supply interruption of the workbench. If the arrival time is later than the time of supply interruption, the trip cannot avoid supply interruption and is directly judged as invalid. (4) Overload check: Calculate the maximum inventory of the workbench after unloading materials. If it exceeds the maximum material inventory capacity of the workbench, the trip is deemed invalid. (5) Vehicle pool verification: Check whether the system has enough available automated guided vehicles during the period from departure to return and preparation. If the number of available vehicles is insufficient, the trip is deemed invalid.

[0111] 5. Dual-objective evaluation and optimal solution update rules

[0112] The algorithm employs a dual-objective evaluation mechanism to evaluate the generated feasible solutions and select the better scheduling scheme. The evaluation rules are as follows: (1) First priority evaluation target: total number of departures. The fewer the total number of departures, the higher the priority of the scheduling scheme, which is consistent with the core optimization target of the model; (2) Second priority evaluation target: End-of-period inventory balance score. The end-of-period inventory balance score is the sum of the end-of-period inventory of all workstations. The higher the score, the more sufficient the inventory of each workstation is and the better the robustness of the system.

[0113] The conditions for updating the optimal scheduling scheme are: the total number of departures of the current feasible scheduling scheme is less than that of the current optimal scheme, or the total number of departures is the same but the ending inventory balance score is higher than that of the current optimal scheme.

[0114] III. Scheduling and Optimization System for Material Delivery of Multi-Load Automated Guided Vehicles on One-Way Loop Lines

[0115] This system is applied to a flexible manufacturing material distribution system consisting of a single replenishment point, multiple workbenches, and a one-way circular track. It is used to execute the aforementioned scheduling optimization method and is divided into three main functional modules, which work together to complete the entire process of scheduling and control from parameter configuration and model building to solution distribution.

[0116] (I) Scene Configuration and Constraint Definition Module

[0117] This module serves as the system's basic configuration entry point, used to solidify the rack-on-vehicle operation mechanism, and to input and verify all scheduling constraints and optimization objectives.

[0118] The module supports automatic parameter collection from the workshop production system and also supports manual input. Core configuration parameters include: physical operating parameters such as the maximum number of rack slots per automated guided vehicle (AGV), minimum safe departure interval, travel time for each route, and stop / unload duration; and production operating parameters such as material consumption rate at each workstation, material capacity per rack, initial inventory, and inventory limit. The module defaults to minimizing the total number of departures within the planning cycle as its core optimization objective, while also establishing boundary constraints of "zero supply disruptions, zero conflicts, and no inventory overload," providing standardized input for subsequent modeling.

[0119] (II) Mixed Integer Linear Programming Model Construction Module

[0120] This module is used to transform practical scheduling problems into solvable mathematical models and construct a mixed-integer linear programming scheduling model that couples entity flow and load flow.

[0121] Based on the input configuration parameters, the module automatically completes the set definition and sets three types of decision variables (vehicle trajectory 0-1 variables, resource status continuous variables, and load flow continuous variables), generates an objective function with minimizing the total number of departures as the core, and builds a complete constraint system covering vehicle resource circulation, online traffic conservation, shelf load flow, single-track implicit anti-conflict, and dynamic inventory balancing. At the same time, through the upper bound binding rules of vehicle trajectory 0-1 variables and load flow continuous variables, it realizes the logical coupling between entity flow and load flow, eliminates the logical loophole of "goods without vehicles", and ensures the rigor of the physical meaning of the model.

[0122] (III) Time-based driven greedy stochastic adaptive search solution module

[0123] This module is the core computation and output module of the system, used to efficiently solve mixed-integer linear programming models and issue the final solution.

[0124] The module incorporates a comprehensive spatiotemporal status tracking mechanism and employs a "firefighting" scheduling logic: first, it locates the first workstation experiencing a supply disruption through a global supply interruption warning, then reverse-engineers the latest departure time and reserves firefighting slots, and finally fills the remaining shelf slots using a greedy random mechanism with a limited candidate set; all candidate trips undergo five-dimensional verification: anti-rear-end collision, workstation mutual exclusion, lifeline, warehouse overload, and vehicle pool, and the corresponding spatiotemporal resources are locked after verification. The module constructs feasible solutions through multiple rounds of iteration, selecting the optimal solution based on a dual rule of "total number of departures first, and end-of-period inventory balance second," and finally distributes the optimal scheduling solution to the automated guided vehicle's onboard terminal, replenishment point operation terminal, and each workstation's production terminal, driving the workshop material distribution system to operate stably according to the plan.

