Distributed task allocation and cooperative control method oriented to heterogeneous multi-agent system

Abstract

The invention discloses a distributed task allocation and cooperative control method oriented to a heterogeneous multi-agent system, and particularly relates to the technical field of artificial intelligence, state information of the multi-agent system is collected, modeling and classification are performed on the multi-agent system in an industrial field, and the task allocation and cooperative control method is established according to the working capability and task requirements of each agent. The method comprises the following steps: preliminarily matching tasks, designing a task allocation mechanism by adopting a reinforcement learning algorithm, performing task decision by utilizing local information among intelligent agents, dynamically adjusting a task allocation mode in a task allocation process, and for each intelligent agent, analyzing the state and task execution condition of the intelligent agent; the method comprises the steps of receiving feedback information, dynamically adjusting task allocation and cooperative control according to the feedback information, regularly evaluating task execution effects, resource consumption and agent load key indexes in a system operation process, and continuously optimizing a task allocation strategy and a cooperative control mechanism according to an evaluation result.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and more specifically, to a method for distributed task allocation and cooperative control of heterogeneous multi-agent systems. Background Technology

[0002] In modern industrial automation systems, with the rapid development of intelligent manufacturing, the Internet of Things, and robotics, field control systems have gradually shifted from traditional centralized control to a more flexible and intelligent distributed control architecture. Multi-agent systems, as a key technology, can improve the performance and efficiency of the entire system through the collaboration and division of labor among multiple agents.

[0003] Currently, existing distributed task allocation methods are based on preset rules or static optimization algorithms, ignoring the dynamic changes of tasks and the differences between agents. Although traditional centralized control systems can handle all tasks through centralized decision-making, they are inefficient in the face of large-scale, dynamic environments and cannot solve the dynamic adjustment problem in multi-agent collaborative work. In this case, distributed control systems and multi-agent collaborative work have become the key to improving the overall performance of the system. Summary of the Invention

[0004] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a distributed task allocation and cooperative control method for heterogeneous multi-agent systems, in order to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] Distributed task allocation and cooperative control methods for heterogeneous multi-agent systems include:

[0007] Collect state information of multi-agent systems, including working capacity, task requirements and environmental change data, and model and classify multi-agent systems in industrial sites;

[0008] Based on the working capabilities and task requirements of each agent, tasks are initially matched. A task allocation mechanism is designed using reinforcement learning algorithms. Local information between agents is used to make task decisions, and the task allocation method is dynamically adjusted during the task allocation process.

[0009] For each agent, the system acquires operational status and task progress information through sensors and real-time monitoring devices, analyzes the agent's status and task execution, and dynamically adjusts task allocation and collaborative control based on feedback information.

[0010] Furthermore, the process of collecting state information of the multi-agent system, including work capacity, task requirements, and environmental change data, and modeling and classifying the multi-agent system in the industrial field, specifically includes:

[0011] The state information of the multi-agent system is obtained by using sensors and monitoring equipment, and integrated into a working capability vector. By aggregating the working capabilities of all agents, the working capability vector of the entire multi-agent system is obtained.

[0012] Use the task management system to obtain the resources required for each task of the multi-agent system, and calculate the total resource requirements for each task.

[0013] Environmental monitoring sensors are used to collect environmental data in industrial sites and calculate environmental variables such as temperature, humidity, and vibration.

[0014] The entire multi-agent system is modeled as a system state model that includes the agent's working capabilities, task requirements, and environmental changes. Based on the data of working capabilities, task requirements, and environmental changes, the state of each agent and the set of all agent states are obtained.

[0015] Furthermore, based on the working capabilities and task requirements of each agent, tasks are initially matched; a task allocation mechanism is designed using reinforcement learning algorithms; task decisions are made using local information among agents; and the task allocation method is dynamically adjusted during the allocation process. Specifically, this includes:

[0016] Initialize the working capabilities, task requirements, and environmental change data of each agent, and perform preliminary task allocation based on the working capabilities and task requirements of the agents through a matching function;

[0017] Based on the state of each agent, the resource requirements of each task, and environmental change data, reinforcement learning algorithms are used to adjust the task allocation strategy and optimize task execution.

[0018] Furthermore, the step of adjusting the task allocation strategy and optimizing task execution performance using a reinforcement learning algorithm based on the state of each agent, the resource requirements of each task, and environmental change data further includes:

[0019] The goal of reinforcement learning is to maximize long-term rewards. At each time step, the agent chooses whether to assign a task based on its current state and calculates the reward based on the performance. The agent updates its Q-value through interaction with the environment to optimize the task assignment strategy.

