AGV cooperative scheduling system and method based on mixed integer programming

Abstract

The invention relates to the field of automatic guided vehicle scheduling, in particular to an AGV collaborative scheduling system and method based on mixed integer programming, and the system comprises a data collection module, a central processing module, a scheme output module, an AGV scheduling module and a conflict detection module, and the data collection module collects storage and transportation operation demand data and task data; the central processing module receives the data and performs solution analysis based on a mixed integer programming model to generate a scheduling scheme, and the model comprises a dual-objective weighting function and a multi-dimensional constraint dynamic balance matrix; the scheme output module displays the scheme and receives an instruction; the AGV scheduling module schedules the AGV according to the scheme and feeds back execution information; and the conflict detection module detects a conflict state and generates conflict scheduling information, and realizes balanced optimization of the resource utilization rate and the energy consumption by establishing a dual-target weighting function which considers the minimum idle time of the AGV and the minimum total driving distance of the AGV at the same time.

Description

Technical Field

[0001] The present invention relates to the field of automated guided vehicle (AGV) scheduling, and specifically to an AGV collaborative scheduling system and method based on mixed integer programming, which is applicable to multi-AGV collaborative operation scenarios in the fields of intelligent warehousing, intelligent manufacturing, intelligent logistics, etc. Background Art

[0002] With the rapid development of Industry 4.0 and intelligent manufacturing, AGVs have been widely used in modern logistics and warehousing systems to realize the automated transportation and handling of goods. In large-scale warehousing environments, the collaborative operation of multiple AGVs has become an important means to improve logistics efficiency. However, with the increase in the number of AGVs and the complexity of the warehousing environment, traditional AGV scheduling methods face many challenges:

[0003] First, traditional AGV scheduling systems often employ single-objective optimization strategies, pursuing either the shortest path or minimum latency, making it difficult to strike a good balance between resource utilization and energy consumption. Second, existing systems often employ static planning methods, lacking the ability to rapidly respond to dynamic environmental changes. Once conflicts or anomalies arise, the entire scheduling system may face the risk of collapse. Furthermore, existing AGV scheduling systems generally lack adaptive learning capabilities and are unable to continuously optimize scheduling strategies based on historical operation data, resulting in a failure to increase system efficiency over time.

[0004] Although some scheduling methods based on mixed integer programming have been proposed, most fail to simultaneously consider two key metrics: AGV idle time and travel distance. They also lack the ability to handle multi-AGV conflict detection and collaborative optimization. Therefore, there is an urgent need for an AGV collaborative scheduling system that can comprehensively consider multiple optimization objectives and possesses dynamic response and adaptive learning capabilities. Summary of the Invention

[0005] The purpose of the present invention is to provide an AGV collaborative scheduling system and method based on mixed integer programming, aiming to solve the problems of single-objective optimization, static planning and lack of adaptive learning ability in the existing technology.

[0006] The present invention proposes an AGV collaborative scheduling system based on mixed integer programming, comprising:

[0007] Data collection module, used to collect warehousing and transportation operation demand data and task data;

[0008] a central processing module, communicatively connected to the data acquisition module, configured to receive the warehousing and transportation operation demand data and the task data sent by the data acquisition module, and generate a scheduling method scheme by performing a solution analysis based on the warehousing and transportation operation demand data and the task data using a mixed integer programming model, wherein the mixed integer programming model includes a dual-objective weighted function that simultaneously considers minimizing AGV idle time and minimizing AGV total travel distance, and a multi-dimensional constraint dynamic balance matrix that includes AGV load constraints, battery power constraints, task time constraints, and AGV travel trajectory constraints;

[0009] A scheme output module, which is in communication with the central processing module and is used to receive the scheduling method scheme sent by the central processing module, display it and receive input instructions;

[0010] an AGV scheduling module, communicatively connected to the central processing module and the solution output module, configured to schedule AGVs according to the scheduling method solution output by the solution output module, and to send AGV execution feedback information to the central processing module;

[0011] The conflict detection module is in communication with the central processing module and the AGV scheduling module, and is used to detect the conflict status during the operation of the AGV, generate conflict scheduling information and send it to the central processing module, so that the central processing module can regenerate the scheduling method solution based on the conflict scheduling information.

[0012] Preferably, the dual-objective weighted function in the mixed integer programming model is expressed as:

[0013] Maximize(Min Time+Min Path),

[0014] Among them, MinTime represents the minimum idle time target, MinPath represents the shortest path target, and t ij represents the difference between the start time of the i-th task and the j-th task, d ij represents the path distance between the i-th task and the j-th task, S i Represents the task information set of the i-th task, including task assignment information, task quantity, task start time, task end time, and task execution type.

[0015] Preferably, the multi-dimensional constraint dynamic balance matrix includes:

[0016] AGV load constraint, expressed as: W i ≤L·(1-α),

[0017] Among them, W i is the weight of the i-th task, L is the AGV load, and α is the battery loss rate;

[0018] Battery power constraint, expressed as: ∑(W i ·α)≤Q0

[0019] Among them, Q is the battery capacity, Q0 is the initial battery capacity;

[0020] Task time constraint, expressed as: t i ≤T and

[0021] Among them, t i is the execution time of the i-th task, T is the total time window of the warehouse, t s is the driving speed of AGV within the safety range, d i is the distance between the center position of the i-th task and the task point;

[0022] AGV driving trajectory constraint, expressed as: (x i ,y i )=(x ij ,y ij ),

[0023] Among them, x i 、y i are the center X-axis coordinates and Y-axis coordinates of the i-th task, respectively. ij 、y ij are the X-axis coordinate and Y-axis coordinate of the starting point of the road where the AGV travels between task i and task j.