[0125] IV. Implementation of the Optimal Scheduling Scheme

[0126] The optimal scheduling scheme generated by the above method is essentially a precise timetable for the departure and unloading of automated guided vehicles (AGVs), specifying the departure time of each AGV, the unloading stations along the route, and the unloading quantity at each station. The entire implementation process requires no complex manual intervention; it can be strictly executed according to the optimal scheme. The core logic is: departure at designated times, unloading at stations, cyclical replenishment, and conflict prevention from the planning level.

[0127] (a) Pre-shift preparation: Issuance of plan and pre-loading of materials

[0128] This step is the preparatory work before the start of the shift, ensuring that the vehicles, materials, workstations, and scheduling plan are aligned.

[0129] 1. The solution is synchronized to all terminals across the entire supply chain.

[0130] The scheduling system will synchronize the final optimal scheduling plan to three execution terminals: the operation terminal at the replenishment point, the material dashboard at each workstation, and the onboard control system of each automated guided vehicle (AGV). The core content of the plan includes: the departure time of each AGV throughout the day, the type / quantity of materials loaded on each vehicle, the unloading volume and estimated arrival time at each workstation.

[0131] 2. Replenishment points prepare goods and load them onto trucks in advance.

[0132] The replenishment clerk, following the loading list of the first shift's automated guided vehicle (AGV), pre-places the material boxes for the corresponding workstations into the three fixed shelf slots of the AGV. The shelves themselves are permanently fixed to the AGV, and only the material boxes are moved, complying with the "shelves never leave the vehicle" rule. At the same time, the AGV's battery level and driving system self-check are completed, and the vehicle is ready at the departure point.

[0133] 3. Confirm initial inventory at the workstation.

[0134] When each workstation starts work, the initial material inventory is checked to ensure it matches the initial inventory preset in the plan, so as to ensure that the timing of the first wave of replenishment is accurately matched with the consumption rhythm.

[0135] (ii) Punctual departure: Departure at fixed intervals to prevent rear-end collisions from the source.

[0136] 1. Automatic departure at the designated time

[0137] Upon arrival at the departure time specified in the plan, the automated guided vehicle automatically enters the circular track from the replenishment point and travels at a constant speed in a clockwise direction, without the need for manual driving throughout the entire process.

[0138] 2. Strictly adhere to safe intervals.

[0139] The departure times of any two automated guided vehicles (AGVs) strictly adhere to the minimum safe interval specified in the plan (e.g., an interval of 5 time steps). The following vehicle will not depart until the preceding vehicle has traveled a sufficient safe distance, thus preventing rear-end collisions on the track at the source. Overtaking and real-time avoidance are not required throughout the entire journey.

[0140] (iii) Loop delivery: stop immediately upon arrival at the station, and leave immediately after unloading.

[0141] The automated guided vehicle passes each workbench sequentially along a circular track, traveling along a preset trajectory throughout. The operating rules are simple and clear. Non-stop workstations: Automated guided vehicles pass through directly at a constant speed without stopping, without affecting the efficiency of track passage; Workstations requiring unloading: The automated guided vehicle slows down in advance and stops precisely, automatically completing the unloading of materials: transferring the material boxes in the corresponding slots to the material racks on the workbench, while the empty racks remain on the automated guided vehicle. Automatic workstation confirmation: The material system at the workstation automatically counts the quantity of goods received, updates the current inventory, and displays the next replenishment time on the Kanban board. Production line workers only need to consume materials during normal production; there is no need to actively request or expedite materials.

[0142] Because the scheduling plan has already calculated the time and space trajectories of all vehicles in advance, only one automated guided vehicle will stop or pass through any workstation at any time. There will never be a situation where two vehicles compete for a workstation or block the workstation. There is no congestion or collision risk throughout the process.

[0143] (iv) Return trip restocking: After completing a round trip, return to the checkpoint and quickly restock the next batch.

[0144] 1. After unloading all the planned materials, the automated guided vehicle continues to travel along the circular track and eventually returns to the replenishment point; 2. The replenishment clerk quickly loads the new material boxes into the empty shelf slots according to the loading list for the next shift, while simultaneously completing a quick vehicle inspection; 3. Once the vehicle is ready, it will wait at the replenishment point until the next departure time, at which point it will set off again to enter the loop and begin a new round of deliveries.

[0145] (v) 24-hour cyclical operation and dynamic fine-tuning

[0146] 1. Normal production runs according to schedule: Under normal production conditions, all automated guided vehicles (AGVs) strictly follow the scheduling plan to dispatch and deliver materials in a cyclical manner, ensuring that the materials at each workstation are always replenished as soon as they are used up, without accumulation or interruption of supply, thus achieving just-in-time delivery.