[0020] Furthermore, for each intelligent agent, the system acquires operational status and task progress information through sensors and real-time monitoring devices, analyzes the agent's status and task execution, and dynamically adjusts task allocation and collaborative control based on feedback information. Specifically, this includes:

[0021] Extract the working capabilities and task resource requirements of agents from the state information of the multi-agent system to obtain the computing resources required for agents to execute tasks, as well as the total computing resource requirements of the entire system at the current moment. By weighted summing of the loads of different resources, the total load of each agent is obtained.

[0022] Furthermore, based on the agent's load results and task progress, the task allocation is dynamically adjusted. When the total load exceeds the preset load threshold, the task is transferred to a neighboring agent with a lighter load. Based on the information of the neighboring agents, the load difference between the neighboring agents is calculated, and the cooperation strategy between the agents is adjusted.

[0023] Furthermore, based on load differences and using a feedback reward function, the Q value is updated according to task completion, load, and progress. Through continuous updates to the Q value, the agent can autonomously optimize its task allocation and collaboration strategies to adapt to dynamically changing environments and task requirements. During system operation, key indicators such as task execution effectiveness, resource consumption, and agent load are periodically evaluated. Based on the evaluation results, the task allocation strategy and collaborative control mechanism are continuously optimized.

[0024] Furthermore, during system operation, key indicators such as task execution effectiveness, resource consumption, and agent load are periodically evaluated. Based on the evaluation results, the task allocation strategy and collaborative control mechanism are continuously optimized, further including:

[0025] Set an execution quality function to evaluate the task execution effect based on the task completion time, agent load, and task progress.

[0026] The performance of the agent's current task is evaluated based on the execution quality function, and the feedback signal is converted into the information required for Q-value updates. After each task is completed, the Q-value is updated based on the current execution quality function value. Based on the updated Q-value, the agent's task allocation strategy is dynamically adjusted to select agents with shorter task execution time and lighter load to perform new tasks.

[0027] Furthermore, it also includes an agent fault adaptive compensation step, specifically: during system operation, the communication link connectivity status, task execution error rate and key component operating temperature of the agent are monitored in real time by sensors. When the agent is detected to meet the fault judgment conditions, a fault marker is triggered and a compensation mechanism is started.

[0028] Based on the working capacity of the remaining agents in the system state set, the total task requirements of the currently unfinished tasks, and environmental change data, the unfinished tasks of the faulty agents are reallocated by improving the matching function, so as to ensure the real-time performance and reliability of task allocation in fault scenarios.

[0029] On the other hand, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements the steps of the distributed task allocation and cooperative control method for heterogeneous multi-agent systems as described above.

[0030] On the other hand, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements the steps of the distributed task allocation and cooperative control method for heterogeneous multi-agent systems as described above.

[0031] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0032] By collecting and analyzing the state information of a multi-agent system in real time, including work capacity, task requirements, and environmental change data, the system can be made more adaptable to changes in the external environment, task requirements, and agent states. A task allocation mechanism is designed using reinforcement learning algorithms, utilizing local information between agents for task decision-making, avoiding the bottleneck problem of traditional centralized methods and improving task execution efficiency. By introducing an adaptive task allocation mechanism and dynamic adjustment of task allocation and collaborative control, the system can rationally allocate tasks according to the actual work capacity and load of each agent, avoiding resource waste. This invention, by combining reinforcement learning, distributed control, real-time monitoring, and feedback mechanisms, can achieve efficient task allocation and collaborative control in complex and dynamic multi-agent systems. Attached Figure Description

[0033] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0034] Figure 1 This is a flowchart of the distributed task allocation and cooperative control method for heterogeneous multi-agent systems according to the present invention. Detailed Implementation

[0035] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0036] Example 1

[0037] This embodiment provides, for example Figure 1 The distributed task allocation and cooperative control method for heterogeneous multi-agent systems shown includes:

[0038] Step 1 involves collecting state information from the multi-agent system, including its capabilities, task requirements, and environmental changes, and then modeling and classifying the multi-agent system in the industrial setting. This ensures that tasks are accurately matched to the actual capabilities of each agent, avoiding resource waste and improving task execution efficiency. Specifically, this includes:

[0039] By acquiring data on the battery level, motor speed, and task execution duration of the intelligent agent system through sensors and monitoring equipment, these data are integrated into a vector of operational capabilities: By aggregating the capabilities of all agents, the capability vector of the entire multi-agent system is obtained as follows: ,in, This represents the working capability vector of agent i. This is the overall system's operational capability vector, where N is the number of agents in the system. This refers to the battery level of agent i. It is the motor speed of agent i. It is the duration for agent i to perform the task;