[0024] Preferably, the central processing module includes:

[0025] A mathematical model building unit, used to build a mathematical model including an objective function and constraint conditions;

[0026] A heuristic algorithm processing unit, configured to solve and analyze the mathematical model based on the heuristic algorithm and historical data to generate a preliminary scheduling method solution;

[0027] The optimization processing unit is used to optimize the preliminary scheduling method solution to generate a final scheduling method solution, and transmit the final scheduling method solution to the solution output module.

[0028] Preferably, the specific processing steps of the heuristic algorithm processing unit include:

[0029] After inputting the number of AGVs and the warehouse operation time window, the heuristic algorithm parameters are initialized, the number of iterations, the initialization population per generation, the mutation probability, the maximum fitness value and the minimum fitness value are set, and the fitness function is set;

[0030] Initialize the population randomly and use the selection, crossover and mutation operations of the genetic algorithm to generate a new scheduling scheme population;

[0031] Calculate the fitness value of the new population, using the crowding distance as the winning criterion. The minimum fitness value is expressed as:

[0032] Min Dist=Min(min(P t -C dist )),

[0033] Among them, C dist Indicates the current distance, P t Represents the target distance. The target object represents the situation where all tasks have the minimum delay value in a given time window. The target distance Pt calculation formula is expressed as:

[0034]

[0035] Among them, S2 represents the total number of tasks in a given time window, and ω represents the task proportion factor;

[0036] The scheduling method with the minimum fitness value is used as the preliminary scheduling method.

[0037] Preferably, the specific processing steps of the optimization processing unit include:

[0038] Conduct cluster analysis on the preliminary scheduling method scheme and decompose it into different task sets;

[0039] Recalculating the fitness value of the task set, wherein the fitness value is recalculated based on historical job data so that the preliminary scheduling method scheme has a positive impact on each task in the task set and has the same impact on different tasks in the task set;

[0040] The task set with the smallest fitness value is selected as the final scheduling method.

[0041] Preferably, the specific steps of recalculating the fitness value of the task set are:

[0042] Randomly perturb the task set in the preliminary scheduling method scheme to obtain multiple perturbed task sets;

[0043] Calculate the delay cost, total delay time and travel cost of each task set;

[0044] Divide the delay cost by the total delay time to calculate the minimum delay cost;

[0045] Divide the driving cost by the total driving time to calculate the driving cost;

[0046] After adding the minimum delay cost and the travel cost, the fitness value of the disturbed task set is obtained;

[0047] The fitness value of the task set is expressed as:

[0048]

[0049] Among them, Delay represents the delay cost, MaxDelay and MaxTime represent the maximum delay time value and the maximum driving time value respectively, DelayCost and DistCost represent the delay cost and distance cost respectively, and Dist represents the driving cost.

[0050] Preferably, the conflict detection module includes:

[0051] The first conflict detection model is used to detect position conflicts of AGVs on the paths within the warehouse;

[0052] The second conflict detection model is used to detect time conflicts of AGVs on the paths within the warehouse;

[0053] A conflict analysis unit is configured to generate conflict scheduling information based on detection results of the first conflict detection model and the second conflict detection model.

[0054] As an option, it also includes:

[0055] An AGV status identification subsystem is communicatively connected to the central processing module and includes a data acquisition unit, an AGV status identification unit, and a data transmission unit, and is used to collect AGV status parameters, identify the AGV's own status and the container status, and store the AGV status parameters in the central processing module;

[0056] The AGV positioning and identification subsystem is in communication with the central processing module and includes an RFID reader. The RFID reader is set at the node where the AGV passes through and is identified by the AGV status identification subsystem, and is used to locate the AGV position in real time and transmit it to the central processing module.

[0057] As an advantage, the method comprises the following steps:

[0058] S1. Collecting storage and transportation operation demand data and task data through the data acquisition module;

[0059] S2. The central processing module receives the warehousing and transportation operation demand data and the task data, and establishes a mathematical model, wherein the mathematical model includes an objective function and constraint conditions;

[0060] S3, the central processing module solves and analyzes the mathematical model based on the heuristic algorithm and historical data to generate a preliminary scheduling method solution;

[0061] S4, the central processing module optimizes the preliminary scheduling method solution to generate a final scheduling method solution, and transmits the final scheduling method solution to the solution output module;

[0062] S5, the solution output module receives the final scheduling method solution, displays it and receives input instructions;

[0063] S6. The AGV scheduling module schedules the AGV according to the final scheduling method scheme output by the scheme output module, and sends AGV execution feedback information to the central processing module;

[0064] S7, the conflict detection module detects the conflict status during the operation of the AGV, generates conflict scheduling information and sends it to the central processing module;

[0065] S8. The central processing module regenerates the scheduling method based on the conflict scheduling information and returns to step S4.

[0066] Beneficial effects of the present invention:

[0067] 1. By establishing a dual-objective weighted function that simultaneously considers minimizing AGV idle time and minimizing the total AGV travel distance, a balanced optimization of resource utilization and energy consumption is achieved. Test data shows that the overall efficiency is improved by 15-20% compared with single-objective optimization.

[0068] 2. Construct a multi-dimensional constraint dynamic balance matrix, so that the system can dynamically adjust the relationship between constraint conditions according to actual conditions, solving the problem of system crash caused by constraint conflicts in actual operations.

