AGV cluster scheduling method and system thereof

Abstract

The invention relates to the technical field of automatic guided vehicles, in particular to an AGV cluster scheduling method and system, and the method comprises the steps: converting multi-source order information into a standard order format, generating an order priority based on the time urgency, order source weight, material importance and historical distribution evaluation, and creating a distribution task; meanwhile, obtaining a distribution constraint condition, performing feasibility analysis on the distribution task, and generating an executable task; according to the AGV state information, the most suitable AGV is selected to execute an executable task through a multi-dimensional attribute matching mechanism; monitoring the execution state of the AGV in real time, and when an abnormal state is detected, storing the context of the task and reselecting the AGV to continue to execute the task; a scheduling instruction is generated, the scheduling instruction is safely packaged and sent to the target AGV, and the target AGV executes a task after executing multi-level safety verification on the received scheduling instruction; and through a multi-source order intelligent fusion mechanism, unified processing of different channel orders is realized, and the service adaptability of the system is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of automated guided vehicles, and in particular to an AGV cluster scheduling method and system thereof, which are used to realize intelligent AGV task allocation and execution management under multi-source orders. Background Art

[0002] With the rapid development of industries like logistics and manufacturing, AGVs (Automated Guided Vehicles) are widely used in factories, warehouses, hospitals, ports, and other scenarios for automated material transportation and distribution. As the scale of these applications expands, a single AGV is no longer sufficient to meet the growing distribution needs, and the coordinated operation of AGV clusters is becoming an inevitable trend.

[0003] Currently, traditional AGV scheduling systems have the following major problems: First, most use scheduling strategies based on simple rules, such as "nearest AGV first" or "fixed path loop", which are inefficient in complex environments; second, they lack the ability to intelligently integrate and process multi-source orders, making them unable to adapt to diverse business needs; third, the AGV selection process often only considers distance factors, resulting in uneven resource utilization, with some AGVs being overused while others are idle; fourth, the ability to handle abnormal situations is limited. Once an AGV fails or is occupied, the task often cannot be successfully completed; finally, the security mechanism is relatively simple, and there is a risk of misoperation or malicious attack.

[0004] To solve the above problems, there is an urgent need for an AGV cluster scheduling method and system that can intelligently process multi-source orders, comprehensively select AGVs based on multi-dimensional factors, flexibly respond to abnormal situations and ensure scheduling safety. Summary of the Invention

[0005] The purpose of the present invention is to provide an AGV cluster scheduling method and system thereof, aiming to solve the problems existing in the traditional AGV scheduling system, such as single order processing, uneven resource allocation, poor abnormal adaptability and insufficient security.

[0006] The present invention proposes an AGV cluster scheduling method, including:

[0007] Acquire multi-source order information, including terminal system orders, mobile client orders, inventory-triggered orders, and feedback-triggered orders;

[0008] Convert the multi-source order information into a standard order format, generate order priorities based on time urgency, order source weight, material importance, and historical delivery evaluation, and create delivery tasks;

[0009] Obtain delivery constraints, perform feasibility analysis on the delivery task based on physical distance constraints, time window constraints, resource availability constraints, and path accessibility constraints, and generate executable tasks;

[0010] Obtaining AGV status information, including basic status information, function parameter information, resource status information, and historical performance information;

[0011] Based on the executable tasks and the AGV status information, the most suitable AGV is selected to perform the executable tasks through a multi-dimensional attribute matching mechanism, wherein the multi-dimensional attribute matching mechanism includes pre-screening, distance scoring, power scoring, load balancing scoring, and time efficiency scoring;

[0012] Monitor the AGV execution status in real time. When abnormal conditions such as AGV being occupied, path blocked, power abnormality, execution timeout, or communication abnormality are detected, save the task context and reselect the AGV to continue the task.

[0013] Generate a dispatch instruction, securely encapsulate the dispatch instruction and send it to the target AGV, which performs multi-level security checks on the received dispatch instruction and then executes the task.

[0014] Preferably, the process of converting multi-source order information into a standard order format, generating order priorities based on time urgency, order source weight, material importance, and historical delivery evaluation, and creating delivery tasks specifically includes:

[0015] Convert order information from different sources into a standard order format that includes a unique order identifier, order source type, order creation timestamp, delivery point coordinates, material type, weight, volume, and expected delivery time window;

[0016] Calculate the time urgency score based on the difference between the expected delivery time and the current time;

[0017] Different basic weights are assigned based on the order source type. Inventory-triggered orders have a higher weight than terminal system orders, which in turn have a higher weight than mobile client orders.

[0018] Calculate the material importance score based on the importance coefficient corresponding to the material type;

[0019] Obtain users' historical delivery satisfaction ratings and calculate historical rating impact factors;

[0020] Calculate a comprehensive priority score based on the time urgency score, order source weight, material importance score, and historical evaluation impact factor;

[0021] According to the comprehensive priority score, a delivery task is generated including a task unique identifier, an associated order identifier, a task priority, delivery start and end point coordinates, material requirement parameters and a task deadline, and the delivery task is placed in a task queue.

[0022] Preferably, the feasibility analysis of the delivery task is performed based on the physical distance constraint, the time window constraint, the resource availability constraint, and the path accessibility constraint to generate an executable task, specifically including:

[0023] Obtain the distance between the delivery point and the nearest AGV station and compare it with the preset maximum delivery radius to determine whether the physical distance constraint is met;

[0024] Based on the current available AGV's moving speed and current position, the estimated completion time is calculated and compared with the task deadline to determine whether the time window constraint is met;

[0025] Check whether there is an AGV among the currently available AGVs capable of handling the specific material type, weight, and volume to determine whether the resource availability constraints are met;

[0026] Analyze whether there are obstacles or restricted areas on the path from the delivery start point to the destination, and determine whether the path accessibility constraints are met;

[0027] When all constraints are met, the task status is updated to "executable" and added to the executable task queue;

[0028] When the physical distance constraint or path accessibility constraint is not met, a "delivery not possible" notification is sent to the user;

[0029] When the time window constraints are not met, provide users with feasible alternative time options;

[0030] When resource availability constraints are temporarily unsatisfied, the task is marked as "waiting for resources" and re-evaluated periodically.

[0031] Preferably, the selecting the most suitable AGV to execute the executable task through a multi-dimensional attribute matching mechanism specifically includes:

[0032] Based on the material type, weight, and volume requirements of the task, select a candidate set of AGVs with execution capabilities from the AGV cluster;

[0033] Calculate the distance from each candidate AGV to the starting and ending points of the task, and generate a distance score based on the inverse of the distance;

[0034] Evaluate whether the power level of each candidate AGV is sufficient to complete the entire mission and generate a power score based on the remaining power;

[0035] Count the number of tasks recently executed by each candidate AGV and generate a load balancing score based on the balance of task distribution;

[0036] Generate a time efficiency score based on the AGV's movement speed and historical task completion efficiency;

[0037] Dynamically adjust the weight coefficients of distance score, power score, load balancing score, and time efficiency score based on the current system status and task characteristics;

[0038] Perform weighted summation on each candidate AGV’s scores to generate a comprehensive score;

[0039] The AGV with the highest comprehensive score is selected as the target AGV, and an assignment result containing task information and execution requirements is generated.