[0147] 2. Automatic adjustment for minor fluctuations: If minor production fluctuations occur (such as increased production speed at a certain workstation leading to faster material consumption), the scheduling system will increase the unloading volume of that workstation on the next shift's automated guided vehicle based on the current inventory, automatically fine-tuning the subsequent plan without interrupting the current production operation.

[0148] 3. Safety in Abnormal Situations: In the event of emergencies such as vehicle malfunction or workstation suspension, the system will automatically postpone the subsequent departure time, always ensuring that the two anti-conflict rules of "departure interval and workstation mutual exclusion" are in effect, and there will never be track congestion or vehicle collision.

[0149] (vi) Demonstration Plan

[0150] Assuming the workshop has 6 workbenches, the circular track takes 30 minutes to complete one rotation, each automated guided vehicle has 3 shelf slots, and each workbench consumes 1 box of materials every 30 minutes.

[0151] Optimal arrangement: 1 automated guided vehicle is dispatched every 10 minutes, and a total of 3 automated guided vehicles run in a loop; each automated guided vehicle is loaded with 3 boxes of materials at a time, and 1 box is delivered to each of the 3 workstations.

[0152] Actual operational results: Each workstation receives exactly one box of materials every 30 minutes, and replenishes it as soon as it is used up, with no empty runs or stockpiling throughout the process; there are always 3 vehicles evenly distributed on the track, preventing rear-end collisions and workstation blockages; the total number of departures is minimized, and the transportation capacity just meets the production needs, with no waste.

[0153] V. Case Analysis and Experimental Verification

[0154] To further verify the effectiveness and superiority of the above-mentioned material distribution scheduling optimization method for unidirectional loop multi-load automated guided vehicles, this embodiment designs multiple sets of simulation experiments for testing. The experiments use a loop track material distribution system in a flexible manufacturing workshop as a background. By setting up simulations of different scales, the method of this invention is compared and analyzed with traditional scheduling methods, verifying the performance advantages of the method of this invention from two dimensions: solution quality and solution efficiency.

[0155] (I) Simulation Experiment Design

[0156] 1. Experimental Environment and Parameters

[0157] This experiment was conducted on a Windows 11 64-bit operating system, and the simulation program was written in Python 3.9. The experimental computer hardware configuration was as follows: Intel(R) Core(TM) i7-12700 CPU, clock speed 2.10GHz, and 16GB of memory. To ensure the reliability of the experimental results, each example was run independently 20 times, and the average value was taken as the final result.

[0158] 2. System parameter settings

[0159] The experimental scenario is a typical flexible manufacturing workshop unidirectional circular track material delivery system. The system consists of one replenishment point, six workbenches, and a unidirectional clockwise circular track. Each automated guided vehicle (AGV) is equipped with three rack slots and adopts a rack-on-vehicle operation mechanism, similar to the aforementioned embodiment. Figure 1 The system architecture is completely identical. The core parameter settings of the system are shown in Table 1.

[0160] Table 1 System Core Parameter Settings

[0161] The operational parameters for each workstation, such as material consumption rate, single shelf material capacity, initial inventory, and maximum inventory capacity, are shown in Table 2.

[0162] Table 2. Parameter settings for the workbench

[0163] 3. Case Design

[0164] To comprehensively verify the performance of the method of this invention at different scales, multiple sets of computational examples were designed to construct scheduling problems of varying complexity by changing the number of automated guided vehicles and the intensity of material demand. The specific computational examples are designed as follows: (1) Case naming rules: Let the case name be α-β, where α represents the number of automated guided vehicles in the system and β represents the material demand intensity level (L for low demand, M for medium demand, and H for high demand). In the low demand scenario, the material consumption rate of each workbench is 0.8 times the baseline value, in the medium demand scenario it is 1.0 times the baseline value, and in the high demand scenario it is 1.2 times the baseline value.

[0165] (2) Setting the number of automated guided vehicles: Set four vehicle scales of 3, 4, 5 and 6 vehicles respectively to examine the impact of vehicle resource sufficiency on scheduling performance.

[0166] (3) Demand intensity setting: Set three demand intensity levels: low, medium and high, and examine the impact of material delivery pressure on the scheduling scheme.

[0167] By combining the above parameters, a total of 4×3=12 calculation examples were generated, covering a variety of typical scenarios from vehicle shortage to vehicle abundance, and from moderate demand to urgent demand.