[0069] 3. An adaptive genetic algorithm with a crowding distance selection mechanism was introduced to reduce the computation time in multi-AGV scenarios by more than 40%, while improving the solution quality and effectively solving the combinatorial explosion problem in complex scenarios.

[0070] 4. Through the task impact matrix driven by historical data and the random disturbance mechanism, the scheduling plan is continuously self-optimized. The longer the system runs, the higher the scheduling efficiency.

[0071] 5. Design a conflict detection and collaborative scheduling linkage mechanism to achieve real-time detection and dynamic adjustment of AGV operation conflicts, enabling the system to cope with complex and changing warehousing environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 Schematic diagram of the structure of the AGV collaborative scheduling system based on mixed integer programming of the present invention;

[0073] Figure 2Schematic diagram of the internal structure of the central processing module of the present invention;

[0074] Figure 3 This is a workflow diagram of the conflict detection module of the present invention;

[0075] Figure 4 This is a schematic diagram of the multi-dimensional constraint dynamic balance matrix of the present invention;

[0076] Figure 5 is a flow chart of the adaptive genetic algorithm of the present invention;

[0077] Figure 6 This is a schematic diagram of the task impact matrix of the present invention;

[0078] Figure 7 Flowchart of the AGV collaborative scheduling method based on mixed integer programming of the present invention;

[0079] Figure 8 The figure is a comparison chart of the application effect of the system of the present invention in an actual warehousing environment. DETAILED DESCRIPTION

[0080] Please refer to the attached Figure 1-8 The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to illustrate and explain the present invention and are not intended to limit the present invention.

[0081] Reference Figure 1 The AGV collaborative scheduling system based on mixed integer programming provided by the present invention includes a data acquisition module 1, a central processing module 2, a solution output module 3, an AGV scheduling module 4 and a conflict detection module 5.

[0082] Data acquisition module 1 is used to collect warehousing and transportation operation demand data and task data. Preferably, the warehousing and transportation operation demand data includes warehouse layout information, cargo storage location information, and AGV quantity and type information; the task data includes task type, task priority, task start and end times, and task start and end locations. In one specific embodiment of the present invention, data acquisition module 1 acquires operation demand data through a warehouse management system interface and simultaneously acquires task status data in real time through a sensor network at a frequency of five times per second to ensure sufficient real-time performance.

[0083] The central processing module 2 is in communication with the data acquisition module 1 and is configured to receive the warehousing and transportation operation demand data and the task data transmitted by the data acquisition module 1. Based on the warehousing and transportation operation demand data and the task data, the central processing module 2 performs a solution analysis using a mixed integer programming model to generate a scheduling method solution. The mixed integer programming model includes a dual-objective weighted function that simultaneously considers minimizing AGV idle time and minimizing AGV total travel distance, as well as a multi-dimensional constraint dynamic balance matrix that includes AGV load constraints, battery power constraints, task time constraints, and AGV travel trajectory constraints.

[0084] like Figure 2 As shown, in a preferred embodiment of the present invention, the central processing module 2 includes a mathematical model building unit 21, a heuristic algorithm processing unit 22 and an optimization processing unit 23.

[0085] The mathematical model construction unit 21 is used to establish a mathematical model including objective functions and constraints. In a specific embodiment of the present invention, the mathematical model is a multi-objective problem. The objective functions include minimizing the AGV idle time and minimizing the total AGV travel distance. The constraints include AGV load constraints, battery power constraints, task start time constraints, task end time constraints, AGV movement range constraints, and AGV travel trajectory constraints.

[0086] Preferably, the objective function of minimizing the AGV idle time and the objective function of minimizing the total AGV travel distance can be expressed by the following formula:

[0087] Maximize(Min Time+Min Path),

[0088] Among them, Maximize represents the optimization method to maximize the function, MinTime and MinPath represent the minimum time and shortest path respectively, ij represents the difference between the start time of the i-th task and the j-th task, d ij represents the path distance between the i-th task and the j-th task, S i Represents the task information set of the i-th task, including task assignment information, task quantity, task start time, task end time, and task execution type.

[0089] In practical applications, the minimum time target MinTime is usually expressed as the minimization of the time difference of all tasks, which can be mathematically expressed as The shortest path goal MinPath is expressed as the minimization of the path distance between all tasks, which can be expressed as By properly setting the weights of these two goals, the system can adjust its emphasis according to actual needs. For example, during peak hours, the weight of MinTime can be set to 0.7 and the weight of MinPath to 0.3 to prioritize timely task completion. During off-peak hours, the weight configuration can be adjusted to the opposite to prioritize energy consumption optimization.

[0090] The multi-dimensional constraint dynamic balance matrix includes multiple constraint conditions, which together constitute a constraint space and limit the range of feasible solutions for AGV scheduling. Specifically, it includes:

[0091] The AGV load constraint is expressed as follows:

[0092] W i ≤L·(1-α),

[0093] Among them, W i is the weight of the i-th task, L is the AGV load, and α is the battery loss rate. In practical applications, α is usually related to the AGV battery characteristics and generally ranges from 0.05-0.15, with a preferred value of 0.1, which means that the AGV consumes 10% additional power when fully loaded.

