AGV task allocation method and system based on multi-objective optimization

Abstract

The invention relates to the technical field of computers, in particular to an AGV task allocation method and system based on multi-objective optimization, and the method comprises the steps: obtaining a task demand set and a state information set of AGV equipment in a current operation scene, and the task demand set comprises a plurality of task units; performing feature extraction processing on the task demand set and the state information set to obtain task feature description of each task unit and capability feature description of each AGV device; performing joint optimization matching processing according to the task feature description and the capability feature description to generate a matching association relationship set of the task and the AGV equipment, wherein the matching association relationship set comprises a corresponding relationship between the task unit and the AGV equipment and a task execution sequence relationship; and generating a task allocation scheme based on the matching association relationship set, wherein the task allocation scheme comprises a to-be-executed task list of each AGV device and execution time sequence arrangement information of each task. Therefore, task requirements and equipment states can be accurately grasped, and efficient and matched innovative task distribution processing is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer, more particularly, to an AGV task allocation method and system based on multi-objective optimization. BACKGROUND

[0002] In the operation scene of modern industry, especially in the field involving automated equipment operation, such as power battery factory, AGV (Automated Guided Vehicle) equipment is widely used in the execution of various tasks, such as carrying, sorting, inspection, etc. Reasonable and efficient task allocation can improve the operation efficiency and resource utilization of the whole operation system.

[0003] The existing AGV task allocation method mainly relies on simple rules and basic judgments. Some methods only allocate tasks according to the arrival time sequence, that is, the tasks that arrive first are allocated to idle AGV equipment first, without considering the specific characteristics of the tasks and the actual capacity of the AGV equipment. Some other methods allocate tasks according to the location of the AGV equipment, that is, tasks closer to the AGV equipment are allocated to it first, but the inherent needs of the tasks and the adaptability of the equipment are not analyzed in depth.

[0004] Based on the above existing technology, on the one hand, due to the lack of consideration of the specific characteristics of the tasks, the task allocation often cannot meet the actual needs of the tasks. On the other hand, the capacity characteristics of the AGV equipment are not fully evaluated, resulting in low utilization efficiency of the equipment. In addition, the existing method does not comprehensively consider the corresponding relationship between the task units and the AGV equipment and the task execution order relationship, which is easy to cause task conflict and path congestion, further reducing the operation efficiency of the operation system.

[0005] In summary, the existing AGV task allocation technology has the problems of unreasonable task allocation, low equipment utilization, and low operation efficiency, which cannot adapt to the needs of complex and variable operation scenes. SUMMARY

[0006] Therefore, the present application provides an AGV task allocation method and system based on multi-objective optimization.

[0007] The embodiment of the present application provides an AGV task allocation method based on multi-objective optimization, which is applied to an AGV task allocation system, and the method comprises: obtaining a task demand set and a state information set of AGV equipment under a current operation scene, wherein the task demand set comprises a plurality of task units; performing feature extraction processing on the task demand set and the state information set to obtain task feature descriptions of each task unit and capacity feature descriptions of each AGV equipment; The joint optimization matching processing is performed according to the task feature description and the capability feature description, a matching association relationship set of the task and the AGV device is generated, and the matching association relationship set includes a corresponding relationship of a task unit and the AGV device and a task execution sequence relationship; A task allocation scheme is generated based on the matching association relationship set, and the task allocation scheme includes a to-be-executed task list of each AGV device and execution timing arrangement information of each task.

[0008] The application further provides an AGV task allocation system, including a memory for storing program instructions and data, and a processor coupled with the memory and executing the instructions in the memory to implement the method as above.

[0009] The application further provides a computer storage medium containing instructions, which, when executed on a processor, implement the method as above.

[0010] The embodiments of the application realize accurate grasping and efficient matching of task requirements and AGV device states in a work scene, and significantly improve the scientificity and rationality of task allocation. By obtaining a task requirement set and a state information set of the AGV device under the current work scene, the work requirements and the device states can be comprehensively understood; feature extraction processing is performed on the task requirement set and the state information set to obtain a task feature description of each task unit and a capability feature description of each AGV device, deep mining of task and device characteristics is realized, and the subsequent matching is more targeted; joint optimization matching processing is performed according to the task feature description and the capability feature description, a matching association relationship set of the task and the AGV device is generated, the corresponding relationship of the task unit and the AGV device and the task execution sequence relationship are comprehensively considered, the blindness and randomness of task allocation are avoided, and the efficiency and coordination of task execution are improved; a task allocation scheme is generated based on the matching association relationship set, the to-be-executed task list of each AGV device and the execution timing arrangement information of each task are determined, the task allocation is clearer and more orderly, device idling and task conflicts are reduced, and the operation efficiency and resource utilization of the entire work system are effectively improved. BRIEF DESCRIPTION OF DRAWINGS

[0011] The accompanying drawings, which are incorporated into and form a part of the specification, illustrate an embodiment consistent with the application and, together with the description, serve to explain the principles of the application.

[0012] Figure 1 A step flowchart of an AGV task allocation method based on multi-objective optimization provided by the embodiments of the application.

[0013] Figure 2 A structure block diagram of an AGV task allocation system provided by the embodiments of the application. DETAILED DESCRIPTION

[0014] The technical solutions in the present application will be described below with reference to the drawings. In the following description, identical numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all the embodiments consistent with the present application. Instead, they are merely examples of apparatuses and methods consistent with some aspects of the present application. It should be noted that the terms "first", "second", and the like in the description of the present application and the above drawings are used to distinguish similar objects, and do not necessarily indicate a specific order or sequence.

[0015] Referring to Figure 1 , Figure 1 is a flowchart of an AGV task allocation method based on multi-objective optimization provided by an embodiment of the present application. The method is applied to an AGV task allocation system and can further include steps 110-140.

[0016] Step 110: Obtain a task demand set and an AGV device state information set in a current work scenario. The task demand set includes multiple task units.

[0017] Taking a power battery factory as an example, the demand for raw materials on the production line of the power battery factory, the storage arrangement of finished products, etc. will give rise to different types of tasks. For example, when the raw materials for a certain link of the production line are about to run out, a task of transporting the raw materials from the warehouse to the position of the production line will be triggered. As a carrier for executing these tasks, the state information of the AGV device is crucial for reasonable task allocation. The position of each AGV device in the factory, the remaining work capacity, and the allocated task situation, etc. all directly affect the decision of task allocation. Therefore, accurately obtaining the task demand set and the AGV device state information set in the current work scenario is the primary prerequisite for subsequent efficient task allocation.

[0018] In one embodiment, the obtaining of the task demand set and the AGV device state information set in the current work scenario includes: Step 111: Listen to a task trigger signal in the work scenario; analyze task basic information in the task trigger signal, extract a task type identifier and a task execution position identifier of each task unit, the task type identifier is used to distinguish between a transportation task, a sorting task, or an inspection task, and the task execution position identifier is a coordinate point or a region number in the work scenario.

[0019] In a power battery factory, the sources of task trigger signals are diverse. Sensors on the production line monitor the inventory of raw materials in real time. When the inventory is below a certain level, a trigger signal for a handling task will be sent. Management personnel may also issue instructions for sorting tasks or inspection tasks through manual operation terminals. The central control server is used to listen to these task trigger signals. Once a signal is received, it will be analyzed in detail.

[0020] Taking a handling task trigger signal as an example, the task type identifier contained in the signal will be carefully identified during the analysis process, and it will be clear that this is a handling task. At the same time, the task execution location identifier is extracted, which can be a specific coordinate point in the factory, such as a shelf location in the raw material warehouse or a workstation coordinate on the production line. By accurately extracting this information, the key elements of each task unit can be clearly defined.

[0021] In detail, the task trigger signal in the work scene is listened to, including: Step 1111: Deploy a signal listening module in the central control server of the work scene, the signal listening module is configured with a communication protocol interface with external systems and manual operation terminals; a real-time data connection channel is established with the external system through the communication protocol interface; a listening queue for task trigger signals is set up in the data connection channel, the listening queue processes received signal data according to the first-in first-out principle; the signal data in the listening queue is processed by protocol analysis, extracting the task generation timestamp in the signal header information and the task basic information in the signal body information; the parsed task basic information is processed by format checking, verifying whether the task type identifier is of a preset type and whether the task execution location identifier is within the work scene coordinate range, and filtering out invalid signal data that does not meet the requirements.

[0022] In the central control server of the power battery factory, a signal listening module is deployed, which is equipped with communication protocol interfaces with external systems (such as sensor systems on the production line) and manual operation terminals. These communication protocol interfaces are configured according to different system characteristics, for example, the industrial Ethernet protocol may be used with the sensor system to ensure the stability and real-time performance of data transmission. Through these interfaces, a real-time data connection channel is established with the external system, so that the central control server can receive task trigger signals in a timely manner.

[0023] In addition, a listening queue is arranged in the data connection channel, which has the function of orderly management of multiple task trigger signals. When multiple signals are received, they are processed according to the first-in first-out principle. For example, the signal received first will be given priority to enter the processing flow, avoiding confusion in signal processing. When the signal data in the queue is protocol-analyzed, attention will be paid to the task generation timestamp in the signal header information, which records the specific time of task triggering. At the same time, the task basic information is extracted from the signal body information, including the task type identifier and the task execution position identifier.

[0024] After the analysis is completed, the task basic information needs to be format-verified. For the task type identifier, it will be checked whether it is a preset carrying, sorting or inspection task type. If the task type identifier received in the signal does not belong to these preset types, it means that the signal may have errors and will be filtered out. For the task execution position identifier, it will be verified whether it is within the coordinate range of the work scene. The work area of the factory has a clear coordinate boundary. If the task execution position identifier exceeds this range, the signal will also be determined as an invalid signal and filtered out. Through the above format verification processing, the accuracy and effectiveness of the received task trigger signal can be ensured.