[0040] Use a task management system to obtain the resources required for each task in the multi-agent system, and represent the resource requirements of task j as follows: For the entire task set, the sum of task requirements is expressed as: ,in, , , These are the computing resources, communication bandwidth, and space resources required for task j, respectively. It is the sum of all task requirements, where M is the number of tasks;

[0041] Environmental monitoring sensors are used to collect environmental data in industrial sites, obtaining environmental variables such as temperature, humidity, and vibration, which are represented as follows: ,in, It is environmental change data. Indicates the ambient temperature. Indicates the humidity of the environment. This indicates the vibration conditions in the environment, where t represents time.

[0042] The entire multi-agent system is modeled as a system state model that includes the agents' capabilities, task requirements, and environmental changes, with each agent's state defined. At time t, it is represented as The set of all agent states is obtained as follows ,in, It is the specific value of the state vector of the Nth agent in the system at time t. It is the set of all agent states of the entire multi-agent system at time t.

[0043] Step 2: Based on the capabilities of each agent and the requirements of the task, tasks are initially matched. A task allocation mechanism is designed using reinforcement learning algorithms, utilizing local information between agents to make task decisions and dynamically adjusting the task allocation method during the allocation process. This reduces reliance on global information, lowers the communication burden, and continuously optimizes the task allocation strategy through reinforcement learning, enhancing the collaborative capabilities among agents. Specifically, this includes:

[0044] Initialize the working capability vector of each agent. Task resource requirements and environmental change data Based on the agent's capabilities and task requirements, preliminary task allocation is performed using a matching function, which is: ,in, f represents the matching decision of whether task j is assigned to agent i, where f is a matching function.

[0045] Based on the state of each agent Resource requirements for each task and environmental change data The Q-learning algorithm is used to adjust the task allocation strategy and optimize task execution performance, further including:

[0046] The goal of reinforcement learning is to maximize long-term reward. At each time step t, the agent adjusts its current state... Choose an action , represents the action taken by agent i at time t, where the action is "whether to assign a task". After the action is completed, a reward is calculated based on the performance of the task. The agent updates its Q-value through interaction with the environment to optimize the task allocation strategy. The Q-value reflects the expected reward of a state-action pair, and the update formula for the Q-value is as follows: Where α is the learning rate, which controls the speed of updates. The agent i takes an action at time t. Q value, This is a discount factor, representing the degree of importance placed on future returns. Indicates the agent's next state The maximum Q value that can be obtained is... Indicates the next state Any action below, It is the reward that agent i receives at time t.

[0047] Step 3: For each agent, acquire its operational status and task progress information through sensors and real-time monitoring devices, analyze the agent's status and task execution, and dynamically adjust task allocation and collaborative control based on feedback information. This also avoids overloading individual agents, enabling the system to operate efficiently even when facing large-scale tasks. Specifically, this includes:

[0048] Extracting the working capabilities and task resource requirements of agents from the state information of a multi-agent system, we obtain the computational resources, communication bandwidth, and spatial resources required by agent i for the task executed at time t. , , The total computing resource requirements, total communication bandwidth requirements, and total spatial resource requirements of the entire multi-agent system at time t are as follows: , , By weighted summing of the loads on different resources, the total load for each agent is obtained. ,in, That is the battery's maximum capacity. λ1 represents the maximum speed of the motor, and λ2, λ3 are the weighting coefficients for different resources. It is the total load of agent i. This refers to the battery level of agent i. It is the motor speed of agent i.

[0049] Based on the agent's load results and task progress, dynamically adjust task allocation and set a load threshold. ,when The task is transferred from agent i to agent k with a lighter load. Based on the information from neighboring agents, the load difference between the neighboring agents is calculated as follows: , Adjusting the cooperation strategy between intelligent agents, among which... It is the total load of agent i. It is the total load of agent k.

[0050] The adjustment of cooperation strategies among intelligent agents is based on a distributed communication protocol, and the specific process is as follows:

[0051] 1. Collaboration Trigger Judgment: Every 500ms, the current agent broadcasts its load status to neighboring agents within a 10-meter radius, while simultaneously receiving load information from neighboring agents and calculating the load difference. .

[0052] 2. Collaboration mode selection: Based on the value range of ΔL, select one of three collaboration modes: task transfer, resource sharing, or maintaining the status quo.