[0040] Preferably, when an abnormal state such as the AGV being occupied, the path being blocked, the power being abnormal, the execution being timed out, or the communication being abnormal is detected, the task context is saved and the AGV is reselected to continue the task, specifically including:

[0041] Receive AGV status feedback information in real time, and identify it as an abnormal state when the AGV status parameters exceed the preset threshold;

[0042] According to the abnormality type, the abnormal state is classified as AGV occupied, path blocked, power abnormality, execution timeout or communication abnormality;

[0043] Save task context information such as the execution progress of the current task, AGV location coordinates, list of executed subtasks, and list of remaining subtasks to be executed;

[0044] Send a task interruption command to the original AGV to confirm the interruption status;

[0045] Evaluate the current state of the task and decide whether to continue or restart from the current position;

[0046] Re-call the multi-dimensional attribute matching mechanism and select a new suitable AGV;

[0047] Pass the saved task context information to the newly selected AGV;

[0048] Adjust the execution plan of the remaining tasks based on the characteristics of the new AGV;

[0049] Send an instruction to the newly selected AGV to continue executing the task from the breakpoint.

[0050] Preferably, the abnormal state is classified into AGV occupied, path blocked, power abnormality, execution timeout or communication abnormality according to the abnormality type, specifically including:

[0051] When a signal is received that the AGV is temporarily occupied by other systems or manual operations, it is determined to be in the AGV occupied state;

[0052] When an obstacle is detected in the AGV's forward path and it cannot be bypassed, it is considered to be in a path-blocked state;

[0053] When the AGV power decreases abnormally or falls below the safety threshold, it is determined to be in an abnormal power state;

[0054] When the time it takes for the AGV to execute a task exceeds the expected time by more than 20%, it is considered to be in execution timeout state;

[0055] When no status feedback information from the AGV is received for three consecutive times or the communication quality is lower than the preset threshold, it is determined to be a communication abnormality state.

[0056] Preferably, the dispatch instruction is securely encapsulated and sent to the target AGV, and the target AGV performs multi-level security verification on the received dispatch instruction before executing the task, specifically including:

[0057] Generate standardized dispatch instructions containing instruction unique identifier, instruction type code, target AGV identification, instruction parameter set, instruction priority, sending timestamp, expected response time and sender identification;

[0058] Perform data verification and calculation and generate digital signatures for scheduling instructions to ensure the integrity and authenticity of the instructions;

[0059] Send dispatch instructions to the target AGV through a secure communication channel;

[0060] After receiving the dispatch instruction, the target AGV first checks the integrity of the instruction data to ensure that there are no transmission errors;

[0061] Verify whether the identity of the instruction sender is an authorized scheduling system;

[0062] Check the timestamp of the instruction to prevent replay attacks;

[0063] Analyze whether the format of the instruction complies with the predetermined specifications;

[0064] Check whether the instruction content is within the capabilities of the AGV;

[0065] Assess whether there are security risks in executing the instruction;

[0066] Check whether the instruction is consistent with the current state of AGV;

[0067] Based on the verification results, a response strategy of accepting, warning, executing, or rejecting the instruction is executed.

[0068] Preferably, the basic status information in the AGV status information includes the AGV unique identifier, current position coordinates and motion status; the functional parameter information includes a list of supported material types, maximum load weight and maximum load volume; the resource status information includes power level, running time and maintenance cycle status; the historical performance information includes historical task completion rate, average delivery time and abnormality rate.

[0069] Preferably, the weight coefficients of the distance score, power score, load balancing score, and time efficiency score are dynamically adjusted according to the current system state and task characteristics, specifically including:

[0070] During system idle periods, the weight of the load balancing score is increased to promote balanced use of AGV resources;

[0071] During busy periods, the weight of the time efficiency score is increased, giving priority to AGVs with high execution efficiency.

[0072] For urgent tasks, increase the weight of the distance score and give priority to AGVs with short distances;

[0073] For long-distance tasks, increase the weight of the power score to ensure that the AGV has enough power to complete the task;

[0074] Based on historical task execution data, the basic weight configuration of each scoring dimension is regularly updated;

[0075] It supports presetting multiple sets of weight configuration schemes according to the needs of different business scenarios to achieve fast switching.

[0076] AGV cluster scheduling system, including:

[0077] An order management module is used to obtain multi-source order information, convert the multi-source order information into a standard order format, generate order priorities based on time urgency, order source weight, material importance, and historical delivery evaluation, and create delivery tasks;

[0078] A feasibility analysis module is used to obtain delivery constraints, perform feasibility analysis on the delivery task based on physical distance constraints, time window constraints, resource availability constraints, and path accessibility constraints, and generate executable tasks;

[0079] An AGV scheduling module is used to obtain AGV status information and select the most suitable AGV to execute the executable task based on the executable task and the AGV status information through a multi-dimensional attribute matching mechanism;

[0080] The status monitoring module is used to monitor the execution status of the AGV in real time. When an abnormal state is detected, it saves the task context and triggers the AGV scheduling module to reselect the AGV to continue the task;

[0081] The instruction management module is used to generate a dispatch instruction, securely encapsulate the dispatch instruction, and send it to the target AGV;

[0082] Communication interface module, used to exchange data with AGV, receive AGV status feedback and forward it to the status monitoring module;

[0083] State storage module, used to store order information, task information, AGV status information and system configuration information;

[0084] The AGV cluster scheduling system realizes intelligent scheduling of the AGV cluster by executing the AGV cluster scheduling method.

[0085] The beneficial effects of the present invention include:

[0086] 1. Through the intelligent fusion mechanism of multi-source orders, unified processing of orders from different channels is achieved, improving the business adaptability of the system and enabling flexible allocation of priorities based on order characteristics;

[0087] 2. Through intelligent feasibility analysis based on the delivery range, infeasible tasks are filtered out in advance to avoid wasted resources and provide users with more accurate delivery information;

[0088] 3. Through a multi-dimensional attribute matching AGV selection mechanism that comprehensively considers factors such as distance, power consumption, load balancing, and efficiency, optimal resource allocation is achieved, reducing average task completion time by 35% and the AGV idle rate by approximately 40%.

[0089] 4. Through the dynamic task reallocation mechanism, it effectively responds to abnormal situations such as AGV occupation and path obstruction, increasing the task execution success rate to 99.5% and reducing the average fault recovery time by 65%;

[0090] 5. Through a multi-level security verification mechanism, the safe and reliable execution of scheduling instructions is ensured, effectively preventing misoperation and malicious attacks.