[0168] (4) Algorithm comparison: Two representative scheduling methods were selected as comparison benchmarks: ① Traditional greedy scheduling algorithm: Delivery is arranged sequentially according to the order of the workstations, and a first-come, first-served strategy is adopted. Each dispatch only replenishes the most urgent workstation at the moment, without overall coordination of transportation capacity.

[0169] ②The GRASP algorithm of this invention: adopts a just-in-time driven greedy random adaptive search algorithm, takes the global supply interruption warning as the trigger, reversely derives the latest departure time, fills the remaining capacity through a restricted candidate set greedy random mechanism, and performs global optimization with the goal of minimizing the total number of departures.

[0170] (II) Experimental Results and Analysis

[0171] 1. Comparative Analysis of Objective Function Values

[0172] Table 3 presents a comparison of the objective function values ​​(i.e., total number of departures) of the traditional scheduling algorithm and the method of this invention under different computational examples.

[0173] Table 3. Comparison of Objective Function Values ​​between Traditional Scheduling Algorithm and GRASP Algorithm

[0174] As shown in Table 3, in all 12 examples, the total number of departures using the method of this invention is significantly lower than that of traditional scheduling algorithms, with an average optimization rate of 15.58%. Detailed analysis follows: (1) As the number of automated guided vehicles increases, the total number of departures for both algorithms decreases. This is because the more abundant the vehicle resources, the higher the scheduling flexibility, and the better the conditions for capacity integration and batch delivery. When the number of vehicles increases from 3 to 6, the average number of departures for the method of this invention decreases from 30 to 25, and the optimization effect remains stable.

[0175] (2) As demand intensity increases, the total number of departures shows a linear growth trend. In high-demand scenarios, the material consumption rate is faster, requiring more frequent replenishment. However, the method of this invention maintains an optimization ratio of approximately 15% to 17% through global coordination of transportation capacity allocation, indicating that the method of this invention has a stable optimization effect under different demand pressures.

[0176] (3) The optimization ratio of the method of the present invention increases slightly with the increase of the number of vehicles. When the number of vehicles is 3, the average optimization ratio is 14.29%; when the number of vehicles increases to 6, the average optimization ratio increases to 16.67%. This shows that the more abundant the vehicle resources, the greater the overall optimization space of the method of the present invention, and the more it can give full play to the advantages of multi-load integrated delivery.

[0177] 2. Comparison and analysis of solution time

[0178] Table 4 presents a comparison of the solution time of the traditional scheduling algorithm and the method of this invention under different computational examples.

[0179] Table 4. Comparison of solution time between traditional scheduling algorithms and GRASP algorithm

[0180] As shown in Table 4, the method of this invention also has a significant advantage in terms of solution efficiency, with an average efficiency improvement of 37.61%. A detailed analysis follows: (1) As the problem size increases (the number of vehicles increases and the demand intensity increases), the solution time of both algorithms shows an upward trend. The solution time of the traditional scheduling algorithm increases faster, while the growth of the GRASP algorithm is relatively slow, indicating that the GRASP algorithm has better scale adaptability.

[0181] (2) The GRASP algorithm adopts a just-in-time driven greedy random adaptive search strategy, and performs targeted search with the supply interruption warning as the guide, avoiding a large number of invalid enumerations and trials in traditional algorithms. Therefore, it can find a high-quality solution in a shorter time.

[0182] (3) In large-scale examples (such as 6-H), the solution time advantage of the GRASP algorithm is more obvious, and the solution efficiency improvement remains above 37%, indicating that the method of the present invention has higher computational efficiency when dealing with complex scheduling problems and is more suitable for large-scale applications in actual industrial scenarios.

[0183] 3. System Performance Index Analysis

[0184] To further evaluate the overall performance of the scheduling scheme, this section compares and analyzes the scheduling results of the two algorithms from three dimensions: average inventory level, inventory turnover rate, and vehicle utilization rate. The results are shown in Table 5.

[0185] Table 5 System evaluation indicators of AGV task scheduling method under different computational examples

[0186] As can be seen from Table 5, the method of the present invention outperforms the traditional scheduling algorithm in all system performance indicators: (1) Average inventory level: The average inventory level of the GRASP algorithm is about 12% to 15% lower than that of the traditional algorithm. This is because the method of the present invention adopts a just-in-time delivery strategy, which can more accurately match the material consumption rhythm, reduce unnecessary inventory backlog, and reduce the inventory space and capital cost of the workstation.

[0187] (2) Inventory turnover rate: The inventory turnover rate of the GRASP algorithm is about 14% to 16% higher than that of the traditional algorithm. A higher inventory turnover rate means smoother material flow and shorter capital occupation period, which reflects the core advantages of just-in-time delivery.