[0053] 3. Collaborative Execution and Feedback: If task transfer is initiated, it will be done through a matching function. ,in, Given the spatial distance between agents i and k, select the optimal receiving object. After the task transfer is completed, both parties synchronously update their local Q values. This is the maximum load threshold. This represents the current total load of agent k. If resource sharing is initiated, a temporary communication link is established through the edge node, and the usage of shared resources is fed back to the execution quality function in real time for subsequent optimization of the cooperation strategy.

[0054] Collaboration termination condition: When the load difference between the current agent and all neighboring agents is satisfied. At that time, the collaborative adjustment will be terminated, and the normal task execution status will be restored.

[0055] Based on load differences and using a feedback reward function, the Q-value is updated according to task completion, load, and progress. Through continuous updates to the Q-value, the agent can autonomously optimize its task allocation and cooperation strategies to adapt to dynamically changing environments and task requirements. During system operation, key indicators such as task execution effectiveness, resource consumption, and agent load are periodically evaluated. Based on the evaluation results, the task allocation strategy and cooperative control mechanism are continuously optimized, further including:

[0056] Set an execution quality function The execution quality function evaluates the task execution performance based on the task completion time, agent load, and task progress. ,in, This is the ideal time to complete the task. This is the actual task completion time. This is the current progress of the task. This is the maximum progress of the task. It is the load of the intelligent agent. , and These are weighting coefficients, used to adjust the impact of time, schedule, and workload on task quality.

[0057] Based on the execution quality function The effectiveness of the agent's current task execution is evaluated by converting the feedback signal into the information needed for Q-value updates through the execution quality function. After each task execution, the Q-value is updated based on the current execution quality function value. Based on the updated Q-value, the agent's task allocation strategy is dynamically adjusted, selecting agents with short task execution times and light loads to execute new tasks.

[0058] This embodiment also provides a computer device, including a memory and a processor. The memory stores a computer program, wherein when the computer program is executed by the processor, it implements the steps of the distributed task allocation and cooperative control method for heterogeneous multi-agent systems as described above in this invention.

[0059] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0060] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the steps of the distributed task allocation and cooperative control method for heterogeneous multi-agent systems proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0061] Example 2

[0062] The following is another embodiment of the present invention, which provides a distributed task allocation and cooperative control method for heterogeneous multi-agent systems. In order to verify the beneficial effects of the present invention, a simulation experiment is conducted for scientific demonstration.

[0063] This experiment used simulated and actual collected state data of a multi-agent system. By analyzing the consistency between the task allocation and execution effects output by the model and the actual operating state, the accuracy and robustness of the system in task allocation and collaborative control were verified. The aim was to improve the working efficiency and collaborative ability of the multi-agent system and achieve efficient task allocation and execution by means of multi-source data acquisition, data preprocessing, feature extraction, reinforcement learning models and task allocation techniques.

[0064] The simulation experiment steps are implemented according to the distributed task allocation and cooperative control method for heterogeneous multi-agent systems provided in Example 1. The specific steps include:

[0065] Collect state information of the multi-agent system, including working capacity, task requirements and environmental change data. The data collection frequency is set to 1Hz and the collection time is set to 30 minutes.

[0066] The collected multi-agent system data is modeled and classified, the capabilities and task requirements of each agent are analyzed, and tasks are assigned to each agent through preliminary matching.

[0067] A task allocation mechanism is designed using reinforcement learning algorithms, which combines local information among intelligent agents to make task decisions and dynamically adjusts the task allocation method during the task allocation process.

[0068] For each agent, the system acquires operational status and task progress information through sensors and real-time monitoring devices, analyzes the agent's status and task execution, and dynamically adjusts task allocation and collaborative control based on feedback information.

[0069] During system operation, key indicators such as task execution effectiveness, resource consumption, and agent load are evaluated periodically, and task allocation strategies and collaborative control mechanisms are continuously optimized based on the evaluation results.

[0070] The specific data from the above simulation experiment are as follows:

[0071] Table 1. Data Recording Table of Simulation Experiments for the Distributed Task Allocation and Cooperative Control Method for Heterogeneous Multi-Agent Systems of the Present Invention

[0072] Experimental Analysis:

[0073] By dynamically adjusting the task allocation and collaborative control mechanism in the multi-agent system, this system can capture changes in the agent's state in real time and accurately adjust the task allocation. When the agent's workload is too high or the task execution is delayed, the system prioritizes task reallocation and collaborative control, ensuring the timely completion and coordination of tasks, and significantly improving the system's efficiency and accuracy.