[0091] In general, the present invention builds a complete intelligent distribution scheduling solution, which significantly improves the scheduling efficiency, system stability and resource utilization of AGV clusters. BRIEF DESCRIPTION OF THE DRAWINGS

[0092] Figure 1 This is the overall architecture diagram of the AGV cluster scheduling system of the present invention;

[0093] Figure 2 Create a flow chart for the intelligent fusion of multi-source orders and tasks of the present invention;

[0094] Figure 3 This is a flow chart of intelligent feasibility analysis based on the delivery range of the present invention;

[0095] Figure 4 AGV selection flow chart for multi-dimensional attribute matching of the present invention;

[0096] Figure 5 This is a flowchart of AGV task reallocation in a dynamic environment of the present invention;

[0097] Figure 6 This is a flowchart of the security verification and intelligent execution of scheduling instructions of the present invention. DETAILED DESCRIPTION

[0098] Please refer to the attached Figure 1-6 , the specific implementation of the present invention is further described in detail below with reference to the accompanying drawings.

[0099] Reference Figure 2-6 ,This embodiment provides an AGV cluster scheduling method, including the steps of intelligent fusion and task creation of multi-source orders, intelligent feasibility analysis based on delivery range, intelligent selection of AGVs based on multi-dimensional attribute matching, AGV task reallocation in a dynamic environment, and security verification and intelligent execution of scheduling instructions.

[0100] First, the present invention obtains multi-source order information, which includes terminal system orders, mobile client orders, inventory trigger orders and feedback trigger orders. In actual applications, terminal system orders are usually initiated directly by the warehouse management terminal for internal material distribution; mobile client orders are initiated by users through mobile applications for external distribution services; inventory trigger orders are automatically generated by the warehousing system based on inventory conditions for replenishment needs; feedback trigger orders are generated based on user delivery feedback and are usually used to handle special service needs. For example, within a manufacturing plant, there may be terminal system orders for production lines requesting raw materials, mobile client orders for employees applying for office supplies through mobile phone APPs, inventory trigger orders automatically generated by the automatic warehousing system when it detects insufficient inventory of a certain component, and feedback trigger orders for the delivery of returned items based on user complaints.

[0101] Next, the present invention converts the multi-source order information into a standard order format, generates order priorities based on time urgency, order source weight, material importance, and historical delivery evaluation, and creates delivery tasks. Figure 2 As shown in Figure 1, the process first converts heterogeneous orders from different sources into a standard order format that contains order unique identifier, order source type, order creation timestamp, delivery point coordinate information, material type, weight, volume, and expected delivery time window.

[0102] Preferably, the order priority is calculated using a multi-factor weighted approach, and the specific formula is as follows:

[0103] ,

[0104] in: The comprehensive priority score of the order, ranging from 0 to 100, with a larger value indicating a higher priority; The time urgency score is calculated by the difference between the expected delivery time and the current time. The calculation formula is: ,in For the expected delivery time, is the current time, The maximum time difference allowed by the system, usually set to 24 hours; is the order source weight. Orders from different sources have different basic weights. For example, the weight of inventory-triggered orders is 0.9, the weight of terminal system orders is 0.7, and the weight of mobile client orders is 0.5. The weight of feedback-triggered orders fluctuates between 0.3 and 0.9 according to the urgency of the feedback. M is the material importance score, which is calculated based on the importance coefficient corresponding to the material type. For example, production-critical materials are 100 points, ordinary materials are 60 points, and low-priority materials are 30 points. H is the historical evaluation impact factor, which is calculated based on the user's historical delivery satisfaction evaluation. The calculation formula is: ,in The average rating of the user's last 10 deliveries, ranging from 1 to 5 points; 、 、 、 are the weight coefficients of each factor, and satisfy .

[0105] In this embodiment, the weight coefficient can be configured according to actual business needs. For example, in the AGV distribution system of a factory, it is usually set to 、 、 、 The reason for this configuration is that in a factory environment, time is often the most critical factor, followed by the importance of the order source, then the importance of the material, and the least impact of historical evaluation. Based on the calculated comprehensive priority score, the system generates a delivery task containing a unique task identifier, an associated order identifier, a task priority, the coordinates of the delivery start and end points, material requirement parameters, and a task deadline, and places it in the task queue.

[0106] Subsequently, the present invention obtains the delivery constraints, performs feasibility analysis on the delivery task based on physical distance constraints, time window constraints, resource availability constraints and path accessibility constraints, and generates executable tasks. Figure 3 As shown, the system performs a multi-dimensional feasibility analysis on each task in the task queue.

[0107] During the physical distance constraint analysis process, the system determines the distance between the delivery point and the nearest AGV station and compares it to the preset maximum delivery radius. In practice, the maximum delivery radius is typically determined by the AGV's range and service range. For example, it might be set at 500 meters in a factory environment, 1000 meters in a large logistics center, and only 100 meters in a small warehouse. For example, if a delivery point is located in a corner of a factory, 600 meters from the nearest AGV station, and the system's maximum delivery radius is 500 meters, the task does not meet the physical distance constraint, and the system will send an "out of service range" notification to the user.

[0108] In the time window constraint analysis, the system calculates the estimated completion time based on the current available AGV's moving speed and current position. The calculation formula for the estimated completion time is:

[0109] ,

[0110] in: Estimated completion time, which indicates the time point when the AGV is expected to complete the entire task; is the current time, that is, the time point when the system performs calculations; is the total travel distance (unit: meter), including the distance from the AGV's current position to the starting point and the distance from the starting point to the end point, usually calculated using Euclidean distance or Manhattan distance; The average moving speed of the AGV (unit: m / s), usually 1.2-2.0 m / s, varies depending on the AGV model and working environment. For example, it can reach 2.0 m / s on a flat factory floor, but may drop to 1.2 m / s in a complex environment. Loading time (unit: seconds), which is approximately 30-120 seconds depending on the type of material. For example, small items take about 30 seconds, while large items or items that require precise positioning may take 120 seconds. The unloading time (in seconds) ranges from 30 to 90 seconds, usually slightly shorter than the loading time. The system compares the estimated completion time with the task deadline to determine whether the time window constraint is met. For example, an AGV is currently located in area A of a factory and needs to pick up goods from area B and deliver them to area C. The total distance is 300 meters, the AGV speed is 1.5 m / s, the loading time is 60 seconds, and the unloading time is 45 seconds. The estimated completion time is the current time plus 305 seconds. If this time is later than the task deadline, the time window constraint is not met.

[0111] During resource availability constraint analysis, the system checks whether any available AGVs are capable of handling specific material types, weights, and volumes. For example, for materials weighing more than 200 kilograms, the system needs to confirm the availability of heavy-duty AGVs; for special materials such as fragile items, the system needs to confirm the availability of AGVs equipped with anti-vibration devices; and for liquid materials, the system needs to confirm the availability of AGVs with sealing and leak-proof functions. In a hospital logistics scenario, if medicines need to be delivered, the system needs to confirm the availability of AGVs with temperature control functions, and the number of such AGVs must be no less than the number required for the task.