[0025] Step 112: Traverse the AGV device communication interface in the work scene to obtain the real-time state data of each AGV device, which includes the device unique identifier, the current position coordinates, the remaining work capacity value and the allocated task sequence list.

[0026] In the power battery factory, each AGV device is equipped with an independent communication interface for data interaction with the central control server. The central control server will traverse these communication interfaces in turn and send a state query instruction to each AGV device. When the AGV device receives the query instruction, it will return its real-time state data to the central control server.

[0027] The device unique identifier is the unique identification of each AGV device, which can be a number or code, used to distinguish different AGV devices. The current position coordinates are obtained through the positioning system on the AGV device, which can accurately reflect the specific position of the AGV device in the factory. The remaining work capacity value represents the amount of work that the AGV device can currently undertake, for example, for carrying tasks, it can be the remaining carrying capacity. The allocated task sequence list records the tasks that the AGV device has been allocated, including the type, execution order and other information of the tasks. By obtaining these real-time state data, the central control server can fully understand the current state of each AGV device.

[0028] Step 113: validity verification processing is performed on the real-time state data, and AGV devices with communication abnormalities or missing state data are filtered out to obtain a set of state information of AGV devices.

[0029] After obtaining the real-time state data of AGV devices, validity verification needs to be performed. Communication abnormalities can cause incomplete or incorrect data transmission, for example, signal interference can cause the returned data of AGV devices to appear garbled. Missing state data can be due to the failure of AGV devices themselves, causing some key information to be unable to return normally.

[0030] The central control server checks the real-time state data of each AGV device. First, the integrity of the data is checked to ensure that the device unique identifier, current position coordinates, remaining work capacity value, and assigned task sequence list are present. If it is found that the state data of a certain AGV device is missing, for example, the current position coordinates are not returned, the device will be marked as a device with missing state data. Then the accuracy of the data is checked, and some obviously illogical data is filtered out. For example, if the remaining work capacity value is negative, it is obviously unreasonable and the device will be considered as a communication abnormal device. By filtering out these AGV devices with communication abnormalities or missing state data, the final set of state information of AGV devices only contains AGV devices with normal and valid state data.

[0031] Step 114: timestamp alignment processing is performed on the task type identifier and task execution location identifier of the task unit, and the device unique identifier, current position coordinates, remaining work capacity value, and assigned task sequence list of the AGV device, to generate a task demand set and a state information set with time synchronization relationship.

[0032] Since there can be differences in the generation time of the task trigger signal and the acquisition time of the AGV device state data, in order to ensure the validity of the task demand and the AGV device state information at the same time point, timestamp alignment processing needs to be performed.

[0033] The central control server records the task generation timestamp in the task trigger signal and the acquisition timestamp of the AGV device state data. When performing timestamp alignment, a certain unified time standard is used as a reference to match the task type identifier and task execution location identifier of the task unit with the relevant information of the AGV device. For example, the task demand generated in the same time period is associated with the AGV device state information acquired in the same time period. Through the above timestamp alignment processing, the generated task demand set and state information set have a time synchronization relationship, so that the subsequent task allocation can be based on accurate time information, improving the rationality and accuracy of task allocation.

[0034] Step 120: Feature extraction processing is performed on the task requirement set and the state information set to obtain task feature descriptions of each task unit and capability feature descriptions of each AGV device.

[0035] After obtaining the task requirement set and the state information set of the AGV devices, feature extraction processing needs to be performed on them in order to have a deeper understanding of the characteristics of the tasks and the AGV devices.

[0036] For a task unit, by analyzing its task type and execution location and other information, information that can describe the characteristics of the task is extracted. Different types of tasks have different characteristics. For example, the characteristics of a carrying task may include the required carrying capacity, carrying distance, carrying time requirement, etc.; the characteristics of a sorting task may include the number of sorted items, the complexity of sorting, etc.; the characteristics of a patrol task may include the area range of patrol, the time interval of patrol, etc. At the same time, the task execution location also affects the task characteristics. For example, a task with a long distance may require more time and resources to complete.

[0037] For an AGV device, by analyzing its location, remaining work capacity and other state information, information that can describe its capability characteristics is extracted. The current location of the device determines the distance and time it takes to reach the task execution location; the remaining work capacity reflects the amount of tasks it can currently undertake; the assigned task sequence list can reflect the degree of busyness of its work. In addition, other attributes of the AGV device, such as travel speed, maximum carrying capacity, etc., will also be extracted as capability features. Through feature extraction processing, the task feature descriptions of each task unit and the capability feature descriptions of each AGV device obtained can more comprehensively reflect the characteristics of the tasks and the AGV devices.

[0038] As an embodiment, the feature extraction processing on the task requirement set and the state information set to obtain the task feature descriptions of each task unit and the capability feature descriptions of each AGV device includes: Step 121: Type association analysis processing is performed on each task unit in the task requirement set to extract historical execution efficiency data, device adaptation preference data and path complexity data corresponding to the task type identifier, and generate task type association features.

[0039] In a power battery factory, type association analysis processing is performed on each task unit in the task requirement set. Different types of tasks will accumulate different efficiency data in the historical execution process. For example, the historical execution efficiency data of a carrying task may include the time required for each carrying, the energy consumption during carrying, etc. Through analysis of these historical execution efficiency data, the execution difficulty and efficiency level of the task type can be understood.

[0040] Device adaptation preference data reflects which AGV devices are more suitable for different types of tasks. For example, some transportation tasks may be more suitable for AGV devices with high carrying capacity and fast driving speed; while sorting tasks may be more suitable for AGV devices with flexible operation. By analyzing historical task allocation and execution, device adaptation preferences for each task type can be summarized.

[0041] Path complexity data is related to the execution path of the task. For transportation tasks, path complexity may include path length, number of turns, whether passing through narrow channels, etc.; for inspection tasks, it may include the complexity of the layout of the inspection area, etc. By analyzing these path complexity data, the difficulty of task execution and the required resources can be evaluated.

[0042] By extracting and analyzing these data, task type-related features are generated, which can more comprehensively describe the characteristics of the task and provide important references for subsequent task allocation.

[0043] Step 122: Perform spatial relationship analysis on the task execution location identifier of the task unit, calculate the distance relationship and path connectivity data between the task execution location and the key nodes in the work scene, and generate task location-related features.

[0044] In the work scene of a power battery factory, there are some key nodes, such as the entrance node, exit node, and storage rack node of the work scene. The spatial relationship analysis of the task execution location identifier of the task unit is to calculate the distance relationship and path connectivity data between the task execution location and these key nodes.

[0045] The distance relationship between the task execution location and the key nodes can be determined by measuring the straight-line distance or actual driving distance between the two points. For example, for a transportation task, the distance between its task execution location and the raw material warehouse node will affect the transportation time and cost. The farther the distance, the more time and resources required for transportation.

[0046] Path connectivity data reflects whether the path between the task execution location and the key nodes is unobstructed. In the factory, there may be some obstacles or narrow passages that will affect the AGV device's driving. By analyzing factors such as the number of obstacles in the path and the number of turns in the path, the connectivity of the path can be evaluated. For example, if there are multiple obstacles and frequent turns in the path, the connectivity of the path is poor, and the difficulty and time of AGV device driving will increase.

[0047] By calculating and analyzing the distance relationship and path connectivity data, task location-related features are generated, which can more accurately describe the characteristics of the task execution location.

[0048] In an example embodiment, the task execution location of the task unit is subjected to spatial relationship analysis processing, the distance relationship and path connectivity data between the task execution location and the key nodes in the job scene are calculated, and task location association features are generated, including: Step 1221: Obtain electronic map data of the job scene.

[0049] In a power battery factory, the electronic map data of the job scene records the layout information in the factory, including the positions of various regions, the distribution of obstacles, the direction of passages, etc. The central control server can obtain the electronic map data from the database, and the electronic map data provides accurate geographic information, making it possible to accurately calculate the distance and path connectivity between the task execution location and the key nodes.

[0050] Step 1222: Map the task execution location identifier into the electronic map data to determine the specific coordinate point of the task execution location in the electronic map.

[0051] After obtaining the electronic map data of the job scene, the task execution location identifier needs to be mapped into the electronic map. The task execution location identifier can be a region number or a coordinate range, and by matching with the electronic map data, the specific coordinate point in the electronic map can be determined. For example, if the task execution location identifier is the number of a warehouse area, the specific coordinates of the warehouse area in the map can be found by querying the electronic map data, so that the task execution location can be accurately located.

[0052] Step 1223: Calculate the Euclidean distance between the task execution location and the entry node, exit node, and storage rack node of the job scene, and generate distance relationship data between the task location and the key nodes.

[0053] After determining the specific coordinate point of the task execution location in the electronic map, the Euclidean distance between the task execution location and the entry node, exit node, and storage rack node of the job scene can be calculated. The Euclidean distance refers to the straight-line distance between two points, and by measuring the straight-line distance between the task execution location and these key nodes, the distance relationship data between the task location and the key nodes can be obtained.

[0054] For example, for a carrying task, the Euclidean distance between the task execution location and the raw material warehouse rack node is calculated, which can reflect the approximate distance of carrying and provide a reference for task allocation and path planning. The distance relationship data can help judge the difficulty of task execution and the required resources. The farther the distance, the more time and energy may be needed to complete the task.

[0055] Step 1224: analyze the path connectivity between the task execution location and each key node, count the number of obstacles and the number of turns in the path, and generate path connectivity data.