[0074] The system's task allocation threshold, through analysis of historical data and assessment of agent load, can effectively determine whether tasks need to be reallocated or workload adjusted, thus avoiding resource waste and task delays.

[0075] Experimental results show that when task execution efficiency decreases or resource consumption is too high, the system can respond in a timely manner and activate the corresponding task adjustment mechanism, accurately allocate tasks and optimize resource use, thereby improving the collaborative ability and operating efficiency of the multi-agent system.

[0076] Example 3

[0077] This embodiment, based on embodiment 1, adds a deployment and execution process for adaptive compensation of intelligent agent faults, as detailed below:

[0078] Fault monitoring deployment: A communication link monitoring module, a task execution error calculation module, and a temperature sensor are added to each intelligent agent. The measurement range is -40℃ to 125℃, and the accuracy is ±0.3℃. The communication link monitoring module checks the communication connectivity status with the edge node every 100ms. The task execution error calculation module compares the preset standard value of the task with the actual execution result in real time. The temperature sensor monitors the operating temperature of key components such as motors and controllers.

[0079] Fault Judgment and Triggering: Set fault judgment thresholds: communication interruption duration ≥3s, task execution error rate ≥15%, or critical component temperature ≥85℃. When the agent meets any of the threshold conditions, it immediately sends a fault signal to the distributed data aggregation node. The node marks the agent as "fault state" and suspends the assignment of new tasks to it. At the same time, it locks its list of unfinished tasks, including task ID, execution progress, and remaining resource requirements.

[0080] Fault Task Reassignment: Distributed data aggregation nodes extract the working capability vectors of remaining "normal state" agents in the system. Total resource requirements for unfinished tasks j represents the unfinished tasks of the faulty agent. Resource requirements for currently unfinished tasks, and current environmental data. Set the weight coefficients of the improved matching function. , , Calculate the matching degree between each normal agent and the faulty task. The top three agents with the highest matching degree are selected as candidate assignment targets, and then the assignment is based on the differences in the load of the candidate agents. The task is assigned to the agent with the lightest workload, among which... The total load of candidate agent i currently being evaluated; For the total load of another candidate agent k, The absolute difference in load between the two candidate agents.

[0081] Optimizing reinforcement learning strategies: Setting compensation incentive coefficients When the faulty task is successfully completed, the reward function... Where R is the base reward value, This is the adjusted reward value; if not completed, Update formula using Q value Continuously optimize decision-making strategies under fault scenarios;

[0082] in, η is the Q-value corresponding to taking action a in state S; η is the learning rate; The new state reached after performing the action; The adjusted instant reward; Discount factor; In the new state The maximum Q value among all possible actions.

[0083] Experimental Verification: Based on the simulation experiment of Example 2, a new fault scenario test was added: This example summarizes the experiments where three faults of the handling robot were artificially triggered during the 15-20 minute high-load phase, such as communication interruption. The test results show:

[0084] The average fault detection response time is 1.2 seconds;

[0085] The average interruption time for faulty tasks was 3.8 seconds, which is 69.6% shorter than the average interruption time of 12.5 seconds for traditional methods without compensation mechanisms.

[0086] Under fault scenarios, the overall task completion rate was 89%, which is 41.3% higher than the 63% completion rate of traditional methods, fully verifying the effectiveness of the fault adaptive compensation mechanism.

[0087] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A method for distributed task allocation and cooperative control for a heterogeneous multi-agent system, characterized in that, include: Collect state information of multi-agent systems, including working capacity, task requirements and environmental change data, and model and classify multi-agent systems in industrial sites; Based on the working capabilities and task requirements of each agent, tasks are initially matched. A task allocation mechanism is designed using reinforcement learning algorithms. Local information between agents is used to make task decisions, and the task allocation method is dynamically adjusted during the task allocation process. For each agent, the system acquires operational status and task progress information through sensors and real-time monitoring devices, analyzes the agent's status and task execution, and dynamically adjusts task allocation and collaborative control based on feedback information.

2. The method of claim 1, wherein: The process involves collecting state information from multi-agent systems, including operational capabilities, task requirements, and environmental change data, and modeling and classifying multi-agent systems in industrial settings. Specifically, this includes: The state information of the multi-agent system is obtained by using sensors and monitoring equipment, and integrated into a working capability vector. By aggregating the working capabilities of all agents, the working capability vector of the entire multi-agent system is obtained. Use the task management system to obtain the resources required for each task of the multi-agent system, and calculate the total resource requirements for each task. Environmental monitoring sensors are used to collect environmental data in industrial sites and calculate environmental variables such as temperature, humidity, and vibration. The entire multi-agent system is modeled as a system state model that includes the agent's working capabilities, task requirements, and environmental changes. Based on the data of working capabilities, task requirements, and environmental changes, the state of each agent and the set of all agent states are obtained.