[0188] (3) Vehicle utilization rate: The vehicle utilization rate of the GRASP algorithm is about 15%~20% higher than that of the traditional algorithm. By coordinating the overall transportation capacity allocation, the delivery needs of multiple workstations are integrated into the same trip, which effectively improves the single vehicle capacity utilization rate and reduces ineffective empty runs and redundant departures.

[0189] 4. Algorithm Convergence Analysis

[0190] To verify the convergence performance of the GRASP algorithm, a representative medium-sized example (4-M) was selected for iterative process analysis. The curves showing the change of the objective function value during the algorithm's iteration process are shown below. Figure 2 As shown.

[0191] The convergence curve shows that: (1) The GRASP algorithm has a fast convergence speed. The objective function value decreases rapidly in the first 50 iterations, and it can quickly find a better solution. (2) When the iteration reaches about 150 times, the objective function value tends to stabilize and the algorithm basically converges, indicating that the GRASP algorithm has good convergence. (3) The convergence curve is smooth overall and there is no obvious oscillation phenomenon, indicating that the greedy random adaptive search mechanism can effectively balance exploration and utilization and avoid getting trapped in local optima.

[0192] (III) Detailed Analysis of Typical Examples

[0193] To more intuitively demonstrate the scheduling effect of the method of this invention, a 4-M case (4 automated guided vehicles, medium demand intensity) is selected as a typical case for detailed analysis. Table 6 provides a summary comparison of the scheduling schemes of the two algorithms under the 4-M case.

[0194] Table 6. Summary Comparison of Scheduling Schemes for the Two Algorithms in the 4-M Example

[0195] As can be seen from Table 6, the method of the present invention achieves the following optimization effects through global overall scheduling: (1) Significant effect of capacity integration: The proportion of fully loaded trips increased from 43.75% to 70.37%, and the average number of delivery workstations per vehicle increased from 2.3 to 2.9, giving full play to the capacity advantages of multi-load automated guided vehicles; (2) More precise inventory management: the average inventory level decreased by 13.0%, the inventory fluctuation range decreased by 33.9%, and the minimum inventory level was higher, indicating that the method of the present invention can better ensure the security of material supply while reducing inventory backlog. (3) Improved overall operating efficiency: The total driving time was reduced by 15.6%, which correspondingly reduced vehicle energy consumption and equipment wear and tear, resulting in significant economic benefits.

[0196] In summary, through 12 sets of simulation examples of different scales, it can be verified that the material delivery scheduling optimization method for unidirectional loop multi-load automated guided vehicles proposed in this invention is superior to traditional scheduling algorithms in terms of solution quality, solution efficiency, and overall system performance. Under the premise of strictly satisfying the loop passage constraints and production inventory constraints, it can effectively reduce the total number of departures, improve the capacity utilization rate, and reduce the workstation inventory level, thereby achieving a synergistic improvement in operational safety, delivery timeliness, and capacity economy.

[0197] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A scheduling optimization method for material delivery of multi-load automated guided vehicles (AGVs) on a unidirectional loop line, characterized in that, The flexible manufacturing material delivery system, which consists of a single replenishment point, multiple workbenches, and a unidirectional circular track, includes the following steps: The system adopts a rack-based operation mechanism with scheduling constraints and optimization objectives. The scheduling constraints include: the number of goods loaded by a single automated guided vehicle (AGV) at one time does not exceed the preset rack slot limit; the departure time interval between any two AGVs is not less than the minimum safety interval; only one AGV is allowed to stop or pass through any workstation at any given time; and the material inventory of each workstation is always greater than or equal to zero and does not exceed the maximum inventory capacity of the corresponding workstation. The optimization objective is to minimize the total number of AGV departures within the planning period. Based on scheduling constraints and optimization objectives, and taking the minimization of the total number of departures under each constraint as the objective function, a mixed-integer linear programming scheduling model coupling entity flow and load flow is constructed. A just-in-time driven greedy stochastic adaptive search algorithm is used to solve the mixed-integer linear programming scheduling model. The latest departure time is derived in reverse from the global supply interruption warning. The remaining capacity of the automated guided vehicles is filled by a greedy stochastic mechanism of a restricted candidate set. After multiple rounds of iterative screening, the optimal automated guided vehicle scheduling scheme that satisfies all constraints is output and sent to the flexible manufacturing material distribution system for the allocation of automated guided vehicles.