[0094] The battery power constraint is expressed as follows:

[0095] ∑(W i ·α)≤Q0,

[0096] Where Q is the battery capacity, and Q0 is the initial battery charge. In the embodiments of the present invention, the AGV battery capacity Q is typically designed to last for 8 hours of operation, typically 5000-8000 mAh. The initial charge Q0 is determined based on the actual charging status of the AGV when it goes online. The system sets a 20% charge margin as a safety margin. When the AGV's charge falls below this threshold, the system automatically schedules its charging task.

[0097] The task start time constraint is expressed as follows:

[0098] t i ≤T,

[0099]

[0100] Among them, t i is the execution time of the i-th task, which is determined by the scheduling method, T is the total time window of the warehouse, and t s is the driving speed of AGV within the safety range, d i is the distance between the center position of the i-th task and the task point. In practical applications, t sIt is generally set to 1.2-1.5m / s, which is determined by taking into account the safety factors in the warehouse and the technical characteristics of the AGV. T is set according to the actual working hours of the warehouse, usually 8 hours or 24 hours.

[0101] The task end time constraint is expressed as follows:

[0102] e i ≤T,

[0103]

[0104] Among them, e i is the end time of the i-th task, t s is the driving speed of AGV within the safety range, d i is the distance between the center position of the i-th task and the task point. This set of constraints ensures that the task is completed within the specified time window and takes into account the minimum time required for the AGV to travel from the starting point to the task location.

[0105] The AGV moving range constraint is expressed as follows:

[0106] (x i ,y i )∈Warehouse area,

[0107] Among them, x i 、y i are the X-axis and Y-axis coordinates of the center of the i-th task, respectively. This constraint ensures that the AGV travels only within the permitted area of ​​the warehouse, avoiding entering restricted areas or collisions. In system implementation, the warehouse area is typically represented by a rectangle or polygon. The system uses a point-in-polygon algorithm to detect whether a coordinate point is within the permitted area.

[0108] The AGV driving trajectory constraint is expressed as follows:

[0109] (x i ,y i )=(x ij ,y ij ),

[0110] Among them, x i 、y i are the center X-axis coordinates and Y-axis coordinates of the i-th task, respectively. ij 、y ij are the X-axis coordinates and Y-axis coordinates of the starting point of the path that the AGV travels between tasks i and j. This constraint ensures that the AGV travels along the predefined path, avoiding safety hazards caused by random driving.

[0111] The heuristic algorithm processing unit 22 is used to solve and analyze the mathematical model based on the heuristic algorithm and historical data to generate a preliminary scheduling method solution. In one embodiment of the present invention, the specific processing steps of the heuristic algorithm processing unit 22 include:

[0112] First, after inputting the number of AGVs and the warehouse operation time window, the heuristic algorithm parameters are initialized, including the number of iterations, the initial population per generation, the mutation probability, the maximum and minimum fitness values, and the fitness function. In one specific example, the number of iterations can be set to 100-200, the initial population size can be set to 50-100, and the mutation probability is preferably set to 0.01-0.05. These parameters are empirically determined through extensive experimentation to ensure algorithm convergence while balancing computational efficiency.

[0113] Then, the population is randomly initialized, and a new population of scheduling schemes is generated using the genetic algorithm's selection, crossover, and mutation operations. In this embodiment, the selection operation uses a strategy combining roulette wheel and elite selection, the crossover operation uses a two-point crossover method, and the mutation operation uses uniform mutation. This combination of operations has been proven in practice to achieve a good balance between convergence and diversity.

[0114] Next, the fitness value of the new population is calculated, using the crowding distance as the winning criterion. The minimum fitness value is expressed as:

[0115] Min Dist=Min(min(P t -C dist )),

[0116] Among them, C dist Indicates the current distance, P t Represents the target distance. The target object represents the situation where all tasks have the minimum delay value in a given time window. The target distance P t The calculation formula is expressed as:

[0117]

[0118] Here, S2 represents the total number of tasks within a given time window, and ω represents the task ratio. In practice, the value of ω is typically derived from historical data statistics, reflecting the relationship between warehouse task load and time. It generally ranges from 0.2 to 0.8, and the specific value is dynamically adjusted based on the actual warehouse operation. For example, in an e-commerce warehouse, ω is often close to 0.8 during promotional seasons, indicating a high density of warehouse tasks; during regular periods, it may be between 0.3 and 0.5, indicating a moderate density of tasks.

[0119] Finally, the scheduling method with the minimum fitness value is used as the preliminary scheduling method, which will be further optimized in subsequent steps.

[0120] The optimization processing unit 23 is used to optimize the preliminary scheduling method solution to generate a final scheduling method solution, and transmit the final scheduling method solution to the solution output module 3. In one embodiment of the present invention, the specific processing steps of the optimization processing unit 23 include:

[0121] First, cluster analysis is performed on the preliminary scheduling method proposal to decompose it into distinct task sets. In this embodiment, cluster analysis uses a modified K-means algorithm to divide tasks into clusters based on their temporal and spatial characteristics. The number of clusters, K, is typically set to 1.2-1.5 times the number of AGVs. This ensures that each AGV is assigned an appropriate number of tasks while retaining some room for adjustment.

[0122] Then, the fitness value of the task set is recalculated. The fitness value is recalculated based on the historical job data so that the initial scheduling method has a positive impact on each task in the task set and has the same impact on different tasks in the task set. Specifically, the steps of recalculating the fitness value of the task set are:

[0123] First, the task set in the preliminary scheduling method solution is randomly perturbed to obtain multiple perturbed task sets. In this embodiment, the random perturbation is achieved by changing the execution order or allocation relationship of the tasks. The number of perturbations is set to 10-20 times to generate sufficient solution diversity.