3. The method of claim 1, wherein: Based on the working capabilities and task requirements of each agent, tasks are initially matched. A task allocation mechanism is designed using reinforcement learning algorithms. Task decisions are made using local information among agents, and the task allocation method is dynamically adjusted during the allocation process. Specifically, this includes: Initialize the working capabilities, task requirements, and environmental change data of each agent, and perform preliminary task allocation based on the working capabilities and task requirements of the agents through a matching function; Based on the state of each agent, the resource requirements of each task, and environmental change data, reinforcement learning algorithms are used to adjust the task allocation strategy and optimize task execution.

4. The distributed task allocation and cooperative control method for heterogeneous multi-agent systems according to claim 3, characterized in that: The step of adjusting the task allocation strategy and optimizing task execution performance using reinforcement learning algorithms based on the state of each agent, the resource requirements of each task, and environmental change data further includes: The goal of reinforcement learning is to maximize long-term rewards. At each time step, the agent chooses whether to assign a task based on its current state and calculates the reward based on the performance. The agent updates its Q-value through interaction with the environment to optimize the task assignment strategy.

5. The distributed task allocation and cooperative control method for heterogeneous multi-agent systems according to claim 1, characterized in that: For each intelligent agent, operational status and task progress information are acquired through sensors and real-time monitoring devices. The status and task execution of the intelligent agent are analyzed, and task allocation and collaborative control are dynamically adjusted based on feedback information. Specifically, this includes: Extract the working capabilities and task resource requirements of the agents from the state information of the multi-agent system to obtain the computing resources required for the agents to execute tasks, as well as the total computing resource requirements of the entire system at the current moment. By weighted summing of the loads of different resources, the total load of each agent is obtained.

6. The distributed task allocation and cooperative control method for heterogeneous multi-agent systems according to claim 5, characterized in that: Based on the agent's load results and task progress, the task allocation is dynamically adjusted. When the total load exceeds the preset load threshold, the task is transferred to the neighboring agent with a lighter load. Based on the information of the neighboring agents, the load difference of the neighboring agents is calculated, and the cooperation strategy between the agents is adjusted.

7. The distributed task allocation and cooperative control method for heterogeneous multi-agent systems according to claim 6, characterized in that: Based on load differences and using a feedback reward function, the Q value is updated according to task completion, load, and progress. Through continuous updates to the Q value, the agent can autonomously optimize its task allocation and collaboration strategies to adapt to dynamically changing environments and task requirements. During system operation, key indicators such as task execution effectiveness, resource consumption, and agent load are periodically evaluated. Based on the evaluation results, the task allocation strategy and collaborative control mechanism are continuously optimized.

8. The distributed task allocation and cooperative control method for heterogeneous multi-agent systems according to claim 7, characterized in that: During system operation, key indicators such as task execution effectiveness, resource consumption, and agent load are periodically evaluated. Based on the evaluation results, the task allocation strategy and collaborative control mechanism are continuously optimized, further including: Set an execution quality function to evaluate the task execution effect based on the task completion time, agent load, and task progress. The performance of the agent in the current task is evaluated based on the execution quality function, and the feedback signal is converted into the information required for Q-value updates. After each task is completed, the Q-value is updated based on the current execution quality function value. Based on the updated Q-value, the agent's task allocation strategy is dynamically adjusted to select agents with shorter task execution time and lighter load to perform new tasks.

9. The distributed task allocation and cooperative control method for heterogeneous multi-agent systems according to claim 1, characterized in that: It also includes an agent fault adaptive compensation step, specifically: during system operation, the communication link connectivity status, task execution error rate and key component operating temperature of the agent are monitored in real time by sensors. When the agent is detected to meet the fault judgment conditions, a fault mark is triggered and the compensation mechanism is started. Based on the working capacity of the remaining agents in the system state set, the total task requirements of the currently unfinished tasks, and environmental change data, the unfinished tasks of the faulty agents are reallocated by improving the matching function, so as to ensure the real-time performance and reliability of task allocation in fault scenarios.

10. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the distributed task allocation and cooperative control method for heterogeneous multi-agent systems as described in any one of claims 1-9.

Images