[0056] After calculating the distance relationship between the task execution location and the key nodes, it is necessary to analyze the path connectivity between them. Through analysis of electronic map data, the number of obstacles and the number of turns in the path between the task execution location and each key node are counted.

[0057] For example, in the path from the task execution location to the raw material warehouse node, if there are multiple obstacles such as machine equipment, goods piles, etc., the AGV device needs to bypass these obstacles during travel, increasing the difficulty and time of travel. The number of turns in the path also affects the efficiency of travel, and frequent turns can reduce the travel speed of the AGV device. By counting this information, path connectivity data is generated, which can more comprehensively evaluate the difficulty of task execution and the resources required.

[0058] Step 1225: normalize the distance relationship data and path connectivity data to eliminate the dimensional differences of different dimensions of data, and generate task location association features with a unified numerical range.

[0059] Since the distance relationship data and path connectivity data have different dimensions, for example, the unit of distance relationship data can be meters, while the path connectivity data can be the number of obstacles and the number of turns, etc. These different dimensions of data cannot be directly compared and fused. Therefore, they need to be normalized.

[0060] The purpose of normalization is to convert data of different dimensions to a unified numerical range, so that they are comparable. For example, both the distance relationship data and the path connectivity data can be converted to the range of 0 to 1. Through normalization, the dimensional differences of different dimensions of data are eliminated, and task location association features with a unified numerical range are generated. The above task location association features can be more conveniently used for subsequent task allocation and model calculation.

[0061] Step 123: perform location matching analysis processing on each AGV device in the state information set, calculate the spatial distance parameter and path accessibility data between the current location coordinates of the AGV device and the task execution location identifier, and generate location matching association features.

[0062] After obtaining the state information set of the AGV device, perform location matching analysis processing on each AGV device. First, calculate the spatial distance parameter between the current location coordinates of the AGV device and the task execution location identifier. This spatial distance can be a straight-line distance or an actual travel distance, which reflects the approximate distance of the AGV device to the task execution location.

[0063] For example, for a transport task, the spatial distance between the current location of an AGV device and the task execution location (raw material warehouse) is calculated. The farther the distance, the more time and resources the AGV device needs to reach the task execution location.

[0064] At the same time, the path accessibility data between the current location of the AGV device and the task execution location is analyzed. Path accessibility takes into account factors such as whether there are obstacles in the path, whether the channel is narrow, etc. If there are a large number of obstacles in the path or the channel is narrow, the AGV device may not be able to reach the task execution location smoothly, or it may need to spend more time and effort to bypass the obstacles. By calculating the spatial distance parameter and analyzing the path accessibility data, the location matching correlation features are generated, which can reflect the matching degree of the AGV device and the task execution location.

[0065] Step 124: Perform capacity adaptation analysis processing on the remaining job capacity value of the AGV device, extract the matching degree data of the remaining job capacity value and the job resource amount required by the task unit, and generate capacity adaptation correlation features in combination with the execution time estimation data of the allocated task sequence list.

[0066] For the remaining job capacity value of the AGV device, capacity adaptation analysis processing is needed. First, extract the matching degree data of the remaining job capacity value and the job resource amount required by the task unit. For example, for a transport task, the job resource amount required by the task unit can be the weight of the goods that need to be transported, and the remaining job capacity value of the AGV device is its remaining carrying capacity. By comparing the two values, the matching degree can be obtained. If the remaining job capacity value is greater than the job resource amount required by the task, it means that the AGV device can undertake the task; otherwise, it may not be able to undertake it.

[0067] At the same time, in combination with the execution time estimation data of the allocated task sequence list. The allocated task sequence list records the tasks that have been allocated to the AGV device. By estimating the execution time of these tasks, the busy degree of the AGV device in the future period of time can be understood. If the AGV device has been allocated a large number of tasks and the execution time is long, then in this case, even if the remaining job capacity value of the AGV device can theoretically undertake a new task, due to its busy job, it may not be able to complete the new task in time, resulting in a delay in task execution. In combination with these two aspects of data, capacity adaptation correlation features are generated, which comprehensively consider the remaining job capacity and the busy degree of the AGV device, and can more accurately reflect the adaptation ability of the AGV device to the task.

[0068] Step 125: performing feature fusion processing on the task type association feature and the task location association feature to generate a task feature description of each task unit; performing feature fusion processing on the location matching association feature and the capacity adaptation association feature to generate a capability feature description of each AGV device.

[0069] After obtaining the task type association feature and the task location association feature respectively, feature fusion processing is needed to generate a task feature description of each task unit. The process of feature fusion processing is to organically combine the two features and fully consider their respective influences on the task. For example, for a carrying task, the historical execution efficiency data in the task type association feature can reflect the speed of completing the task in the past, while the distance relationship data and the path connectivity data in the task location association feature reflect the spatial difficulty of task execution. Through weighted fusion, these different aspects of features are integrated to form a task feature description that can comprehensively describe the task unit.

[0070] Similarly, for an AGV device, the location matching association feature and the capacity adaptation association feature are fused to generate its capability feature description. The spatial distance parameter and the path accessibility data in the location matching association feature reflect the difficulty of the AGV device to reach the task execution location, and the capacity adaptation association feature comprehensively reflects the remaining work capacity and the work busy degree. The fusion of these two features enables the capability feature description to fully reflect the ability of the AGV device to execute tasks, including its moving ability and carrying ability, etc.

[0071] Step 130: performing joint optimization matching processing according to the task feature description and the capability feature description to generate a matching association relationship set of tasks and AGV devices, the matching association relationship set containing the corresponding relationship of task units and AGV devices and the task execution order relationship.

[0072] After obtaining the task feature description of each task unit and the capability feature description of each AGV device, joint optimization matching processing is needed. The core of this process is to find the best matching relationship between tasks and AGV devices, which not only considers the corresponding relationship of task units and AGV devices, i.e., which task should be assigned to which AGV device, but also determines the task execution order relationship, i.e., these tasks should be executed in what order on the AGV device. Through joint optimization matching processing, the capabilities of AGV devices can be fully utilized to improve the efficiency and quality of task execution.

[0073] Optionally, the joint optimization matching processing according to the task feature description and the capability feature description to generate a matching association relationship set of tasks and AGV devices comprises: Step 131: input the task feature description and the capability feature description into the input layer of the multi-objective optimization model for feature dimension alignment processing to unify the feature dimension quantity of the task feature description and the capability feature description.

[0074] In the embodiment of the application, the input layer of the multi-objective optimization model is used to realize feature dimension alignment. Since the task feature description and the capability feature description can have different feature dimension quantities, this will bring difficulties to subsequent model processing. Therefore, it is necessary to process them in the input layer to unify the feature dimension quantity of the two. For example, the task feature description can contain more feature dimensions related to the time requirement of the task, while the capability feature description has fewer dimensions in this respect. Through feature dimension alignment processing, the two feature descriptions can be analyzed and matched in the same dimension system in the subsequent analysis and matching, ensuring that the model can accurately process and compare them.

[0075] Step 132: calculate the correlation degree value between the task feature description of each task unit and the capability feature description of each AGV device through the feature correlation layer of the multi-objective optimization model, the correlation degree value being used to represent the adaptation degree of the task and the AGV device.

[0076] In the feature correlation layer of the multi-objective optimization model, the input task feature description and the capability feature description are further processed. The main function of this layer is to calculate the correlation degree value between the task feature description of each task unit and the capability feature description of each AGV device. The calculation of the correlation degree value is based on various features of the task and the AGV device, and comprehensively considers the requirements of the task and the capabilities of the AGV device. For example, for a carrying task that needs to be completed quickly and has high requirements on carrying capacity, if a certain AGV device has high driving speed and large carrying capacity, then the correlation degree value between them will be relatively high, indicating that the adaptation degree of the AGV device to the task is good. The correlation degree value can directly reflect the adaptation relationship between the task and the AGV device.

[0077] Step 133: use the constraint condition layer of the multi-objective optimization model to apply a preset task allocation constraint rule, the task allocation constraint rule including that the remaining work capacity of the AGV device is not less than the required resource quantity of the task, the task execution path does not conflict with the allocated task path, and the task execution time does not exceed the preset deadline.

[0078] Further, the constraint condition layer of the multi-objective optimization model applies preset task allocation constraint rules to ensure the rationality and feasibility of task allocation. First, the rule that the remaining job capacity of AGV equipment is not less than the required resource amount of the task ensures that AGV equipment has sufficient capacity to complete the task. For example, for a carrying task, if the carrying capacity required by the task exceeds the remaining job capacity of the AGV equipment, the AGV equipment should not be allocated the task. Second, the rule that the task execution path does not conflict with the allocated task path avoids collisions or congestion of AGV equipment during travel. Finally, the rule that the task execution time does not exceed the preset deadline ensures that the task can be completed on time to meet the time requirements of production. Through the application of these constraint rules, the matching relationship between tasks and AGV equipment that meet the conditions is screened out.

[0079] Step 134: Through the optimization decision layer of the multi-objective optimization model, based on the correlation value and the task allocation constraint rule, a multi-objective optimization algorithm is used to globally optimize the matching relationship between tasks and AGV equipment, generating a plurality of candidate matching association relationship sets.

[0080] In the optimization decision layer of the multi-objective optimization model, the correlation value and the task allocation constraint rule are combined, and a multi-objective optimization algorithm is used to globally optimize the matching relationship between tasks and AGV equipment. The purpose of the multi-objective optimization algorithm is to find the optimal task allocation scheme under the premise of meeting the constraint conditions. The multi-objective optimization algorithm will consider the correlation value and try to maximize the adaptability of tasks and AGV equipment. In the optimization process, the correspondence between tasks and AGV equipment and the task execution order are constantly adjusted to generate a plurality of different candidate matching association relationship sets. These candidate sets contain different task allocation schemes, each of which meets the requirements of task allocation to some extent, but may differ in some aspects.