[0112] During path accessibility constraint analysis, the system analyzes whether there are obstacles or restricted areas along the route from the delivery origin to the destination. This path analysis uses the A* or Dijkstra algorithm, combined with environmental map data, for planning. For example, within a factory, if the direct route from Area A to Area C is temporarily closed due to equipment maintenance, the system will attempt to find an alternative route. If no valid alternative route can be found or the alternative route is too long (exceeding 150% of the direct route), the path accessibility constraint is determined to be unmet.

[0113] When all constraints are met, the system updates the task status to "executable" and adds it to the executable task queue; when the physical distance constraint or path accessibility constraint is not met, the system sends a "non-deliverable" notification to the user and explains the specific reason; when the time window constraint is not met, the system provides the user with feasible alternative time options, such as postponing the estimated delivery time by 30 minutes or 1 hour; when the resource availability constraint is temporarily not met, the system marks the task as "waiting for resources" and re-evaluates it every 30 seconds until resources are available or the waiting time exceeds a preset threshold (such as 20 minutes), then cancels the task.

[0114] Next, the present invention obtains AGV status information, which includes basic status information, functional parameter information, resource status information, and historical performance information. AGV status information is an important basis for AGV selection, and the system obtains this information through real-time collection and processing.

[0115] Basic status information includes the AGV's unique identifier, current position coordinates, and motion state. The AGV's unique identifier is typically a combination of letters and numbers, such as "AGV001" or "AGV002." The current position coordinates are expressed in two-dimensional or three-dimensional coordinates, such as (x,y) in a flat factory environment or (x,y,z) in a multi-story warehouse. Coordinate units are typically meters. Motion states include stationary, accelerating, constant speed, decelerating, and turning. The system determines the AGV's motion state based on changes in speed and direction.

[0116] Functional parameter information includes a list of supported material types, maximum load weight, and maximum load volume. The material type list is a collection of material types that the AGV can handle, such as "general goods, fragile goods, liquids, and hazardous materials." The maximum load weight is typically the AGV's rated load, such as "100kg," "300kg," or "500kg." Different AGV models have different load capacities. The maximum load volume is the maximum space the AGV can load, such as "0.5m³," "1m³," or "2m³." This depends on the size of the AGV's loading platform.

[0117] Resource status information includes battery level, elapsed time, and maintenance cycle status. The battery level is typically expressed as a percentage, such as "85%" indicating 85% of the battery is available. Elapsed time is the continuous operating time of the AGV during the current shift, measured in hours, and is used to assess AGV fatigue. The maintenance cycle status indicates the time until the next routine maintenance due, and is used to assess AGV reliability. It is typically expressed in hours or days remaining.

[0118] Historical performance information includes historical task completion rates, average delivery time, and exception rates. The historical task completion rate is the percentage of tasks that the AGV successfully completes. For example, a value of "98%" indicates that the AGV has a high task success rate. The average delivery time is the average time it takes the AGV to complete a task and is used to evaluate AGV efficiency. The exception rate is the frequency of exceptions encountered during the AGV's task execution. For example, a value of "2%" indicates that the AGV rarely encounters exceptions.

[0119] Based on the executable tasks and the AGV status information, the present invention selects the most suitable AGV to execute the executable tasks through a multi-dimensional attribute matching mechanism, which includes pre-screening, distance scoring, power scoring, load balancing scoring and time efficiency scoring. Figure 4 As shown, the AGV selection process is divided into multiple steps.

[0120] First, the system selects a candidate set of AGVs from the AGV cluster that are capable of performing the task based on the material type, weight, and volume requirements. For example, for a material weighing 350 kg, the system selects AGVs with a maximum load capacity of at least 350 kg; for a material with a volume of 1.5 cubic meters, the system selects AGVs with a maximum load capacity of at least 1.5 cubic meters; and for fragile materials, the system selects AGVs equipped with shockproof features. For example, in a factory, if there is a task to transport a precision instrument weighing 120 kg and with a volume of 0.8 cubic meters, the system will select 5 AGVs with shockproof features, a load capacity exceeding 120 kg, and a load volume exceeding 0.8 cubic meters from a pool of 10 AGVs as candidates.

[0121] The system then calculates the distance from each candidate AGV to the mission start and end points, and generates a distance score based on the inverse of the distance. The distance score calculation formula is:

[0122] ,

[0123] in: is the distance score, ranging from 0 to 1, where a larger value indicates a better AGV position; The distance from the AGV's current position to the mission starting point (unit: meter), calculated using the actual path distance or Euclidean distance; is the distance from the starting point to the end point of the task (unit: meter); and are the weight coefficients of the starting distance and the end distance, respectively. , This setting is because the distance from the AGV to the starting point has a greater impact on the task start time, while the distance from the starting point to the end point has a greater impact on the total task time. Considering both can more accurately evaluate the AGV's position advantage. For example, if AGV1 is currently 200 meters away from the task starting point, and the distance from the starting point to the end point is 300 meters; AGV2 is 100 meters away from the starting point, and the distance from the starting point to the end point is 300 meters; then AGV1's distance score is 0.6×(1 / 200)+0.4×(1 / 300) / (0.6+0.4)=0.0043, and AGV2's distance score is ,AGV2 has a higher distance score and a better position.

[0124] Next, the system evaluates whether the power level of each candidate AGV is sufficient to complete the entire task and generates a power score based on the remaining power. The power score calculation formula is:

[0125] ,

[0126] in: The battery rating ranges from 0 to 1, with a larger value indicating a better battery condition. The current battery level of the AGV is in the range of 0-100%. The basic power percentage required to complete the task is calculated by Calculate, where is the total driving distance (unit: meter), The maximum driving distance of the AGV when fully charged (unit: meter); The safety margin power percentage is usually 20%, ensuring that the AGV has enough power to cope with complex road conditions and return to the charging station after completing the task; It is fully charged, that is, 100%.

[0127] When the required power plus the safety margin is less than the full power, the score is based on the ratio of the remaining available power to the total available power. When the required power plus the safety margin exceeds the full power, the score is based on the ratio of the excess to the safety margin, ensuring that the formula can reasonably evaluate the power status of the AGV in all situations.

[0128] For example, if the current battery level of an AGV is 85% and the total mission distance is 10 kilometers, and the AGV can travel 20 kilometers on a full charge, then the basic battery level required to complete the mission is (10 / 20)×100%=50%, plus a 20% safety margin of 70%. Since 85%>(50%+20%) and (50%+20%)<100%, the battery level score of the AGV is (85%-(50%+20%)) / (100%-(50%+20%))=0.5, indicating that the battery level is moderate.

[0129] Then, the system counts the number of tasks recently executed by each candidate AGV and generates a load balancing score based on the balance of task distribution. The load balancing score calculation formula is:

[0130] ,

[0131] in: The load balance score ranges from 0 to 1, with a larger value indicating a more balanced AGV load. The number of tasks that the AGV has performed recently (e.g., in the past 4 hours); The maximum number of tasks executed by a single AGV in the same period. For example, among the 10 AGVs in a factory, the number of tasks executed in the past 4 hours were 5, 8, 3, 6, 7, 4, 9, 2, 6, and 5 respectively. , the load balancing score of AGV8 is , while the load balancing score of AGV7 is , which shows that AGV8 has a lower task load and is more suitable for assigning new tasks.