[0124] Next, the delay cost, total delay time, and travel cost are calculated for each task set. The delay cost is divided by the total delay time to obtain the minimum delay cost; the travel cost is divided by the total travel time to obtain the travel cost. The minimum delay cost and travel cost are added together to obtain the fitness value of the perturbed task set.

[0125] The fitness value of the task set is expressed as:

[0126]

[0127] Among them, Delay represents the delay cost, MaxDelay and MaxTime represent the maximum delay time value and the maximum driving time value respectively, DelayCost and DistCost represent the delay cost and distance cost respectively, and Dist represents the driving cost.

[0128] The delay cost DelayCost is expressed as:

[0129] DelayCost=∑T ij,

[0130] Among them, T ij represents the delay cost from task i to task j, T ij =a ij -c ij , a ij Indicates the actual completion time, c ij Indicates the planned completion time.

[0131] The distance cost DistCost is expressed as:

[0132] DistCost=∑Dist(i,j),

[0133] Among them, Dist(i,j) represents the driving cost from task i to task j, which is usually proportional to the distance and also takes into account the congestion level and road conditions of the path.

[0134] In an embodiment of the present invention, the weight ratio of delay cost and distance cost is generally set according to business needs. In scenarios with high timeliness requirements, the delay cost weight can be set to 0.7-0.8; in energy optimization scenarios, the distance cost weight can be increased to 0.6-0.7.

[0135] Finally, the task set with the smallest fitness value is selected as the final scheduling method solution. This final solution will be transmitted to the solution output module 3 for subsequent scheduling execution.

[0136] The plan output module 3 is in communication with the central processing module 2 and is configured to receive the scheduling method plan sent by the central processing module 2, display it, and accept input instructions. In a preferred embodiment of the present invention, the plan output module 3 includes a display unit and a control unit. The display unit uses a high-resolution touch screen to visually display the scheduling plan and real-time execution status; the control unit provides a manual intervention interface, allowing management personnel to fine-tune the scheduling plan when necessary.

[0137] The AGV scheduling module 4 is in communication with the central processing module 2 and the solution output module 3, and is used to schedule AGVs according to the scheduling method solution output by the solution output module 3, and send AGV execution feedback information to the central processing module 2. In one embodiment of the present invention, the AGV scheduling module 4 establishes a connection with each AGV through a wireless communication network (such as Wi-Fi or 5G private network), issues scheduling instructions in real time, and receives status feedback. The AGV execution feedback information includes key parameters such as position, speed, power, and load status. The feedback frequency is 5 times per second, which is consistent with the collection frequency of the data acquisition module, ensuring the synchronization and consistency of the system.

[0138] The conflict detection module 5 is in communication with the central processing module 2 and the AGV scheduling module 4, and is configured to detect conflict states during AGV operation, generate conflict scheduling information, and send it to the central processing module 2, so that the central processing module 2 can regenerate the scheduling method based on the conflict scheduling information. In a preferred embodiment of the present invention, the conflict detection module 5 includes a first conflict detection model 51, a second conflict detection model 52, and a conflict analysis unit 53.

[0139] The first conflict detection model 51 is used to detect position conflicts between AGVs on the warehouse path. This model establishes a spatial relationship matrix between AGV positions. When the distance between two AGVs is less than a safety threshold (usually set to 1.5 times the AGV length, approximately 1.5-2.5 meters), it is determined to be a potential position conflict. The mathematical expression of this model can be simplified as follows:

[0140]

[0141] Among them, C pos (i, j) represents the position conflict state between AGVi and j, p i and p j Represent the position coordinates of AGVi and j respectively, d safe represents the safety distance threshold, and ∥·∥ represents the Euclidean distance.

[0142] The second conflict detection model 52 is used to detect time conflicts on the AGV's path within the warehouse. This model focuses on conflicts that may occur when AGVs are at the same location at different times, especially in critical areas such as intersections or narrow passages. The mathematical expression of time conflicts can be expressed as:

[0143]

[0144] Among them, C time (i, j, p) represents the time conflict state between AGVi and j at position p, t i (p) and t j (p) represents the time when AGVi and j are at position p, Δt safe Indicates a safe time interval, usually set to 5-10 seconds.

[0145] The conflict analysis unit 53 is configured to generate conflict scheduling information based on the detection results of the first conflict detection model 51 and the second conflict detection model 52. The conflict scheduling information includes the conflict type, conflict location, conflicting AGV number, conflict severity, etc., providing a basis for rescheduling for the central processing module 2.

[0146] In a preferred embodiment of the present invention, Figure 1As shown, the AGV collaborative scheduling system based on mixed integer programming also includes an AGV state identification subsystem 6 and an AGV positioning identification subsystem 7.

[0147] The AGV status identification subsystem 6 is communicatively connected to the central processing module 2 and includes a data acquisition unit 61, an AGV status identification unit 62, and a data transmission unit 63. These subsystems are used to collect AGV status parameters, identify the AGV's own status and the container's status, and store these AGV status parameters in the central processing module 2. In this embodiment, the AGV status parameters include GPS data, altitude data, and pressure data. GPS data is used to locate the AGV's position coordinates, typically with an accuracy of 0.5-1 meter. Altitude data uses ultrasonic or infrared sensors to measure the height of objects above the AGV and is used to determine the load status. Pressure data is acquired via a pressure sensor and directly reflects the AGV's load status.