[0081] As a preferred embodiment, the multi-objective optimization algorithm is used to globally optimize the matching relationship between tasks and AGV equipment based on the correlation value and the task allocation constraint rule, generating a plurality of candidate matching association relationship sets, including: Step 1341: Initialize the population individuals of the multi-objective optimization algorithm, each individual representing a matching scheme of tasks and AGV equipment, the matching scheme including the correspondence between task units and AGV equipment and the task execution order relationship; When starting the optimization process of the multi-objective optimization algorithm, the population individuals are first initialized. Each population individual represents a matching scheme of tasks and AGV devices, which explicitly shows the correspondence between task units and AGV devices and the task execution order relationship. For example, a population individual can represent that task A is assigned to AGV device 1, task B is assigned to AGV device 2, and task A is executed first and task B is executed later. By initializing multiple population individuals, an initial population can be constructed.

[0082] Step 1342: Calculate the objective function value of each individual, which is composed of the correlation value, the AGV device job load balancing degree, and the total task execution time; apply the task allocation constraint rule to check the feasibility of the individual, eliminate the individual that does not satisfy the AGV device remaining job capacity constraint, the path conflict constraint, and the time deadline constraint; perform cross operation and mutation operation on the feasible individual to generate a new generation of population individuals.

[0083] After initializing the population individuals, the objective function value of each individual is calculated. The objective function value is the result of considering multiple factors, and the embodiment of the application is composed of the correlation value, the AGV device job load balancing degree, and the total task execution time. The correlation value reflects the degree of adaptation of tasks to AGV devices, the AGV device job load balancing degree reflects whether the task allocation between AGV devices is uniform, and the total task execution time relates to the efficiency of the entire task allocation scheme. By calculating the objective function value, each individual can be evaluated to understand its degree of excellence.

[0084] Then, the task allocation constraint rule is applied to check the feasibility of the individual. For the individual that does not satisfy the AGV device remaining job capacity constraint, the path conflict constraint, and the time deadline constraint, it will be eliminated. For example, if the task allocation in a certain individual causes the remaining job capacity of the AGV device to be less than the required resource amount of the task, or the task execution path conflicts with the allocated task path, or the task execution time exceeds the preset deadline, the individual does not meet the requirements and will be eliminated.

[0085] For the feasible individuals that remain after the feasibility check, cross operation and mutation operation are performed on them. Cross operation is to exchange part of the information of two feasible individuals to generate a new individual. For example, exchange part of the correspondence between tasks and AGV devices in one individual with part of the correspondence in another individual. Mutation operation is to randomly change some information in the individual. Through these operations, a new generation of population individuals can be generated, providing more possibilities for the next round of optimization iteration.

[0086] Step 1343: Repeat the target function value calculation, feasibility check, crossover operation and mutation operation of step 1342 until a preset number of iterations is reached or the target function value converges, generating a plurality of candidate matching association relationship sets.

[0087] After completing a target function value calculation, feasibility check, crossover operation and mutation operation, these steps will be repeated. Each iteration optimizes the population individuals, causing the target function value to develop in a more optimal direction. This process continues until a preset number of iterations is reached or the target function value converges. The preset number of iterations is a pre-set value, and when the number of iterations reaches this value, the optimization process stops. Target function value convergence indicates that after multiple iterations, the change in the target function value is very small, close to the optimal convergence interval. When one of the two conditions is met, the optimization process ends, and a plurality of candidate matching association relationship sets are finally generated, which contain different task allocation schemes, each of which meets the requirements and constraints of task allocation to a certain extent.

[0088] Step 135: Call the evaluation and screening layer of the multi-objective optimization model to comprehensively evaluate the candidate matching association relationship sets, and select the candidate matching association relationship set with the highest comprehensive evaluation value as the final matching association relationship set.

[0089] After generating a plurality of candidate matching association relationship sets, the evaluation and screening layer of the multi-objective optimization model will comprehensively evaluate these sets. The evaluation and screening layer will score each candidate matching association relationship set according to certain evaluation criteria to obtain a comprehensive evaluation value. The comprehensive evaluation value will take into account factors such as association degree value, AGV device job load balancing degree, and total task execution time. For example, a candidate matching association relationship set with a high association degree value, good AGV device job load balancing degree, and short total task execution time will have a relatively high comprehensive evaluation value. By comparing the comprehensive evaluation values of each candidate set, the candidate matching association relationship set with the highest comprehensive evaluation value is selected as the final matching association relationship set, which represents the optimal task and AGV device matching scheme under the current conditions.

[0090] Step 140: Generate a task allocation scheme based on the matching association relationship set, which contains a list of tasks to be executed by each AGV device and execution timing arrangement information for each task.

[0091] After obtaining the final matching association relationship set, a task allocation scheme is generated based on the set. The task allocation scheme is to refine and organize the information in the matching association relationship set, and to clearly list the to-be-executed task list of each AGV device and the execution time sequence arrangement information of each task. The to-be-executed task list lists the tasks that each AGV device needs to execute, including the type of the task and the specific requirements of the task. The execution time sequence arrangement information determines the execution order and time arrangement of each task, ensuring that the tasks can be executed in a reasonable order and time, improving the efficiency of the entire work flow.

[0092] As an implementation manner, the task allocation scheme is generated based on the matching association relationship set, including: Step 141: Analyzing the correspondence between the task units and the AGV devices in the matching association relationship set, establishing a to-be-executed task list for each AGV device, and the to-be-executed task list contains the task type identifier and the task execution location identifier of the task unit corresponding to the AGV device.

[0093] When generating the task allocation scheme, the correspondence between the task units and the AGV devices in the matching association relationship set is first analyzed. Through the analysis of this correspondence, a to-be-executed task list is established for each AGV device. The to-be-executed task list records in detail the task type identifier and the task execution location identifier of the task unit corresponding to the AGV device. For example, for AGV device 1, its to-be-executed task list may show that task A is a carrying task and the task execution location is in the raw material warehouse; task B is a sorting task and the task execution location is in the sorting area. The above to-be-executed task list can clearly let the AGV device know which tasks it needs to execute and the specific information of the tasks.

[0094] Step 142: According to the task execution order relationship in the matching association relationship set, performing execution time sequence sorting processing on the to-be-executed task list of each AGV device, and generating a task sequence arranged in execution time sequence.

[0095] After establishing the to-be-executed task list of each AGV device, the to-be-executed task list is sorted according to the task execution order relationship in the matching association relationship set. The purpose of sorting is to generate a task sequence arranged in execution time sequence. For example, according to the task execution order relationship, it is determined that task A is executed first and task B is executed later, so in the task sequence of AGV device 1, task A will be arranged before task B. Through this sorting processing, the AGV device can execute the tasks in a reasonable order, avoid the confusion of task execution, and improve the efficiency of task execution.

[0096] Step 143: path planning processing is performed on the task sequence of each AGV device, in combination with the current position coordinates of the AGV device and the task execution position identifier, to generate optimal travel path information from the current position to each task execution position.

[0097] After obtaining the task sequence of each AGV device in the order of task execution time, path planning processing is performed thereon. The path planning processing needs to be combined with the current position coordinates of the AGV device and the task execution position identifier. By analyzing the electronic map data of the work scene, the optimal travel path information from the current position of the AGV device to each task execution position is found. The optimal travel path information takes into account factors such as path length, travel time, and degree of path unobstructedness. For example, when planning a path from the current position of the AGV device to the raw material warehouse to perform a carrying task, a route that avoids obstacles, has a shorter path length, and has less travel time is selected.

[0098] In a preferred embodiment, the path planning processing performed on the task sequence of each AGV device, in combination with the current position coordinates of the AGV device and the task execution position identifier, to generate optimal travel path information from the current position to each task execution position, includes: Step 1431: extract the current position coordinates of the AGV device and the coordinate points of the task execution position identifiers in the task sequence, to construct a set of starting points and ending points for path planning.

[0099] When performing path planning, the current position coordinates of the AGV device and the coordinate points of the task execution position identifiers in the task sequence are first extracted. These coordinate points constitute a set of starting points and ending points for path planning. For example, the AGV device is currently located at a certain position in the factory, and its coordinates are a certain value. The task sequence has the execution position coordinates of task A and the execution position coordinates of task B. Collecting these coordinate points forms a set containing a starting point (the current position of the AGV device) and multiple ending points (the execution positions of the tasks), providing a basis for subsequent path search.

[0100] Step 1432: apply a target search algorithm to search for an initial path from the starting point to the first task execution position in the electronic map data of the work scene, the initial path avoiding obstacle coordinates and selecting a route with the shortest path length.

[0101] After the start and end point set of path planning is constructed, a target search algorithm (A* algorithm) is applied to search in the electronic map data of the work scene. The target search algorithm starts from the start point (the current position of the AGV device) and searches for the initial path to the first task execution position. In the search process, obstacles coordinates are avoided to ensure the smooth travel of the AGV device. At the same time, the shortest path is selected to improve the travel efficiency. For example, on the electronic map, there may be multiple routes from the start point to the first task execution position, but the algorithm will filter out the route that avoids obstacles and has the shortest path length as the initial path.

[0102] Step 1433: Take the first task execution position as the new start point, search for the intermediate path to the second task execution position, considering the minimum turning radius and travel speed limit of the AGV device; repeat the path search operation until the complete path sequence from the current position to all task execution positions is generated.