[0132] Finally, the system generates a time efficiency score based on the AGV's moving speed and historical task completion efficiency. The time efficiency score calculation formula is:

[0133] ,

[0134] in: It is a time efficiency score ranging from 0 to 1, where a larger value indicates a higher AGV efficiency; is the moving speed of AGV (unit: m / s); is the maximum moving speed of the AGV in the cluster (unit: m / s); The historical task completion efficiency of the AGV is usually expressed as the ratio of expected completion time to actual completion time. For example, 0.95 means that the actual time is slightly longer than expected. The highest historical task completion efficiency of AGV in the cluster; and are the weight coefficients of speed and efficiency, respectively. , indicating that both are equally important. For example, the moving speed of an AGV is 1.8 m / s, while the fastest AGV in the cluster is 2.0 m / s. The historical task completion efficiency of the AGV is 0.92, and the highest efficiency in the cluster is 0.98. Then the time efficiency score of the AGV is .

[0135] Based on the current system status and task characteristics, the system dynamically adjusts the weight coefficients of each scoring dimension. During idle periods (e.g., when the number of tasks in the task queue is less than 20% of the total number of AGVs), the system increases the weight of the load balancing score (e.g., from 0.2 to 0.4) to promote balanced use of AGV resources. During busy periods (e.g., when the number of tasks in the task queue exceeds 80% of the total number of AGVs), the system increases the weight of the time efficiency score (e.g., from 0.2 to 0.4), prioritizing AGVs with high execution efficiency. For urgent tasks (e.g., tasks in the top 10% of priority), the system increases the weight of the distance score (e.g., from 0.3 to 0.5), prioritizing AGVs with close proximity. For long-distance tasks (e.g., delivery distances exceeding 500 meters), the system increases the weight of the battery life score (e.g., from 0.2 to 0.4), ensuring that AGVs have sufficient power to complete the task.

[0136] The comprehensive score calculation formula is:

[0137] ,

[0138] in It is a comprehensive score ranging from 0 to 1, where a larger value indicates that the AGV is more suitable for the task; are the weight coefficients of distance score, power score, load balancing score and time efficiency score, and satisfy In a standard configuration, these weighting factors are usually set to 、 、 、 , indicating that distance and efficiency are the main considerations, followed by power and load balancing.

[0139] The system selects the AGV with the highest overall score as the target AGV and generates an assignment result containing task information and execution requirements. If multiple AGVs have the same overall score (score difference less than 0.01), the system selects the AGV with the lower number to ensure the certainty of the result. For example, in a factory environment, for a task to transport raw materials, the system may calculate that AGV2 has an overall score of 0.78, AGV5 has an overall score of 0.65, and AGV8 has an overall score of 0.71. The system will select AGV2, which has the highest score, to perform the task.

[0140] During the task execution process, the present invention monitors the execution status of the AGV in real time. When an abnormal state such as AGV being occupied, path blocked, power abnormality, execution timeout or communication abnormality is detected, the task context is saved and the AGV is reselected to continue the task. Figure 5 As shown in the figure, the system has designed a complete exception handling mechanism.

[0141] The system receives real-time feedback on AGV status and identifies abnormal conditions when AGV status parameters exceed preset thresholds. For example, if the AGV's position hasn't updated for more than 10 seconds, this could indicate a communication anomaly; if the AGV's battery level drops by more than 150% of its normal value, this could indicate a battery anomaly; and if the AGV remains in the same location for more than 30 seconds without loading or unloading, this could indicate a blocked path. These thresholds are set based on actual operational experience and can be adjusted to suit different environments and AGV models.

[0142] The system categorizes abnormal conditions into different situations based on the type of anomaly. When a signal is received that an AGV is temporarily occupied by another system or manual operation, the system identifies it as occupied. This is common in environments where multiple systems share AGV resources, such as when an AGV is temporarily deployed by an emergency maintenance system. When an obstacle is detected in the AGV's forward path and it cannot be circumvented, the system identifies it as blocked. For example, in a factory, a passageway might be blocked by a temporary pile of materials. When the AGV's battery level drops abnormally (e.g., a normal 5% per hour drop exceeds 8%) or falls below a safety threshold (typically 30%), the system identifies it as abnormal. This may be due to AGV battery aging or excessive load. When an AGV takes more than 20% longer to complete a task, the system identifies it as timed out. For example, if a task originally scheduled for 30 minutes has not been completed after 36 minutes, the system identifies it as a communication anomaly. This may be due to network issues or a malfunctioning AGV communication module.

[0143] After detecting an anomaly, the system saves task context information, including the current task's execution progress, the AGV's location coordinates, a list of completed subtasks, and a list of remaining subtasks. This information is encapsulated as a "task context" data packet, which contains sufficient information for a new AGV to resume the task from the point where it left off. For example, in a pickup and delivery task, if an AGV has already completed the pickup but encounters an anomaly midway, the system saves information such as the pickup, current location, and destination, allowing the newly assigned AGV to resume the delivery directly from its current location.

[0144] The system then sends a task interruption command to the original AGV to confirm the interruption status. This command contains information such as the task ID, the reason for the interruption, and the interruption time. The system waits for the AGV to return an interruption confirmation message. If it does not receive confirmation within 10 seconds, it resends the command up to three times. This design ensures that the AGV has received the interruption command and prevents multiple AGVs from executing the same task simultaneously.

[0145] The system then evaluates the task's current status and determines whether to continue or restart from the current location. For partially completed tasks (e.g., loading completed but not yet transported), the system typically chooses to continue from the current location. For tasks that have just begun (e.g., not yet reached the starting point), the system may choose to restart. This decision is based on a combination of factors, including the task type, completion progress, and the new AGV's location.

[0146] The system then re-invokes its multi-dimensional attribute matching mechanism to select a new, suitable AGV. To avoid selecting an AGV with the same potential for problems, the system adjusts its selection criteria appropriately. For example, in the case of blocked paths, the system increases the score of AGVs with strong path planning capabilities; in the case of abnormal battery life, the system increases the score of AGVs with good battery management records. For example, in a factory, if AGV3 is interrupted during a mission due to an abnormal battery life, the system will specifically consider AGVs with good battery conditions, such as AGV6 and AGV9, when reselecting, and ultimately select the one with the higher overall score.

[0147] The system transfers the saved task context information to the newly selected AGV, adjusts the execution plan for the remaining tasks based on the characteristics of the new AGV, and sends instructions to the newly selected AGV to resume the task from the point where it stopped. For example, if the original AGV stopped at a corner, the system will use that location as the starting point for the new AGV's task and plan the optimal path from that point to the destination.

[0148] Finally, the present invention generates a dispatch instruction, securely encapsulates the dispatch instruction and sends it to the target AGV, which performs a multi-level security check on the received dispatch instruction and then executes the task. Figure 6As shown, the system has designed a complete instruction security mechanism.