[0148] The AGV positioning and identification subsystem 7 is in communication with the central processing module 2 and includes an RFID reader. The RFID reader is located at the nodes that the AGV passes through, as identified by the AGV status identification subsystem 6. The RFID reader is used to locate the AGV's position in real time and transmit it to the central processing module 2. In this embodiment, the RFID reader operates in the 13.56MHz frequency band with a reading range of 0.5-1 meter. RFID tags are placed at key nodes in the warehouse (such as intersections, picking points, charging stations, etc.) to accurately locate the AGV as it passes through, compensating for the lack of GPS positioning accuracy indoors.

[0149] Example 2

[0150] Reference Figure 7 The present invention provides an AGV collaborative scheduling method based on the system, comprising the following steps:

[0151] S1, collecting storage and transportation operation demand data and task data through the data collection module 1;

[0152] In this step, data acquisition module 1 acquires operational demand data, including cargo information, storage location, and inbound and outbound requirements, through an interface with the warehouse management system (WMS). Simultaneously, a sensor network distributed throughout the warehouse collects real-time task data, including AGV status and environmental conditions. In a preferred embodiment of the present invention, data acquisition utilizes a distributed architecture, preprocessing data through edge computing nodes to reduce the computational burden on the central system.

[0153] S2, the central processing module 2 receives the warehousing and transportation operation demand data and the task data, and establishes a mathematical model, the mathematical model including an objective function and constraint conditions;

[0154] In this step, Central Processing Module 2 first cleans and standardizes the received data and then constructs a mixed integer programming model based on a pre-set template. As previously mentioned, this model includes a dual-objective weighted function and a multi-dimensional constrained dynamic balance matrix, comprehensively considering the various constraints and optimization objectives of AGV scheduling.

[0155] S3, the central processing module 2 solves and analyzes the mathematical model based on the heuristic algorithm and historical data to generate a preliminary scheduling method solution;

[0156] In this step, the central processing module 2 invokes the heuristic algorithm processing unit 22 to run an adaptive genetic algorithm to solve the mixed integer programming model. The algorithm first initializes parameters based on historical job data. Then, through iterative calculations, it generates a series of candidate solutions. It then uses a congestion distance mechanism to select the optimal solution set that balances efficiency and energy consumption. Ultimately, the solution with the lowest fitness is selected as the initial scheduling method.

[0157] S4, the central processing module 2 optimizes the preliminary scheduling method solution to generate a final scheduling method solution, and transmits the final scheduling method solution to the solution output module 3;

[0158] In this step, the optimization processing unit 23 of the central processing module 2 further optimizes the preliminary solution. The optimization process first decomposes the tasks into different sets through cluster analysis. Then, random perturbations and fitness recalculations are performed on each task set. The task set with the lowest fitness value is finally selected as the final scheduling method solution, taking into account both delay costs and distance costs.

[0159] S5, the solution output module 3 receives the final scheduling method solution, displays it and receives input instructions;

[0160] In this step, the plan output module 3 displays the final scheduling plan in a visual form on the display interface, including task allocation diagrams, path planning diagrams, time schedules, etc., and provides an interactive interface to allow managers to view details, modify parameters or manually intervene.

[0161] S6, the AGV scheduling module 4 schedules the AGV according to the final scheduling method scheme output by the scheme output module 3, and sends the AGV execution feedback information to the central processing module 2;

[0162] In this step, the AGV scheduling module 4 converts the scheduling plan into a sequence of instructions executable by the AGVs and sends it to each AGV via the wireless network. Each AGV performs its assigned task and provides real-time feedback on its position, speed, battery level, and other status information. This feedback is used for system monitoring and potential conflict detection.

[0163] S7, the conflict detection module 5 detects the conflict state during the operation of the AGV, generates conflict scheduling information and sends it to the central processing module 2;

[0164] In this step, the conflict detection module 5 simultaneously runs two models: position conflict detection and time conflict detection, monitoring the AGV's operating status in real time. When a potential conflict is detected, the system immediately generates conflict dispatch information, including the conflict type, location, involved AGVs, and severity, and sends it to the central processing module 2 for processing.

[0165] S8. The central processing module 2 regenerates the scheduling method based on the conflicting scheduling information and returns to step S4.

[0166] In this step, after receiving the conflict scheduling information, the central processing module 2 implements appropriate adjustment strategies based on the conflict type and severity. For minor conflicts, only partial adjustments to the schedules of the relevant AGVs may be necessary; for severe conflicts, the entire scheduling plan may need to be replanned. The adjusted plan returns to step S4 for optimization, resulting in a new final scheduling plan.

[0167] By cyclically executing the above steps, the present invention realizes the coordinated scheduling of multiple AGVs, can efficiently handle dynamically changing storage environments and task requirements, maximize AGV utilization and reduce energy consumption.

[0168] Example 3

[0169] In another preferred embodiment of the present invention, the mixed integer programming model of the AGV collaborative scheduling system also considers the positional relationship constraints between AGVs. Specifically, the positional relationship constraints between AGVs include the distribution rules of AGVs at the initial node:

[0170] ∥p i -p j ∥≥d AGV +δ,

[0171] Among them, d AGV is the length of the AGV, usually 1.2-1.8 meters; δ is the safety distance, preferably set to 0.5-1 meter; p i and p j are the position coordinates of the i-th AGV and the j-th AGV at the initial node, respectively. This constraint ensures that the AGVs are not too concentrated in the initial state, reducing the risk of congestion during the startup phase.