2. The scheduling optimization method for material distribution of unidirectional loop multi-load automated guided vehicles according to claim 1, characterized in that, The process of constructing a mixed-integer linear programming scheduling model is as follows: Define the set parameters, automated guided vehicle (AGV) driving time parameters, and material delivery business parameters corresponding to the model as the input basis for the model; Three types of decision variables are set up: vehicle spatiotemporal trajectory type 0-1 discrete variables used to track the spatiotemporal operation trajectory of automated guided vehicles; resource system state type continuous variables used to characterize system resources and inventory status; and shelf load flow type continuous variables used to track changes in the quantity of goods carried by automated guided vehicles. The objective function is determined to minimize the total number of automated guided vehicle (AGV) departures within the planning period. A constraint system is constructed that covers vehicle resource cycle constraints, online traffic conservation constraints, shelf load flow constraints, single-track implicit anti-conflict constraints, and dynamic inventory balance constraints. The logical coupling of entity flow and load flow is achieved by binding the upper bound of the vehicle spatiotemporal trajectory class 0-1 discrete variables with the shelf load flow class continuous variables.

3. The scheduling optimization method for material distribution of unidirectional loop multi-load automated guided vehicles according to claim 2, characterized in that, The set parameters include: the discretized time step set T = {1, 2, ..., T} max }, where T max The planning cycle length is defined by the following: the set of workbench nodes is n={1,...,i,...,N}, where i is the workbench number and N is the total number of workbench nodes; the replenishment point is denoted as node 0; the set of unloaded goods in a single operation is K={1,2,3}. The parameters for the Automated Guided Vehicle (AGV) operation time include: the total physical number V of AGVs in the flexible manufacturing material delivery system. total Offline preparation time t after the automated guided vehicle returns to the replenishment point refill Minimum safe time interval H between departures, the time it takes for the vehicle to travel from the replenishment point to the first workbench. The number of time steps required to travel at a constant speed from the i-th workbench to the (i+1)-th workbench. The number of steps required for the vehicle to travel at a constant speed from workbench N back to the replenishment point. The total fixed time steps t required for an automated guided vehicle to stop and unload k goods u (k), where k is the number of fully loaded shelves in a single unloading, and the value is 1, 2, or 3; Material distribution operation parameters include: the material consumption D per unit time for the i-th workstation. i The precise quantity Q of materials contained in a single fully loaded shelf corresponding to the i-th workbench. i The initial inventory I of the i-th workbench i,0 The maximum material inventory capacity of the i-th workbench .

4. The scheduling optimization method for material distribution of unidirectional loop multi-load automated guided vehicles according to claim 3, characterized in that, The vehicle spatiotemporal trajectory class of 0-1 discrete variables includes: the departure state variable u at time t. t This indicates whether an automated guided vehicle (AGV) carrying goods has been dispatched from the replenishment point at time t; 1 indicates yes, 0 indicates no. The state variable w of the replenishment point is returned at time t. t The value p indicates whether an automated guided vehicle (AGV) that has completed its task returns to the replenishment point and begins preparation at time t; 1 indicates yes, 0 indicates no. The state variable p represents the vehicle traveling straight past the workbench at time t. i,t The variable t represents whether an automated guided vehicle (AGV) arrives at the i-th workbench and proceeds straight through at time t; 1 indicates yes, 0 indicates no. The variable t also represents the state variables for unloading k goods at time t. , indicates whether at time t there is an automated guided vehicle that arrives at the i-th workbench and starts to dock and unload k goods, 1 for yes and 0 for no; Continuous variables in the resource system state class include: the number of automated guided vehicles (AGVs) available at the replenishment point at time t (A). t , represents the number of automated guided vehicles (AGVs) in an available standby state at the replenishment point at time t; I represents the material inventory at time t. i,t , represents the remaining inventory of materials at time t for the i-th workbench; Continuous variables in the shelf load flow class include: the number of remaining goods when they arrive at the workbench. , representing the remaining fully loaded cargo carried by the automated guided vehicle (AGV) that arrives at the i-th workbench at time t; and the remaining cargo after passing the workbench. , representing the remaining cargo carried by the automated guided vehicle (AGV) that travels straight past the i-th workbench at time t; and the remaining cargo after stopping and unloading. , represents the number of goods remaining on the automated guided vehicle after stopping at the i-th workbench and unloading k goods at time t.