[0149] The system generates standardized scheduling instructions containing the instruction unique identifier, instruction type code, target AGV identification, instruction parameter set, instruction priority, sending timestamp, expected response time and sender identity. The instruction unique identifier is typically a system-generated UUID (Universally Unique Identifier), such as "XYZ." The instruction type code identifies the instruction type, such as "001" for task assignment, "002" for path adjustment, and "003" for status query. The instruction parameter set contains specific execution parameters, such as the task start coordinates (120.5, 56.8), end coordinates (230.2, 78.4), and the material ID "MAT2458." The instruction priority indicates the urgency of the instruction and is typically divided into five levels, with level 1 being the highest and level 5 being the lowest. The sending timestamp records the exact time the instruction was sent, such as "2025-05-29T10:15:30.123Z." The expected response time specifies how long the AGV should respond to the instruction, such as "5 seconds." The sender identity is used to verify the legitimacy of the instruction sender, such as "SYS_SCHEDULER."

[0150] The system performs a data checksum calculation and digital signature on the dispatch instruction to ensure its integrity and authenticity. The data checksum is calculated using the CRC32 or MD5 algorithm, such as "EFG". The digital signature uses the RSA or ECDSA algorithm and the system private key to sign the instruction digest, generating a signature string such as "ABC".

[0151] Scheduling instructions are sent to the target AGV via a secure communication channel (such as a TLS-encrypted channel). During communication, the system monitors the data transmission status and resends the instruction if an anomaly is detected (e.g., packet loss exceeding 1%). This monitoring mechanism is essential in industrial environments, where electromagnetic interference or network congestion can affect communication quality.

[0152] After receiving a dispatch instruction, the target AGV first checks the integrity of the instruction data to ensure there are no transmission errors. The AGV recalculates the checksum of the received data and compares it with the checksum included in the instruction. If they do not match, it determines a transmission error and requests a retransmission. For example, if an AGV receives a instruction and calculates a checksum of "EFG," which matches the checksum included in the instruction, the integrity check passes.

[0153] The AGV then verifies that the instruction sender is an authorized dispatching system. The AGV uses the pre-stored dispatching system public key to verify the digital signature, confirming that the instruction originated from the legitimate dispatching system and preventing forged instruction attacks. In a factory environment, this verification mechanism prevents unauthorized systems (such as hacked terminals) from sending malicious instructions to the AGV.

[0154] Next, the AGV checks the command's timestamp to prevent replay attacks. If the difference between the command's timestamp and the current time exceeds a preset threshold (typically 30 seconds), the AGV rejects the command, preventing attackers from intercepting and replaying older commands. For example, if the AGV receives a command with a timestamp of "2025-05-29T10:15:30.123Z" and the current time is "2025-05-29T10:20:45.789Z," the difference exceeds 30 seconds, and the AGV rejects the command.

[0155] The AGV then analyzes the instruction format to see if it complies with predefined specifications, checks whether the instruction content is within the AGV's capabilities, assesses whether executing the instruction poses a safety risk, and checks whether the instruction is consistent with the AGV's current state. For example, if the instruction requires the AGV to move 500 kg of material, but the AGV's maximum load capacity is 300 kg, the AGV will reject the instruction; if the instruction requires the AGV to go to an area that is prohibited, the AGV will also refuse to execute it.

[0156] Finally, based on the verification results, the AGV executes a response strategy: accept, execute with a warning, or reject the command. If all verifications pass, the AGV accepts the command and begins executing it, sending an acceptance confirmation to the dispatch system. If a non-critical verification fails (e.g., a minor timestamp timeout of 5 seconds), the AGV sends a warning but continues to execute the command. If a critical verification fails (e.g., identity verification), the AGV rejects the command and returns a detailed failure reason, such as "Signature verification failed, the command may have been tampered with."

[0157] Reference Figure 1 This embodiment provides an AGV cluster scheduling system, including an order management module 1, a feasibility analysis module 2, an AGV scheduling module 3, a status monitoring module 4, an instruction management module 5, a communication interface module 6 and a status storage module 7.

[0158] Order management module 1 is used to obtain multi-source order information, convert it into a standard order format, generate order priorities based on time urgency, order source weight, material importance, and historical delivery evaluation, and create delivery tasks. Order management module 1 includes order acquisition unit 11, order standardization unit 12, priority calculation unit 13, and task creation unit 14.

[0159] The order acquisition unit 11 obtains raw order data from the terminal system, mobile client, warehouse system, and user feedback system through an API interface or message queue. Order data from different sources has different formats and needs to be processed uniformly. For example, the terminal system may use XML format, the mobile client may use JSON format, the warehouse system may use CSV format, and the user feedback system may use a custom binary format.

[0160] The order standardization unit 12 converts the acquired raw order data into a system-standard format. For example, it converts dates and times in various formats into ISO 8601 format (e.g., "2025-05-29T10:15:30Z"), converts weights in various units into kilograms, and converts volumes in various units into cubic meters. The priority calculation unit 13 calculates the priority score for each order using the priority calculation formula described in Example 1. Based on the standardized order and priority score, the task creation unit 14 creates a delivery task and places it into the task queue.

[0161] Feasibility Analysis Module 2 is used to obtain delivery constraints and perform feasibility analysis on the delivery task based on physical distance constraints, time window constraints, resource availability constraints, and path accessibility constraints to generate executable tasks. Feasibility Analysis Module 2 includes a constraint acquisition unit 21, a multi-dimensional constraint analysis unit 22, and a task status update unit 23.

[0162] The constraint acquisition unit 21 retrieves various constraint parameters from system configuration and environmental data, such as the maximum delivery radius and average AGV speed. In a factory application, these parameters might include a maximum delivery radius of 500 meters, an average AGV speed of 1.5 meters per second, and a maximum waiting time of 20 minutes. The multidimensional constraint analysis unit 22 performs constraint analysis across four dimensions: physical distance, time window, resource availability, and path accessibility. The task status update unit 23 updates the task status based on the analysis results and adds executable tasks to the execution queue.

[0163] The AGV scheduling module 3 is used to obtain AGV status information and, based on the executable tasks and the AGV status information, select the most suitable AGV to perform the executable tasks through a multi-dimensional attribute matching mechanism. The AGV scheduling module 3 includes a status acquisition unit 31, a candidate AGV screening unit 32, a multi-dimensional scoring unit 33, and an AGV selection unit 34.

[0164] The status acquisition unit 31 obtains real-time AGV status information through the communication interface module 6. For example, it sends a status query request to all AGVs every 2 seconds and receives a response containing information such as location, power level, and motion status. The candidate AGV screening unit 32 selects a set of candidate AGVs capable of executing the task based on the task requirements. The multidimensional scoring unit 33 calculates the distance score, power level score, load balancing score, and time efficiency score for each candidate AGV and dynamically adjusts the weights based on the system status. The AGV selection unit 34 selects the optimal AGV based on the comprehensive score and generates an allocation result.