[0103] After obtaining the initial path from the start point to the first task execution position, the first task execution position is taken as the new start point to continue searching for the intermediate path to the second task execution position. In searching for the intermediate path, the minimum turning radius and travel speed limit of the AGV device are considered. The minimum turning radius of the AGV device determines its flexibility when turning, and the travel speed limit relates to the safety and efficiency of travel. For example, if the minimum turning radius of the AGV device is large, it will avoid too narrow turns when selecting a path. By repeatedly performing the above path search operation, the new task execution position is taken as the start point to search for the path to the next task execution position, and finally a complete path sequence from the current position of the AGV device to all task execution positions is generated.

[0104] Step 1434: Smooth the complete path sequence to adjust the sharp turning points in the path to circular arc transition segments, generating a smooth path that meets the kinematic constraints of the AGV device.

[0105] After generating the complete path sequence, it will be smoothed. Since some sharp turning points may appear in the path search process, these sharp turning points are not conducive to the travel of the AGV device and may cause unstable travel or increase travel time. Therefore, the sharp turning points in the path will be adjusted to circular arc transition segments. Circular arc transition segments allow the AGV device to turn more smoothly and meet its kinematic constraints. Through smoothing, the generated smooth path can improve the travel efficiency and stability of the AGV device.

[0106] Step 1435: Calculate the total length and total travel time of the smooth path, and select the path with the shortest total length or the least total travel time as the optimal travel path information.

[0107] After obtaining the smooth path, the total length and the total travel time of the smooth path are calculated. The total length refers to the length of the entire smooth path, and the total travel time is the travel time calculated by considering the travel speed and path conditions of the AGV device. By comparing the total length and the total travel time of different smooth paths, the path with the shortest total length or the shortest total travel time is selected as the optimal travel path information. For example, if there are two smooth paths, one has a shorter total length but a longer travel time, and the other has a longer total length but a shorter travel time, the more suitable path is selected as the optimal travel path information according to the actual situation to ensure that the AGV device can efficiently complete the task.

[0108] Step 144: Extract the path length data and the estimated travel time data in the optimal travel path information, and generate the estimated start time identifier and the estimated end time identifier of each task unit by combining the time data required for task execution. After extracting the path length data and the estimated travel time data in the optimal travel path information, the time data required for the execution of each task unit can be further combined to generate the estimated start time identifier and the estimated end time identifier of each task unit. For each task unit, the determination of the estimated start time needs to consider the estimated travel time of the AGV device to reach the task execution position. If it is the first task, the estimated start time is usually the time after the estimated travel time required for the AGV device to travel from the current position to the task execution position along the optimal travel path. If it is a subsequent task, the estimated start time needs to consider the estimated end time of the previous task and the estimated travel time of the AGV device from the position where the previous task is completed to the current task execution position.

[0109] Taking an exemplary task sequence as an example, the AGV device is currently located at the PA point of the factory, the first task is to perform cargo handling at the PB point, and the second task is to perform cargo sorting at the PC point. It is known that the estimated travel time of the optimal travel path from the PA point to the PB point is a set time length, and the execution time required for the handling task B is also known. Then the estimated start time of task B is the time after the estimated travel time required for the AGV device to travel from the PA point to the PB point along the optimal path. The estimated start time of task C is the estimated end time of task B plus the estimated travel time of the AGV device from the PB point to the PC point.

[0110] In addition, the calculation of the estimated end time is based on the estimated start time plus the time required for task execution. Continuing with the above example, the estimated end time of task B is the estimated start time plus the execution time required for handling task B; the estimated end time of task C is the estimated start time plus the execution time required for sorting task C.

[0111] Through the above calculation logic, the predicted start time identifier and the predicted end time identifier can be accurately generated for each task unit by comprehensively considering the predicted travel time of the optimal travel path and the time required for task execution, which provides important time reference for subsequent task execution arrangement and scheduling, making the entire task allocation scheme more accurate and operable.

[0112] Step 145: Information integration processing is performed on the to-be-executed task list, the task sequence arranged in task execution time order, the optimal travel path information, the predicted start time identifier and the predicted end time identifier to generate a task allocation scheme containing task allocation details and execution time sequence arrangement.

[0113] After obtaining the to-be-executed task list, the task sequence arranged in task execution time order, the optimal travel path information, the predicted start time identifier and the predicted end time identifier respectively, these information needs to be integrated to generate a complete task allocation scheme. The process of information integration processing is to combine these different aspects of information to form a comprehensive and detailed task allocation scheme.

[0114] The to-be-executed task list specifies the tasks that each AGV device needs to execute, including basic information such as the type of the task and the task execution location. The task sequence arranged in task execution time order determines the execution order of these tasks, providing a basis for the ordered execution of the tasks. The optimal travel path information specifies the best travel route for the AGV device from the current location to each task execution location, ensuring that the AGV device can efficiently reach the task location. The predicted start time identifier and the predicted end time identifier provide a clear definition of the execution time for each task, making the time arrangement of the tasks more reasonable.

[0115] In integrating these information, first of all, the AGV device will be taken as a unit, and its corresponding to-be-executed task list, task sequence arranged in task execution time order, optimal travel path information and predicted start time identifier and predicted end time identifier of each task will be associated. For example, for a certain AGV device, the tasks to be executed will be arranged in task execution time order, with the predicted start time and predicted end time of each task marked, and the optimal travel path information from the current location to each task execution location attached.

[0116] Then, the relevant information of all AGV devices is summarized to form a complete task allocation scheme containing all task allocation details and execution time sequence arrangement. The scheme not only clearly shows the tasks that each AGV device needs to perform and the execution order of the tasks, but also provides detailed time arrangement and travel path information, so that the execution process of the tasks has high planability and controllability. Through the above information integration processing, the generated task allocation scheme can provide comprehensive and accurate guidance for the actual task execution, improving the efficiency and coordination of the entire operation process.

[0117] As an independent embodiment, after the task allocation scheme is generated based on the matching association relationship set, the following steps are further included: Step 210: Obtain real-time state feedback data in the task execution process. The real-time state feedback data includes the actual position coordinates of the AGV device, the task execution progress identifier, and the path deviation degree identifier.

[0118] After the task allocation scheme starts to execute, real-time state feedback data of the AGV device needs to be obtained in order to monitor and adjust the task execution. The AGV device will collect its actual position coordinates in real time through its own positioning system, sensors, and other devices. The actual position coordinates can accurately reflect the current position of the AGV device in the factory, and by comparing it with the position expected in the task allocation scheme, it can be determined whether the AGV device is traveling according to the planned path.

[0119] The task execution progress identifier is used to indicate the completion degree of the task execution. Different types of tasks have different progress measurement methods. For example, for a carrying task, the progress may be determined by whether the goods have been loaded, transported to the designated location, etc.; for a sorting task, the progress may be measured by the proportion of the number of sorted items to the total number of items. By obtaining the task execution progress identifier in real time, it can be known whether the task is proceeding according to the expected time and plan.

[0120] The path deviation degree identifier reflects the deviation between the actual travel path of the AGV device and the optimal travel path. In the actual travel process, due to various factors such as temporary obstacles in the factory, interference from other devices, etc., the AGV device may deviate from the pre-planned optimal travel path. By detecting the actual travel trajectory of the AGV device through sensors and other devices, and comparing it with the optimal travel path, the path deviation degree identifier can be obtained. This identifier can help determine whether the AGV device is traveling normally and whether it needs to be adjusted.

[0121] Step 220: Perform consistency check on the real-time state feedback data with the expected start time identifier, expected end time identifier, and optimal travel path information in the task allocation scheme, and generate an execution deviation feature set, which includes time deviation, position deviation, and path deviation.

[0122] After obtaining the real-time state feedback data, it needs to be consistent with the relevant information in the task allocation scheme. First, the time aspect of the check, the actual task execution progress and the expected start time identifier and the expected end time identifier are compared. If the actual start time of the task is later than the expected start time, or the actual end time is later than the expected end time, then there will be a time deviation. Time deviation reflects the delay of task execution in time, which may affect the progress of the entire work flow.

[0123] The position aspect of the check is to compare the actual position coordinates of the AGV device with the expected position in the task allocation scheme. If there is a difference between the actual position and the expected position, then there will be a position deviation. Position deviation can help determine whether the AGV device deviates from the predetermined task execution position, which may cause the task to be unable to proceed normally.

[0124] The path aspect of the check is to compare the actual travel path with the optimal travel path information, and determine the path deviation according to the path offset degree identifier. Path deviation reflects the deviation degree between the actual travel path of the AGV device and the optimal travel path. If the deviation is too large, it may increase the travel time and energy consumption, and even affect the completion quality of the task.

[0125] Through the check of time, position and path, the time deviation, position deviation and path deviation obtained are integrated to generate an execution deviation feature set, which comprehensively reflects the deviation between the task execution process and the task allocation scheme.

[0126] Step 230: Based on the execution deviation feature set, call the preset adjustment rule library to adaptively modify the task allocation scheme, and the adjustment rule library includes deviation threshold matching rules, resource reallocation rules, and path re-planning trigger rules.

[0127] After generating the execution deviation feature set, the preset adjustment rule library is called based on the set to adaptively modify the task allocation scheme. The adjustment rule library contains a variety of rules to deal with different types and degrees of deviation.

[0128] The deviation threshold matching rule is used to determine whether the deviation reaches a degree that requires adjustment. For the time deviation amount, the position deviation amount, and the path deviation amount, corresponding deviation thresholds are set. When the execution of a certain deviation amount in the deviation feature set exceeds the corresponding deviation threshold, the corresponding adjustment operation is triggered. For example, if the time deviation amount exceeds the preset time deviation threshold, it indicates that the task execution delay is serious, and corresponding adjustment is required.