5. The scheduling optimization method for material distribution of unidirectional loop multi-load automated guided vehicles according to claim 4, characterized in that, The specific contents of each type of constraint in the constraint system are as follows: Vehicle resource cycle constraints are used to regulate the resource conservation relationship throughout the entire lifecycle of automated guided vehicles (AGVs), from departure and delivery to return to the replenishment point for preparation. Specifically, this includes: The vehicle return constraint is expressed as follows: ； In the formula, express At any given moment, if an automated guided vehicle arrives at workbench N and moves straight through, 1 indicates yes, and 0 indicates no. This indicates whether an automated guided vehicle (AGV) arrives at workbench N at time t and begins to dock and unload k goods; 1 indicates yes and 0 indicates no. Indicates any one item; Vehicle balance constraints can be used, expressed as: ； In the formula, A t-1 This indicates the number of automated guided vehicles (AGVs) that are available and ready to go at the replenishment point at time t-1. express If at any given moment an automated guided vehicle (AGV) that has completed its mission returns to the replenishment point and begins preparation, 1 indicates yes, and 0 indicates no. The departure quantity limit constraint is expressed as follows: ； The initial condition constraints are expressed as follows: ； In the formula, A0 represents the number of automated guided vehicles (AGVs) that are in an available standby state at the initial replenishment point; Online traffic conservation constraints are used to ensure the continuity and uniqueness of the spatiotemporal trajectory of the automated guided vehicle on the circular track, specifically including: The first node enters the conservation constraint, expressed as: ； In the formula, p 1,t This indicates whether an automated guided vehicle (AGV) arrives at the first workbench and passes through it at time t; 1 indicates yes, and 0 indicates no. This indicates whether at time t, an automated guided vehicle (AGV) has arrived at the first workbench and started to dock and unload k goods; 1 indicates yes, 0 indicates no. express Whether an automated guided vehicle (AGV) carrying goods has been dispatched from the replenishment point at any given time is 1 if yes and 0 if no. The conservation constraint passed down through cascading workbenches is expressed as follows: ； In the formula, p i,t This indicates whether an automated guided vehicle (AGV) arrives at the i-th workbench and passes straight through at time t; 1 indicates yes, and 0 indicates no. This indicates whether at time t an automated guided vehicle (AGV) has arrived at the i-th workbench and started to dock and unload k goods; 1 indicates yes, 0 indicates no. The time steps required to travel at a constant speed from the (i-1)th workbench to the ith workbench; express At any given moment, if an automated guided vehicle arrives at the (i-1)th workstation and moves straight past it, the value is 1; otherwise, it is 0. express At any given time, if an automated guided vehicle arrives at the (i-1)th workbench and begins to dock and unload k goods, 1 indicates yes and 0 indicates no. Shelf load flow constraints are used to track changes in the quantity of goods carried during the movement of automated guided vehicles (AGVs) and to match the correspondence between load flow and physical operation actions. Specifically, they include: The initial departure capacity constraint is expressed as: ； In the formula, Let t be the remaining fully loaded cargo carried by the automated guided vehicle that arrives at the first workbench at time t; Load propagation constraints across nodes are expressed as follows: ； In the formula, Let t be the remaining fully loaded cargo carried by the automated guided vehicle that arrives at the i-th workstation at time t; for The number of remaining goods carried by the automated guided vehicle that passes the (i-1)th workbench at any given time; for The number of remaining goods on the vehicle is automatically guided after the (i-1)th workbench stops and k goods are unloaded. The expression for unloading the deduction constraint is: ； In the formula, Let t be the number of remaining goods carried by the automated guided vehicle that passes the i-th workbench in a straight line at time t; Let t be the number of goods remaining on the autoguide vehicle after the i-th workbench stops and k goods are unloaded; Physical capacity binding constraints include straight-through scenario binding constraints and docking / unloading scenario binding constraints, with the following expressions: ； ； Implicit anti-collision constraints on monorails are used to avoid track collisions and platform congestion from a planning perspective. Specifically, they include: The departure safety interval constraint is expressed as: ； In the formula, For the time step summation variable; for Whether an automated guided vehicle (AGV) carrying goods has been dispatched from the replenishment point at any given time is 1 if yes and 0 if no. The physical mutual exclusion constraint of the workbench is expressed as follows: ； In the formula, express At any given moment, 1 indicates whether an automated guided vehicle (AGV) has arrived at the i-th workbench and started to dock and unload k goods; 0 indicates otherwise. Dynamic inventory balancing constraints are used to maintain a reasonable range of material inventory at each workstation, ensuring zero supply disruptions and no inventory overflows. Specifically, this includes: The dynamic inventory update constraint is expressed as follows: ； In the formula, I i,t-1 This represents the remaining inventory of materials at the i-th workbench at time t-1; express At any given moment, 1 indicates whether an automated guided vehicle (AGV) has arrived at the i-th workbench and started to dock and unload k goods; 0 indicates otherwise. Inventory boundary constraints, expressed as: 。 6. The scheduling optimization method for material distribution of unidirectional loop multi-load automated guided vehicles according to claim 5, characterized in that, The logical coupling between the physical flow and the load flow is achieved through physical capacity binding constraints. The physical capacity binding constraints use the vehicle spatiotemporal trajectory class 0-1 discrete variables as the upper bound control factor of the shelf load flow class continuous variables, binding the physical existence state of the automated guided vehicle with the quantity of goods carried, including binding constraints for straight-through scenarios and binding constraints for docking and unloading scenarios. Through the two binding constraints, when no automated guided vehicle arrives at the corresponding workbench, the value of the corresponding shelf load flow variable is limited to zero.