[0165] The state monitoring module 4 is used to monitor the AGV's execution status in real time. When an abnormal state is detected, it saves the task context and triggers the AGV scheduling module 3 to reselect an AGV to continue the task. The state monitoring module 4 includes a state monitoring unit 41, an abnormality identification unit 42, a context storage unit 43, and a reallocation triggering unit 44.

[0166] The status monitoring unit 41 periodically (typically every second) obtains status feedback from the AGVs. In one factory implementation, this unit might be configured to collect approximately 20KB of status data per second from 20 AGVs, including location, speed, and battery level. The anomaly identification unit 42 identifies abnormal conditions based on preset thresholds and rules. The context storage unit 43 saves task context information when an anomaly is detected. The reallocation trigger unit 44 sends a reallocation request to the AGV scheduling module 3, triggering the reselection process.

[0167] The instruction management module 5 is used to generate a dispatch instruction, securely encapsulate the dispatch instruction, and send it to the target AGV. The instruction management module 5 includes an instruction generation unit 51, a secure encapsulation unit 52, and an instruction sending unit 53.

[0168] The instruction generation unit 51 generates standardized dispatch instructions based on task assignment results or system control requirements. For example, for a task assignment instruction, a data structure containing parameters such as the task ID, start and end coordinates, and material information is generated. The security encapsulation unit 52 performs data verification and calculations on the instruction, digitally signing it to ensure its security. The instruction transmission unit 53 securely transmits the instruction to the target AGV via the communication interface module 6.

[0169] The communication interface module 6 is used to exchange data with the AGV, receive AGV status feedback and forward it to the status monitoring module 4. The communication interface module 6 includes a data sending unit 61, a data receiving unit 62 and a protocol conversion unit 63.

[0170] The data transmission unit 61 is responsible for sending system commands to the AGV via the network. For example, it uses the TCP / IP protocol to send command packets and ensure successful data transmission. The data reception unit 62 receives status feedback and response information from the AGV. The protocol conversion unit 63 converts between different communication protocols to ensure system compatibility with various AGV models. For example, some older AGV models may use the RS485 communication protocol, while newer models use Ethernet or wireless communication. The protocol conversion unit ensures that the system can communicate properly with all AGV models.

[0171] State Storage Module 7 stores order information, task information, AGV status information, and system configuration information. State Storage Module 7 utilizes a distributed database architecture to ensure high data availability and consistency. Key stored data includes order tables, task tables, AGV status tables, scheduling log tables, and system configuration tables. For example, in a large logistics center, this module processes approximately 100,000 order records and 500,000 AGV status records daily, retaining 30 days of historical data for analysis and optimization.

[0172] The AGV cluster scheduling system in this embodiment implements intelligent scheduling of the AGV cluster by executing the AGV cluster scheduling method described in Example 1. The modules work together through standardized interfaces and data flows to form a complete scheduling closed-loop system.

[0173] Preferably, the system supports presetting multiple sets of weight configuration schemes according to the needs of different business scenarios to achieve fast switching. For example, in the logistics and distribution scenario, three schemes can be preset: "efficiency priority", "balance priority" and "energy saving priority", which are suitable for scheduling needs during peak, off-peak and off-peak periods respectively. In the "efficiency priority" scheme, the time efficiency score weight can be set to 0.5, the distance score weight is 0.3, and the power score and load balancing score weights are 0.1 each; in the "balance priority" scheme, the load balancing score weight can be increased to 0.4, and the other three items are 0.2 each; in the "energy saving priority" scheme, the power score weight can be increased to 0.4, the distance score weight is 0.3, and the time efficiency and load balancing score weights are 0.15 each.

[0174] The system also boasts excellent scalability and adaptability, supporting the collaborative operation of different types of AGVs and adapting to the needs of diverse application scenarios such as factories, warehouses, hospitals, and ports. For example, in a hospital setting, the system can simultaneously dispatch general delivery AGVs, pharmaceutical-specific AGVs, and medical waste disposal AGVs, selecting the most appropriate AGV type based on the specific characteristics of each task.

[0175] In summary, the AGV cluster scheduling method and system provided by the present invention, through innovative technologies such as intelligent fusion of multi-source orders, intelligent feasibility analysis based on delivery range, AGV selection based on multi-dimensional attribute matching, dynamic task reallocation, and scheduling instruction security verification, have built a complete set of intelligent distribution scheduling solutions, significantly improving the scheduling efficiency, system stability and resource utilization of AGV clusters.

[0176] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be included within the scope of protection of the present invention.

Claims

1. AGV cluster scheduling method, characterized in that: include: Acquire multi-source order information, including terminal system orders, mobile client orders, inventory-triggered orders, and feedback-triggered orders; Convert the multi-source order information into a standard order format, generate order priorities based on time urgency, order source weight, material importance, and historical delivery evaluation, and create delivery tasks; Obtain delivery constraints, perform feasibility analysis on the delivery task based on physical distance constraints, time window constraints, resource availability constraints, and path accessibility constraints, and generate executable tasks; Obtaining AGV status information, including basic status information, function parameter information, resource status information, and historical performance information; Based on the executable tasks and the AGV status information, the most suitable AGV is selected to perform the executable tasks through a multi-dimensional attribute matching mechanism, wherein the multi-dimensional attribute matching mechanism includes pre-screening, distance scoring, power scoring, load balancing scoring, and time efficiency scoring; Monitor the execution status of the AGV in real time. When it is detected that the AGV is occupied, the path is blocked, the power is abnormal, the execution timeout or the communication is abnormal, the task context is saved and the AGV is reselected to continue the task. Generate a dispatch instruction, securely encapsulate the dispatch instruction and send it to the target AGV, which performs multi-level security checks on the received dispatch instruction and then executes the task.

2. The AGV cluster scheduling method according to claim 1, characterized in that: The process of converting multi-source order information into a standard order format, generating order priorities based on time urgency, order source weight, material importance, and historical delivery evaluation, and creating delivery tasks specifically includes: Convert order information from different sources into a standard order format that includes a unique order identifier, order source type, order creation timestamp, delivery point coordinates, material type, weight, volume, and expected delivery time window; Calculate the time urgency score based on the difference between the expected delivery time and the current time; Different basic weights are assigned based on the order source type. Inventory-triggered orders have a higher weight than terminal system orders, which in turn have a higher weight than mobile client orders. Calculate the material importance score based on the importance coefficient corresponding to the material type; Obtain users' historical delivery satisfaction ratings and calculate historical rating impact factors; Calculate a comprehensive priority score based on the time urgency score, order source weight, material importance score, and historical evaluation impact factor; According to the comprehensive priority score, a delivery task is generated including a task unique identifier, an associated order identifier, a task priority, delivery start and end point coordinates, material requirement parameters and a task deadline, and the delivery task is placed in a task queue.