[0129] The resource reallocation rule is to adjust the task allocation of the AGV device when a deviation occurs. If a certain AGV device cannot complete the task on time due to various reasons, or its task execution progress is seriously behind schedule, according to the resource reallocation rule, part or all of the tasks of the device can be reallocated to other capable AGV devices, which can ensure that the overall progress of the entire task allocation scheme is not greatly affected.

[0130] The path re-planning trigger rule is to trigger the re-planning of the AGV device's travel path when the path deviation amount exceeds a certain threshold. When the actual travel path of the AGV device deviates too much from the optimal travel path, it may lead to reduced travel efficiency or failure to reach the task execution location. At this time, according to the path re-planning trigger rule, a new travel path will be planned for the AGV device to ensure that it can successfully complete the task.

[0131] By invoking these rules, the task allocation scheme is modified according to the specific circumstances in the execution deviation feature set, making it better adapt to the actual task execution situation.

[0132] Step 240: Send the modified task allocation scheme to the control module of the corresponding AGV device to generate a task execution correction scheme containing adjustment instructions.

[0133] After the adaptive correction of the task allocation scheme, the modified task allocation scheme is sent to the control module of the corresponding AGV device. The control module of the AGV device is the core part of its task execution, responsible for receiving and executing various instructions. Sending the modified task allocation scheme to the control module actually conveys new task arrangements and adjustment instructions to the AGV device.

[0134] These adjustment instructions may include task reallocation, execution time adjustment, travel path change, etc. For example, if a task is transferred from one AGV device to another AGV device according to the resource reallocation rule, the new task allocation scheme will explicitly inform the AGV device receiving the task that it needs to execute the task, as well as the specific requirements and execution time of the task. If the travel path is re-planned according to the path re-planning trigger rule, the adjustment instruction will instruct the AGV device to travel according to the new path.

[0135] By sending the modified task allocation scheme to the control module of the AGV device, a task execution correction scheme containing adjustment instructions is generated, which can help the AGV device correct the deviation in the task execution process in time, ensure that the task can proceed smoothly according to the modified plan, and improve the stability and efficiency of the entire operation process.

[0136] As an independent embodiment, after generating the task allocation scheme based on the matching association relationship set, it further includes: Step 310: Extract the task sequence of each AGV device in the task allocation scheme in the order of task execution time and the corresponding optimal travel path information, and construct a space-time trajectory set of the multi-AGV collaborative execution scene, which contains the position coordinate mapping relationship of each AGV device at different time points.

[0137] After the task allocation scheme is generated, in order to realize the collaborative execution of multiple AGV devices, it is necessary to construct a space-time trajectory set of the multi-AGV collaborative execution scene. First, extract the task sequence of each AGV device in the task allocation scheme in the order of task execution time and the corresponding optimal travel path information. The task sequence in the order of task execution time clearly shows the tasks that each AGV device needs to execute at different times, while the optimal travel path information specifies the travel route of the AGV device from the current position to each task execution position.

[0138] According to these information, combined with the expected start time identifier and the expected end time identifier of each task, the position coordinates of each AGV device at different time points can be calculated. For example, for a certain AGV device, before the task starts, its position is the current position coordinate; during the process of traveling along the optimal travel path to the first task execution position, the position coordinates at different times can be calculated according to the travel time and speed; during the execution of the task, the position coordinate is the task execution position; after completing the task, the position coordinates at different time points are further calculated according to the travel path and time to the next task execution position.

[0139] Integrate all the position coordinates of the AGV devices at different time points to construct a space-time trajectory set, which contains the position coordinate mapping relationship of each AGV device at different time points, and can comprehensively show the space-time distribution of multiple AGV devices in the entire task execution process. By analyzing the space-time trajectory set, possible conflicts and problems can be found in advance.

[0140] Step 320: Cross-overlapping detection processing is performed on the spatio-temporal trajectory set to identify AGV device pairs with path space overlap or time interval overlap, and a collaborative conflict feature set is generated, which includes conflict position identifier, conflict time interval identifier, and involved AGV device identifier.

[0141] After the spatio-temporal trajectory set is constructed, cross-overlapping detection processing needs to be performed. Cross-overlapping detection is to identify AGV device pairs that may have conflicts during task execution. The conflict may manifest as path space overlap, i.e., two or more AGV devices appearing at the same location or nearby at the same time, which may cause collision or congestion; or time interval overlap, i.e., different AGV devices needing to use the same resource or execute the same regional task at the same time, which may cause resource competition and task interference.

[0142] By analyzing the position coordinates of each AGV device in the spatio-temporal trajectory set at different time points, and comparing the travel paths and time arrangements of different AGV devices, situations with path space overlap or time interval overlap are found. When it is found that the position coordinates of two or more AGV devices overlap at a certain time point or time period, it is determined that there is a conflict.

[0143] For each conflict situation, the conflict position identifier, i.e., the specific position coordinates where the conflict occurs, the conflict time interval identifier, i.e., the specific time range when the conflict occurs, and the involved AGV device identifier, which specifies which AGV devices have conflicts, are recorded. These information is integrated to generate a collaborative conflict feature set, which can clearly identify the possible conflict situations of multiple AGV devices during collaborative execution.

[0144] Step 330: Based on the collaborative conflict feature set, a multi-AGV priority sorting strategy is called to prioritize the conflicting devices, and a collaborative adjustment strategy is generated by combining the task urgency level identifier and the device remaining job capacity value, which includes task timing rearrangement instructions or path detour instructions.

[0145] After obtaining the collaborative conflict feature set, a multi-AGV priority sorting strategy is called to prioritize the conflicting devices based on the set. The multi-AGV priority sorting strategy determines the priority of the conflicting AGV devices according to certain rules and standards, which takes into account the task urgency level identifier and the device remaining job capacity value.

[0146] The task urgency identifier reflects the importance and time urgency of each task. For example, some tasks can be for the urgent needs of the production line, and if they cannot be completed on time, they may affect the progress of the entire production process, so the urgency of these tasks is high. The device remaining job capacity value reflects the current carrying capacity and executable task capacity of the AGV device. If the remaining job capacity of an AGV device is large, it means that it has more capacity to adjust the task arrangement.

[0147] According to the task urgency identifier and the device remaining job capacity value, the priority of the conflicting devices is divided. The AGV device with high priority has the right to execute the task first in the conflict processing. Based on the priority division result, combined with the information in the collaborative conflict feature set, a collaborative adjustment strategy is generated.

[0148] The collaborative adjustment strategy can include task timing rearrangement instructions or path bypass instructions. If the conflict is caused by overlapping time intervals, the task execution order of the conflicting device can be adjusted through the task timing rearrangement instructions, and the tasks of the device with low priority can be postponed or advanced to avoid time conflicts. If the conflict is caused by path space overlap, the device with low priority can be instructed to change the travel path to avoid the conflict position with the device with high priority through the path bypass instructions.

[0149] Through the above-mentioned collaborative adjustment strategy, the conflict problem that occurs in the collaborative execution process of multiple AGV devices can be effectively solved, and the collaborative work efficiency of multiple AGV devices can be improved.

[0150] Step 340: Embed the collaborative adjustment strategy into the original task allocation scheme to generate a revised task allocation scheme containing multiple device collaboration constraints.

[0151] After generating the collaborative adjustment strategy, it is embedded into the original task allocation scheme, which means that the task timing rearrangement instructions or path bypass instructions and other collaborative adjustment strategies are integrated with the contents such as the task list to be executed, the task sequence arranged in the order of task execution time, and the optimal travel path information in the original task allocation scheme.

[0152] If the collaborative adjustment strategy contains task timing rearrangement instructions, the task execution time order of the conflicting device in the original task allocation scheme needs to be adjusted. For example, the task of a certain AGV device with low priority is postponed, and the expected start time identifier and the expected end time identifier of the task are modified accordingly, and the time arrangement of other related tasks is updated to ensure the rationality of the entire task sequence.

[0153] If the cooperative adjustment strategy contains path bypass instructions, the optimal travel path information of the conflict device in the original task allocation scheme needs to be modified. Replace the original optimal travel path information with the new bypass path information to ensure that the AGV device can travel according to the new path and avoid path space overlap conflicts.

[0154] By embedding the cooperative adjustment strategy into the original task allocation scheme, a revised task allocation scheme containing multi-device cooperative constraints is generated. The revised task allocation scheme fully considers the conflict problems of multi-AGV devices during cooperative execution, and through reasonable adjustment and constraints, multi-AGV devices can complete tasks more coordinately, improving the efficiency and stability of the entire operation process.

[0155] As an independent embodiment, after generating the task allocation scheme based on the matching association relationship set, it further includes: Step 410: Collect historical execution data after the task allocation scheme is executed. The historical execution data includes the actual completion time identifier of the task unit, the actual job load value of the AGV device, and the actual traffic efficiency value of the task execution path.

[0156] After the task allocation scheme is executed, relevant historical execution data needs to be collected. These data can reflect the effect of the task allocation scheme in the actual execution process, providing an important basis for subsequent optimization. First, the actual completion time identifier of the task unit. By recording the actual completion time of each task unit, it can be compared with the estimated end time identifier in the task allocation scheme to understand the deviation of task execution in time.

[0157] The actual job load value of the AGV device refers to the actual amount of work undertaken by the AGV device during task execution. For transportation tasks, the actual job load value can be the weight of the transported goods; for sorting tasks, it can be the number of sorted items. The actual job load value can reflect the actual use of the AGV device, and by comparing it with the estimated job load in the task allocation scheme, the rationality of task allocation can be evaluated.

[0158] The actual traffic efficiency value of the task execution path is a measure of the efficiency of the task execution path in actual use. It can be evaluated by factors such as actual travel time and number of stops during travel. If the actual traffic efficiency value is low, it may indicate that there is a problem with the path planning, which needs to be adjusted.