[0172] In addition, during the scheduling process, the system will dynamically calculate and maintain the safe distance between AGVs, and adaptively adjust the safe distance based on the speed and load status of the AGV:

[0173] dsafe (i,j)=d base +α·v i +β·L i ,

[0174] Among them, d base is the basic safety distance, which is usually set to 1.2 times the length of the AGV; α is the speed impact factor, with a preferred value of 0.5-1.0; v i is the current speed of AGVi; β is the load influence factor, the preferred value is 0.2-0.5; L i is the current load rate of the AGVi. This dynamic safety distance mechanism enables the system to flexibly adjust the safety strategy based on the actual operating status of the AGV, ensuring operational safety while avoiding efficiency losses caused by over-conservatism.

[0175] Example 4

[0176] In another preferred embodiment of the present invention, the adaptive genetic algorithm further includes a variety of improvement strategies to further enhance the performance and applicability of the algorithm:

[0177] 1. Adaptive crossover and mutation rate: The algorithm dynamically adjusts the crossover rate and mutation rate according to the diversity of the population. When the population tends to be consistent, the mutation rate is increased to increase diversity. When the population is too dispersed, the crossover rate is increased to accelerate convergence. Specifically, the crossover rate P c and mutation rate P m The calculation formula is:

[0178]

[0179] Among them, P c_base and P m_base are the basic crossover rate and mutation rate, respectively, which are usually set to 0.6-0.8 and 0.01-0.05; γ is the adjustment coefficient, with an optimal value of 0.1-0.3; D pop is the diversity measure of the current population, D max is the preset maximum diversity threshold.

[0180] 2. Elite retention strategy: In each iteration, a certain proportion of individuals with the highest fitness (usually 5-10%) are directly retained to the next generation, avoiding the risk of losing excellent solutions due to random selection.

[0181] 3. Local Search Enhancement: Apply local search algorithms (such as simulated annealing or tabu search) to some high-quality solutions to further improve their quality. The trigger condition for local search is that there is no significant improvement in the optimal solution of the population for N consecutive generations (usually N = 5-10).

[0182] Through the comprehensive application of these improvement strategies, the adaptive genetic algorithm of the present invention performs well in complex scenarios, can significantly improve computational efficiency while ensuring solution quality, and is suitable for the actual needs of large-scale AGV collaborative scheduling.

[0183] This invention provides an AGV collaborative scheduling system and method based on mixed integer programming. Through innovative technical approaches such as a dual-objective weighted function, a multi-dimensional constrained dynamic balance matrix, an adaptive genetic algorithm, and a historical data-driven task impact matrix, it achieves efficient scheduling of multiple AGV collaborative operations. Compared to existing technologies, this invention can simultaneously optimize AGV idle time and total travel distance, dynamically handle various constraints, detect and resolve conflicts in real time, and continuously optimize scheduling strategies through adaptive learning. Practical applications have demonstrated that this invention can significantly improve AGV utilization, reduce idle travel distance, and lower energy consumption and maintenance costs, providing a substantive and innovative technical solution for the field of intelligent warehousing and logistics.

[0184] Although the present invention has been disclosed above with reference to preferred embodiments, this is not intended to limit the present invention. Anyone skilled in the art may make various modifications and alterations without departing from the spirit and scope of the present invention. Therefore, the scope of protection of the present invention shall be determined by the claims.

Claims

1. The AGV collaborative scheduling system based on mixed integer programming is characterized by: Includes: Data collection module, used to collect warehousing and transportation operation demand data and task data; a central processing module, communicatively connected to the data acquisition module, configured to receive the warehousing and transportation operation demand data and the task data sent by the data acquisition module, and generate a scheduling method scheme by performing a solution analysis based on the warehousing and transportation operation demand data and the task data using a mixed integer programming model, wherein the mixed integer programming model includes a dual-objective weighted function that simultaneously considers minimizing AGV idle time and minimizing AGV total travel distance, and a multi-dimensional constraint dynamic balance matrix that includes AGV load constraints, battery power constraints, task time constraints, and AGV travel trajectory constraints; A scheme output module, which is in communication with the central processing module and is used to receive the scheduling method scheme sent by the central processing module, display it and receive input instructions; an AGV scheduling module, communicatively connected to the central processing module and the solution output module, configured to schedule AGVs according to the scheduling method solution output by the solution output module, and to send AGV execution feedback information to the central processing module; The conflict detection module is in communication with the central processing module and the AGV scheduling module, and is used to detect the conflict status during the operation of the AGV, generate conflict scheduling information and send it to the central processing module, so that the central processing module can regenerate the scheduling method solution based on the conflict scheduling information.

2. The AGV collaborative scheduling system based on mixed integer programming according to claim 1 is characterized in that: The dual-objective weighted function in the mixed integer programming model is expressed as: Maximize(Min Time+Min Path), Among them, MinTime represents the minimum idle time target, MinPath represents the shortest path target, and t ij represents the difference between the start time of the i-th task and the j-th task, d ij represents the path distance between the i-th task and the j-th task, S i Represents the task information set of the i-th task, including task assignment information, task quantity, task start time, task end time, and task execution type.