7. The scheduling optimization method for material distribution of unidirectional loop multi-load automated guided vehicles according to claim 6, characterized in that, The specific process of solving the problem using the time-sensitive driven greedy stochastic adaptive search algorithm is as follows: Initialize the algorithm parameters, set the maximum number of iterations, and initialize the optimal scheduling scheme and the optimal end-of-period inventory balance score; The process of constructing a heuristic process to generate a feasible scheduling scheme is as follows: the first workstation in the entire field with a risk of supply interruption is located by global supply interruption early warning monitoring and the corresponding time of supply interruption. Based on the time of supply interruption, the latest departure time is deduced in reverse. The legal departure window is searched forward and a fire-fighting slot is reserved for the workstation with supply interruption. The remaining shelf slots of the automated guided vehicle are filled by a greedy random mechanism of a limited candidate set. After the feasibility is verified, the corresponding spatiotemporal resources are locked and the system status is updated. The global supply interruption early warning monitoring and slot filling process is executed cyclically until all workstations have no risk of supply interruption within the planning period, and a feasible scheduling scheme is obtained. Perform a dual-objective evaluation on the generated feasible scheduling schemes, and update the current optimal scheduling scheme based on the evaluation results; Repeat the heuristic process of constructing the optimal solution and the optimal solution update steps until the preset maximum number of iterations is reached, and output the final optimal automated guided vehicle scheduling scheme.

8. The scheduling optimization method for material distribution of unidirectional loop multi-load automated guided vehicles according to claim 7, characterized in that, The specific steps for filling the remaining capacity of automated guided vehicles using a greedy random mechanism with a limited candidate set are as follows: Construct a candidate workbench set: Iterate through all workbench, tentatively add an unloading slot to each workbench and verify the validity of the process, add the workbench that passes the verification to the candidate set, and calculate the supply interruption urgency score of the corresponding workbench. The earlier the supply interruption occurs, the higher the urgency. Candidate set sorting: Sort the workbenches in the candidate set from highest to lowest according to the urgency of the supply disruption; Random selection filling: Randomly select one workbench from the top three workbench of the sorted candidate set and add an unloading slot to it; Repeat the steps of building, sorting and randomly selecting to fill the candidate set until all remaining shelf slots of the automated guided vehicle are filled, or there are no valid candidate workbenches.

9. The scheduling optimization method for material distribution of unidirectional loop multi-load automated guided vehicles according to claim 7, characterized in that, The rules for dual-objective evaluation are as follows: The first priority evaluation target is the total number of departures; the fewer the total number of departures, the higher the priority of the scheduling scheme. The second priority evaluation target is the ending inventory balance score, which is the sum of the ending inventory of all workstations. When the total number of departures is the same for two scheduling schemes, the scheduling scheme with the higher end-of-period inventory balance score is judged to be better. The conditions for updating the optimal scheduling scheme are: the total number of departures of the current feasible scheduling scheme is less than that of the current optimal scheme, or the total number of departures is the same but the ending inventory balance score is higher than that of the current optimal scheme.

10. A scheduling optimization system for material delivery of unidirectional loop multi-load automated guided vehicles (AGVs), characterized in that, An application to a flexible manufacturing material distribution system consisting of a single replenishment point, multiple workstations, and a unidirectional circular track, for executing the scheduling optimization method for material distribution of a unidirectional circular multi-load automated guided vehicle (AGV) as described in any one of claims 1-9, the system comprising: The scenario configuration and constraint definition module is used to configure scheduling constraints and optimization objectives when adopting the rack-on-vehicle operation mechanism. The mixed-integer linear programming model building module is used to construct a mixed-integer linear programming scheduling model that couples entity flow and load flow. The just-in-time driven greedy stochastic adaptive search solution module is used to solve the mixed-integer linear programming scheduling model, output the optimal automated guided vehicle scheduling scheme that satisfies all constraints, and send it to the flexible manufacturing material distribution system.

Images