[0159] By collecting these historical execution data, the actual execution of the task allocation scheme can be comprehensively understood.

[0160] Step 420: Matching effect evaluation processing is performed on the historical execution data and the expected start time identification in the task allocation scheme, the AGV device job load balancing degree target value, and the optimal travel path information, a feature matching evaluation report is generated, and the evaluation report includes the adaptability score of the task features and the capability features, the achievement degree score of each dimension of the target function, and the satisfaction degree score of the constraint rules.

[0161] After collecting the historical execution data, matching effect evaluation processing is performed on the historical execution data and the expected start time identification in the task allocation scheme, the AGV device job load balancing degree target value, and the optimal travel path information. First, for the adaptability score of the task features and the capability features, the degree of fit between the characteristics of the task units and the capability characteristics of the AGV device in actual execution is comprehensively considered.

[0162] Taking a carrying task as an example, the task features may include the required carrying capacity, carrying distance, carrying time requirement, etc., and the capability features of the AGV device include the maximum carrying capacity, travel speed, etc. The actual situation in the historical execution data is compared with the features relied on in the task allocation. If in actual execution, the carrying capacity of the AGV device can well meet the carrying demand of the task, and the travel speed matches the carrying time requirement, then the adaptability of the task features and the capability features is high, and the corresponding adaptability score is also high. Conversely, if the AGV device frequently appears insufficient carrying capacity or slow travel speed during task execution, resulting in task delay, etc., the adaptability score will be lower.

[0163] For the achievement degree score of each dimension of the target function, the target function is usually composed of the correlation value, the AGV device job load balancing degree, and the total task execution time. For the correlation value, the actual adaptation of the task and the AGV device in the historical execution is compared with the correlation value calculated in the task allocation. If the actual adaptation is consistent with the correlation value, it means that the achievement degree of the correlation value dimension is high. For the AGV device job load balancing degree, the actual job load of each AGV device in the historical execution data is compared with the job load balancing degree target value in the task allocation scheme. If the actual job load of each AGV device is relatively balanced and close to the target value, then the achievement degree of the job load balancing degree dimension is high; otherwise, it is low. For the total task execution time, the actual total time in the historical execution is compared with the expected total time in the task allocation scheme. The closer the actual total time is to the expected total time, the higher the achievement degree of the total task execution time dimension.

[0164] The satisfaction score of the constraint rule is based on the constraint rule set when the task is assigned, such as the remaining work capacity of the AGV device is not less than the required resource amount of the task, the task execution path does not conflict with the allocated task path, the task execution time does not exceed the preset deadline, etc. The historical execution data is evaluated. If all tasks can meet these constraint rules in actual execution, the satisfaction score of the constraint rule is full score; if some tasks violate the constraint rules, such as the remaining work capacity of a certain AGV device is less than the required resource amount of the task, or the task execution path conflicts, or the task execution time exceeds the preset deadline, the satisfaction score will be reduced accordingly according to the rule violation situation.

[0165] By comprehensively evaluating the adaptability of task characteristics and capability characteristics, the achievement degree of each dimension of the objective function, and the satisfaction degree of the constraint rule, a feature matching evaluation report is generated, which can comprehensively reflect the effect of the task allocation scheme in actual execution.

[0166] Step 430: Based on the feature matching evaluation report, iteratively update the task type associated features in the task feature description, the task location associated features, and the location matching associated features and capacity adaptation associated features in the capability feature description, and generate an optimized task feature description set and a capability feature description set.

[0167] According to the scores in the feature matching evaluation report, iteratively update the task feature description and the capability feature description. For the task type associated features in the task feature description, if the evaluation report shows that the historical execution efficiency data of a certain type of task in actual execution is different from the data previously relied on, such as the actual execution time of a certain type of carrying task is longer than previously estimated, the historical execution efficiency data corresponding to the task type needs to be updated. At the same time, if it is found in actual execution that the adaptation preference of some AGV devices for this type of task is different from the previous record, the device adaptation preference data also needs to be updated. In addition, if the actual path complexity is different from the previous evaluation, such as a new obstacle appears to increase the path complexity, the path complexity data also needs to be updated accordingly.

[0168] For the task location associated features, the distance relationship data and the path connectivity data are updated according to the actual distance relationship between the task execution location and the key nodes in the historical execution data and the path connectivity situation. If the distance between the task execution location and the entry node of the work scene measured in actual execution is different from the previous record, or the number of obstacles or the number of turns in the actual execution path is different from the previous statistics, these data need to be corrected. After the above updating process, the task location associated features can more accurately reflect the actual situation of the task execution location.

[0169] In terms of capability feature description, for the location matching associated feature, the spatial distance parameter and path accessibility data of the AGV device to the task execution location are updated according to the actual driving situation of the AGV device in historical execution. If the actual distance or path accessibility of the AGV device to some task execution locations changes due to the adjustment of the factory layout or the appearance of new obstacles in actual execution, these data need to be updated. For the capacity adaptation associated feature, the matching degree data of the remaining job capacity value and the required job resource amount of the task unit and the execution time estimation data of the allocated task sequence list are updated in combination with the actual job load situation of the AGV device in historical execution and the matching situation of the required resource amount of the task and the actual execution time of the allocated task sequence. For example, if it is found in actual execution that the required resource amount of some tasks is larger than previously expected, or the job efficiency of the AGV device in executing tasks is different from the expectation, causing the matching situation of the remaining job capacity value and the required resource amount of the task to change, the matching degree data need to be adjusted.

[0170] By iteratively updating the task type associated feature, the task location associated feature in the task feature description, and the location matching associated feature and the capacity adaptation associated feature in the capability feature description, the optimized task feature description set and the capability feature description set are generated, which can more accurately reflect the actual situation of the tasks and the AGV device.

[0171] Step 440: input the optimized feature description set into the multi-objective optimization model for parameter calibration to generate an iterative optimization model with historical experience learning ability.

[0172] After obtaining the optimized task feature description set and the capability feature description set, they are input into the multi-objective optimization model for parameter calibration. The multi-objective optimization model works based on the original task feature description and capability feature description in previous task allocation. As the tasks are continuously executed and historical execution data are accumulated, the original parameters may no longer be suitable for new situations.

[0173] After inputting the optimized feature description set into the model, the model will adjust its parameters according to these new feature data. For example, in the feature association layer, the model will re-learn the association between the task feature and the capability feature, adjust the parameters for calculating the association degree value, so that the association degree value can more accurately reflect the adaptation degree of the task and the AGV device. In the optimization decision layer, the model will adjust the parameters of the multi-objective optimization algorithm, such as the initialization method of the population individual, the probability of crossover operation and mutation operation, etc., to better achieve the global optimization matching of the task and the AGV device.

[0174] Through the above parameter calibration process, the multi-objective optimization model can adapt to new tasks and AGV device characteristics, generating an iterative optimization model with historical experience learning ability. This iterative optimization model can use the experience in historical execution data to make more reasonable and efficient decisions in subsequent task allocation. For example, in the next task allocation, the model can more accurately match the task characteristics and the ability characteristics of the AGV device, improving the adaptability of the task and the AGV device, while better meeting the constraint rules, achieving better optimization of each dimension of the objective function. The iterative optimization model with historical experience learning ability can continuously improve the quality and efficiency of task allocation, making the entire power battery factory operation process smoother and more efficient.

[0175] In the embodiments of the present application, the cosine similarity calculation correlation degree value can be introduced in the feature correlation layer, the path conflict judgment rule (such as the coordinate overlap threshold) is stored in the preset database in the constraint condition layer, and the population individual is screened through the tournament selection mechanism in the optimization decision layer, and the adaptive probability parameter is set for the crossover and mutation operation.

[0176] In actual application, the unified job capacity value dimension can be "standard carrying unit" (such as single battery module weight), the path length unit is unified as meter, and time unit (second) is added for description; in path planning, the heuristic function definition of A* algorithm and the curvature radius calculation method of circular arc transition section can be combined to realize.

[0177] In addition, in the multi-device collaborative conflict detection process, the four-way tree space index can be used to accelerate the trajectory comparison; the historical learning mechanism can introduce the sliding window model to update the task feature weight. In the capacity adaptation analysis, the task resource quantity quantification rule (such as carrying task resource quantity = cargo weight / standard carrying unit) can be used to realize further optimization.

[0178] To sum up, the embodiment of the present application realizes accurate grasping and efficient matching of task demand and AGV device state in a work scene, and significantly improves the scientificity and rationality of task allocation. By obtaining a task demand set and a state information set of AGV devices under the current work scene, the work demand and the device state can be comprehensively understood; the task demand set and the state information set are subjected to feature extraction processing to obtain task feature description of each task unit and capability feature description of each AGV device, realizing deep mining of task and device characteristics, so that the subsequent matching is more targeted; the task feature description and the capability feature description are subjected to joint optimization and matching processing to generate a matching association relationship set of tasks and AGV devices, comprehensively considering the corresponding relationship of the task unit and the AGV device and the task execution order relationship, avoiding blindness and randomness of task allocation, and improving the efficiency and coordination of task execution; a task allocation scheme is generated based on the matching association relationship set, clearly defining a to-be-executed task list of each AGV device and execution timing arrangement information of each task, so that the task allocation is clearer and more orderly, reducing device idling and task conflict, and effectively improving the operation efficiency and resource utilization rate of the entire work system. In this way, the embodiment of the present application can improve the problems of unreasonable task allocation, low device utilization rate, and low work efficiency, thereby adapting to complex and variable work scene demands.