3. The AGV collaborative scheduling system based on mixed integer programming according to claim 1 is characterized in that: The multi-dimensional constraint dynamic balance matrix includes: AGV load constraint, expressed as: W i ≤L·(1-α), Among them, W i is the weight of the i-th task, L is the AGV load, and α is the battery loss rate; Battery power constraint, expressed as: ∑(W i ·α)≤Q0 Among them, Q is the battery capacity, Q0 is the initial battery capacity; Task time constraint, expressed as: t i ≤T and Among them, t i is the execution time of the i-th task, T is the total time window of the warehouse, t s is the driving speed of AGV within the safety range, d i is the distance between the center position of the i-th task and the task point; AGV driving trajectory constraint, expressed as: (x i ,y i )=(x ij ,y ij ), Among them, x i 、y i are the center X-axis coordinates and Y-axis coordinates of the i-th task, respectively. ij 、y ij are the X-axis coordinate and Y-axis coordinate of the starting point of the road where the AGV travels between task i and task j.

4. The AGV collaborative scheduling system based on mixed integer programming according to claim 1 is characterized in that: The central processing module includes: A mathematical model building unit, used to build a mathematical model including an objective function and constraint conditions; A heuristic algorithm processing unit, configured to solve and analyze the mathematical model based on the heuristic algorithm and historical data to generate a preliminary scheduling method solution; The optimization processing unit is used to optimize the preliminary scheduling method solution to generate a final scheduling method solution, and transmit the final scheduling method solution to the solution output module.

5. The AGV collaborative scheduling system based on mixed integer programming according to claim 4 is characterized in that: The specific processing steps of the heuristic algorithm processing unit include: After inputting the number of AGVs and the warehouse operation time window, the heuristic algorithm parameters are initialized, the number of iterations, the initialization population per generation, the mutation probability, the maximum fitness value and the minimum fitness value are set, and the fitness function is set; Initialize the population randomly and use the selection, crossover and mutation operations of the genetic algorithm to generate a new scheduling scheme population; Calculate the fitness value of the new population, using the crowding distance as the winning criterion. The minimum fitness value is expressed as: Min Dist=Min(min(P t -C dist )), Among them, C dist Indicates the current distance, P t Represents the target distance. The target object represents the situation where all tasks have the minimum delay value in a given time window. The target distance Pt calculation formula is expressed as: Among them, S2 represents the total number of tasks in a given time window, and ω represents the task proportion factor; The scheduling method with the minimum fitness value is used as the preliminary scheduling method.

6. The AGV collaborative scheduling system based on mixed integer programming according to claim 4 is characterized in that: The specific processing steps of the optimization processing unit include: Conduct cluster analysis on the preliminary scheduling method scheme and decompose it into different task sets; Recalculating the fitness value of the task set, wherein the fitness value is recalculated based on historical job data so that the preliminary scheduling method scheme has a positive impact on each task in the task set and has the same impact on different tasks in the task set; The task set with the smallest fitness value is selected as the final scheduling method.

7. The AGV collaborative scheduling system based on mixed integer programming according to claim 6 is characterized in that: The specific steps of recalculating the fitness value of the task set are: Randomly perturb the task set in the preliminary scheduling method scheme to obtain multiple perturbed task sets; Calculate the delay cost, total delay time and travel cost of each task set; Divide the delay cost by the total delay time to calculate the minimum delay cost; Divide the driving cost by the total driving time to calculate the driving cost; After adding the minimum delay cost and the travel cost, the fitness value of the disturbed task set is obtained; The fitness value of the task set is expressed as: Among them, Delay represents the delay cost, MaxDelay and MaxTime represent the maximum delay time value and the maximum driving time value respectively, DelayCost and DistCost represent the delay cost and distance cost respectively, and Dist represents the driving cost.

8. The AGV collaborative scheduling system based on mixed integer programming according to claim 1 is characterized in that: The conflict detection module includes: The first conflict detection model is used to detect position conflicts of AGVs on the paths within the warehouse; The second conflict detection model is used to detect time conflicts of AGVs on the paths within the warehouse; A conflict analysis unit is configured to generate conflict scheduling information based on detection results of the first conflict detection model and the second conflict detection model.

9. The AGV collaborative scheduling system based on mixed integer programming according to claim 1, characterized in that: Also includes: An AGV status identification subsystem is communicatively connected to the central processing module and includes a data acquisition unit, an AGV status identification unit, and a data transmission unit, and is used to collect AGV status parameters, identify the AGV's own status and the container status, and store the AGV status parameters in the central processing module; The AGV positioning and identification subsystem is in communication with the central processing module and includes an RFID reader. The RFID reader is set at the node where the AGV passes through and is identified by the AGV status identification subsystem, and is used to locate the AGV position in real time and transmit it to the central processing module.

10. The AGV collaborative scheduling method based on the system of claim 1, using the system of any one of claims 1-9, characterized in that: The following steps are involved: S1. Collecting storage and transportation operation demand data and task data through the data acquisition module; S2. The central processing module receives the warehousing and transportation operation demand data and the task data, and establishes a mathematical model, wherein the mathematical model includes an objective function and constraint conditions; S3, the central processing module solves and analyzes the mathematical model based on the heuristic algorithm and historical data to generate a preliminary scheduling method solution; S4, the central processing module optimizes the preliminary scheduling method solution to generate a final scheduling method solution, and transmits the final scheduling method solution to the solution output module; S5, the solution output module receives the final scheduling method solution, displays it and receives input instructions; S6. The AGV scheduling module schedules the AGV according to the final scheduling method scheme output by the scheme output module, and sends AGV execution feedback information to the central processing module; S7, the conflict detection module detects the conflict status during the operation of the AGV, generates conflict scheduling information and sends it to the central processing module; S8. The central processing module regenerates the scheduling method based on the conflict scheduling information and returns to step S4.

Images