3. The AGV cluster scheduling method according to claim 1, characterized in that: The feasibility analysis of the delivery task is performed based on the physical distance constraint, the time window constraint, the resource availability constraint, and the path accessibility constraint to generate an executable task, specifically including: Obtain the distance between the delivery point and the nearest AGV station and compare it with the preset maximum delivery radius to determine whether the physical distance constraint is met; Based on the current available AGV's moving speed and current position, the estimated completion time is calculated and compared with the task deadline to determine whether the time window constraint is met; Check whether there is an AGV among the currently available AGVs capable of handling the specific material type, weight, and volume to determine whether the resource availability constraints are met; Analyze whether there are obstacles or restricted areas on the path from the delivery start point to the destination, and determine whether the path accessibility constraints are met; When all constraints are met, the task status is updated to "executable" and added to the executable task queue; When the physical distance constraint or path accessibility constraint is not met, a "delivery not possible" notification is sent to the user; When the time window constraints are not met, provide users with feasible alternative time options; When resource availability constraints are temporarily unsatisfied, the task is marked as "waiting for resources" and re-evaluated periodically.

4. The AGV cluster scheduling method according to claim 1, characterized in that: The method of selecting the most suitable AGV to execute the executable task through the multi-dimensional attribute matching mechanism specifically includes: Based on the material type, weight, and volume requirements of the task, select a candidate set of AGVs with execution capabilities from the AGV cluster; Calculate the distance from each candidate AGV to the starting and ending points of the task, and generate a distance score based on the inverse of the distance; Evaluate whether the power level of each candidate AGV is sufficient to complete the entire mission and generate a power score based on the remaining power; Count the number of tasks recently executed by each candidate AGV and generate a load balancing score based on the balance of task distribution; Generate a time efficiency score based on the AGV's movement speed and historical task completion efficiency; Dynamically adjust the weight coefficients of distance score, power score, load balancing score, and time efficiency score based on the current system status and task characteristics; Perform weighted summation on each candidate AGV’s scores to generate a comprehensive score; The AGV with the highest comprehensive score is selected as the target AGV, and an assignment result containing task information and execution requirements is generated.

5. The AGV cluster scheduling method according to claim 1, characterized in that: When it is detected that the AGV is occupied, the path is blocked, the power is abnormal, the execution timeout or the communication is abnormal, the task context is saved and the AGV is reselected to continue the task, specifically including: Receive AGV status feedback information in real time, and identify it as an abnormal state when the AGV status parameters exceed the preset threshold; According to the abnormality type, the abnormal state is classified as AGV occupied, path blocked, power abnormality, execution timeout or communication abnormality; Save the execution progress of the current task, AGV location coordinates, the list of executed subtasks, and the list of remaining subtasks to be executed, as well as the task context information; Send a task interruption command to the original AGV to confirm the interruption status; Evaluate the current state of the task and decide whether to continue or restart from the current position; Re-call the multi-dimensional attribute matching mechanism and select a new suitable AGV; Pass the saved task context information to the newly selected AGV; Adjust the execution plan of the remaining tasks based on the characteristics of the new AGV; Send an instruction to the newly selected AGV to continue executing the task from the breakpoint.

6. The AGV cluster scheduling method according to claim 5, characterized in that: According to the abnormality type, the abnormal state is classified into AGV occupied, path blocked, power abnormality, execution timeout or communication abnormality, specifically including: When a signal is received that the AGV is temporarily occupied by other systems or manual operations, it is determined to be in the AGV occupied state; When an obstacle is detected in the AGV's forward path and it cannot be bypassed, it is considered to be in a path-blocked state; When the AGV power decreases abnormally or falls below the safety threshold, it is determined to be in an abnormal power state; When the time it takes for the AGV to execute a task exceeds the expected time by more than 20%, it is considered to be in execution timeout state; When no status feedback information from the AGV is received for three consecutive times or the communication quality is lower than the preset threshold, it is determined to be a communication abnormality state.

7. The AGV cluster scheduling method according to claim 1, characterized in that: The dispatch instruction is securely encapsulated and sent to the target AGV, and the target AGV performs multi-level security verification on the received dispatch instruction and then executes the task, specifically including: Generate standardized dispatch instructions containing instruction unique identifier, instruction type code, target AGV identification, instruction parameter set, instruction priority, sending timestamp, expected response time and sender identification; Perform data verification and calculation and generate digital signatures for scheduling instructions to ensure the integrity and authenticity of the instructions; Send dispatch instructions to the target AGV through a secure communication channel; After receiving the dispatch instruction, the target AGV first checks the integrity of the instruction data to ensure that there are no transmission errors; Verify whether the identity of the instruction sender is an authorized scheduling system; Check the timestamp of the instruction to prevent replay attacks; Analyze whether the format of the instruction complies with the predetermined specifications; Check whether the instruction content is within the capabilities of the AGV; Assess whether there are security risks in executing the instruction; Check whether the instruction is consistent with the current state of AGV; Based on the verification results, a response strategy of accepting, warning, executing, or rejecting the instruction is executed.

8. The AGV cluster scheduling method according to claim 1, characterized in that: The basic status information in the AGV status information includes the AGV unique identifier, current position coordinates and motion status; the functional parameter information includes a list of supported material types, maximum load weight and maximum load volume; the resource status information includes power level, running time and maintenance cycle status; the historical performance information includes historical task completion rate, average delivery time and abnormality rate.

9. The AGV cluster scheduling method according to claim 4, characterized in that: The weight coefficients of the distance score, power score, load balancing score, and time efficiency score are dynamically adjusted according to the current system status and task characteristics, specifically including: During system idle periods, the weight of the load balancing score is increased to promote balanced use of AGV resources; During busy periods, the weight of the time efficiency score is increased, giving priority to AGVs with high execution efficiency. For urgent tasks, increase the weight of the distance score and give priority to AGVs with short distances; For long-distance tasks, increase the weight of the power score to ensure that the AGV has enough power to complete the task; Based on historical task execution data, the basic weight configuration of each scoring dimension is regularly updated; It supports presetting multiple sets of weight configuration schemes according to the needs of different business scenarios to achieve fast switching. 10.AGV cluster scheduling system, characterized by: include: An order management module is used to obtain multi-source order information, convert the multi-source order information into a standard order format, generate order priorities based on time urgency, order source weight, material importance, and historical delivery evaluation, and create delivery tasks; A feasibility analysis module is used to obtain delivery constraints, perform feasibility analysis on the delivery task based on physical distance constraints, time window constraints, resource availability constraints, and path accessibility constraints, and generate executable tasks; An AGV scheduling module is used to obtain AGV status information and select the most suitable AGV to execute the executable task based on the executable task and the AGV status information through a multi-dimensional attribute matching mechanism; The status monitoring module is used to monitor the execution status of the AGV in real time. When an abnormal state is detected, it saves the task context and triggers the AGV scheduling module to reselect the AGV to continue the task; The instruction management module is used to generate a dispatch instruction, securely encapsulate the dispatch instruction, and send it to the target AGV; Communication interface module, used to exchange data with AGV, receive AGV status feedback and forward it to the status monitoring module; State storage module, used to store order information, task information, AGV status information and system configuration information; The AGV cluster scheduling system realizes intelligent scheduling of the AGV cluster by executing the AGV cluster scheduling method according to any one of claims 1 to 9.

Images