[0179] Further, Figure 2 A structural block diagram of the AGV task allocation system 300 is shown, which includes a memory 310 for storing program instructions and data, and a processor 320 coupled with the memory 310 for executing instructions in the memory 310 to realize the above method.

[0180] Further, a computer storage medium is also provided, which contains instructions that, when executed on a processor, realize the above method.

[0181] The above only describes the preferred embodiments of the present application and is not used to limit the present application. For those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. An AGV task allocation method based on multi-objective optimization, characterized in that: The method comprises: Obtaining a task requirement set and an AGV device status information set in the current operation scenario, wherein the task requirement set includes multiple task units; Performing feature extraction processing on the task requirement set and the state information set to obtain a task feature description of each task unit and a capability feature description of each AGV device; Performing joint optimization matching processing based on the task feature description and the capability feature description to generate a matching association relationship set between the task and the AGV device, wherein the matching association relationship set includes a correspondence between the task unit and the AGV device and a task execution order relationship; A task allocation plan is generated based on the matching association relationship set, and the task allocation plan includes a list of tasks to be executed for each AGV device and execution timing arrangement information of each task.

2. The AGV task allocation method based on multi-objective optimization according to claim 1 is characterized in that: The joint optimization matching process is performed based on the task feature description and the capability feature description to generate a matching association relationship set between the task and the AGV device, including: Inputting the task feature description and the capability feature description into the input layer of the multi-objective optimization model to perform feature dimension alignment processing to unify the number of feature dimensions of the task feature description and the capability feature description; Calculate the correlation value between the task feature description of each task unit and the capability feature description of each AGV device through the feature association layer of the multi-objective optimization model, and the correlation value is used to represent the degree of adaptation between the task and the AGV device; Using the constraint layer of the multi-objective optimization model, the preset task allocation constraint rules are applied, wherein the task allocation constraint rules include that the remaining operating capacity of the AGV equipment is not less than the amount of resources required for the task, the task execution path does not conflict with the assigned task path, and the task execution time does not exceed the preset deadline; Through the optimization decision layer of the multi-objective optimization model, based on the association value and the task allocation constraint rule, a multi-objective optimization algorithm is used to perform global optimization processing on the matching relationship between the task and the AGV equipment, and multiple candidate matching association relationship sets are generated; The evaluation and screening layer of the multi-objective optimization model is called to perform comprehensive evaluation processing on the candidate matching association relationship set, and the candidate matching association relationship set with the highest comprehensive evaluation value is selected as the final matching association relationship set.

3. The AGV task allocation method based on multi-objective optimization according to claim 2 is characterized in that: The optimization decision layer of the multi-objective optimization model uses a multi-objective optimization algorithm to perform global optimization processing on the matching relationship between tasks and AGV equipment based on the association value and the task allocation constraint rule, and generates multiple candidate matching association relationship sets, including: Initialize the population of individuals in the multi-objective optimization algorithm, where each individual represents a matching scheme between a task and an AGV device. The matching scheme includes the correspondence between the task unit and the AGV device and the task execution order. Calculate the objective function value of each individual, which is composed of the correlation value, the AGV equipment operation load balance and the total task execution time; Apply task allocation constraint rules to verify the feasibility of individuals and eliminate individuals that do not meet the remaining operation capacity constraints, path conflict constraints, and time deadline constraints of AGV equipment; Perform crossover and mutation operations on feasible individuals to generate a new generation of population individuals; The objective function value calculation, feasibility check, crossover operation and mutation operation are repeatedly performed until the preset number of iterations is reached or the objective function value converges, and multiple candidate matching association relationship sets are generated.

4. The AGV task allocation method based on multi-objective optimization according to claim 1 is characterized in that: Generating a task allocation plan based on the matching association relationship set includes: Analyze the correspondence between the task units and the AGV devices in the matching association relationship set, and establish a to-be-executed task list for each AGV device, wherein the to-be-executed task list includes a task type identifier and a task execution location identifier of the task unit corresponding to the AGV device; According to the task execution order relationship in the matching association relationship set, the list of tasks to be executed by each AGV device is sorted in execution time sequence to generate a task sequence arranged in the order of task execution time; Perform path planning on the task sequence of each AGV device, combine the current position coordinates of the AGV device and the task execution location identifier, and generate the optimal driving path information from the current position to each task execution location; Extracting the path length data and the estimated travel time data from the optimal travel path information, and combining them with the time data required for task execution, to generate an estimated start time identifier and an estimated end time identifier for each task unit; The list of tasks to be executed, the task sequence arranged in order of task execution time, the optimal driving path information, the estimated start time identifier and the estimated end time identifier are integrated and processed to generate a task allocation plan including task allocation details and execution timing arrangement.

5. The AGV task allocation method based on multi-objective optimization according to claim 4 is characterized in that: The path planning process is performed on the task sequence of each AGV device, and the optimal driving path information from the current position to each task execution position is generated by combining the current position coordinates of the AGV device and the task execution position identifier, including: Extract the current position coordinates of the AGV device and the coordinate points of each task execution position in the task sequence, and construct the starting and ending points of the path planning; Applying a target search algorithm to search for an initial path from a starting point to a first task execution location in electronic map data of the work scene, wherein the initial path avoids obstacle coordinates and selects a route with the shortest path length; Taking the first task execution location as the new starting point, search for an intermediate path to the second task execution location, the intermediate path taking into account the minimum turning radius and driving speed limit of the AGV equipment; Repeat the path search operation until a complete path sequence from the current position to all task execution positions is generated; Smoothing the complete path sequence, adjusting sharp turning points in the path to arc transition segments, and generating a smooth path that complies with the kinematic constraints of the AGV device; The total length and total travel time of the smooth path are calculated, and the path with the shortest total length or the shortest total travel time is selected as the optimal travel path information.

6. The AGV task allocation method based on multi-objective optimization according to claim 1 is characterized in that: The step of obtaining the task requirement set and the AGV device status information set in the current operation scenario includes: Listen for task trigger signals in the operation scene; parse the basic task information in the task trigger signal, extract the task type identifier and task execution location identifier of each task unit, the task type identifier is used to distinguish between handling tasks, sorting tasks or inspection tasks, and the task execution location identifier is a coordinate point or area number in the operation scene; Traverse the AGV device communication interfaces in the operation scene and obtain the real-time status data of each AGV device, wherein the real-time status data includes the device unique identifier, current location coordinates, remaining operation capacity value and assigned task sequence list; Performing validity verification on the real-time status data, filtering out AGV devices with abnormal communication or missing status data, and obtaining a set of status information of the AGV devices; The task type identifier and task execution location identifier of the task unit are timestamp aligned with the device unique identifier, current location coordinates, remaining job capacity value, and assigned task sequence list of the AGV device to generate a task requirement set and status information set with a time synchronization relationship.

7. The AGV task allocation method based on multi-objective optimization according to claim 6 is characterized in that: The monitoring of the task trigger signal in the operation scenario includes: A signal monitoring module is deployed in the central control server of the operation scene, and the signal monitoring module is configured with a communication protocol interface with the external system and the manual operation terminal; Establishing a real-time data connection channel with an external system through the communication protocol interface; Setting a monitoring queue for task trigger signals in the data connection channel, wherein the monitoring queue processes received signal data according to a first-in-first-out principle; Performing protocol parsing on the signal data in the monitoring queue to extract the task generation timestamp in the signal header information and the task basic information in the signal body information; The parsed task basic information is format checked to verify whether the task type identifier is the preset type and whether the task execution location identifier is within the coordinate range of the operation scene, and invalid signal data whose format does not meet the requirements is filtered out.

8. The AGV task allocation method based on multi-objective optimization according to claim 1 is characterized in that: The feature extraction process is performed on the task requirement set and the state information set to obtain a task feature description of each task unit and a capability feature description of each AGV device, including: Performing type association analysis on each task unit in the task requirement set, extracting historical execution efficiency data, device adaptation preference data, and path complexity data corresponding to the task type identifier, and generating task type association features; Performing spatial relationship analysis on the task execution location identifier of the task unit, calculating the distance relationship and path connectivity data between the task execution location and key nodes in the operation scene, and generating task location association features; Performing position matching analysis on each AGV device in the state information set, calculating the spatial distance parameters and path reachability data between the current position coordinates of the AGV device and the task execution position identifier, and generating position matching association features; Performing capacity adaptation analysis on the remaining operating capacity of the AGV device, extracting matching data between the remaining operating capacity and the amount of operating resources required by the task unit, and combining this with the estimated execution time of the assigned task sequence list to generate capacity adaptation correlation features; The task type association features and the task location association features are subjected to feature fusion processing to generate a task feature description of each task unit; the location matching association features and the capacity adaptation association features are subjected to feature fusion processing to generate a capability feature description of each AGV device.

9. The AGV task allocation method based on multi-objective optimization according to claim 8 is characterized in that: The spatial relationship analysis processing of the task execution location identifier of the task unit is performed, the distance relationship and path connectivity data between the task execution location and the key nodes in the operation scene are calculated, and the task location association feature is generated, including: Obtain electronic map data of the operation scene; Mapping the task execution location identifier to the electronic map data to determine the specific coordinate point of the task execution location in the electronic map; Calculate the Euclidean distance between the task execution location and the entry node, exit node, and storage shelf node of the operation scene, and generate distance relationship data between the task location and key nodes; Analyze the path connectivity between the task execution location and each key node, count the number of obstacles and path turns in the path, and generate path connectivity data; The distance relationship data and path connectivity data are normalized to eliminate dimensional differences in data of different dimensions and generate task location association features with a unified numerical range.

10. An AGV task allocation system, characterized in that: include: Memory, used to store program instructions and data; A processor, coupled to a memory, and configured to execute instructions in the memory to implement the method according to any one of claims 1 to 9.

Images