A method and device for cooperative control of a multi-agent system

Abstract

The application provides a kind of multi-agent system's cooperative control method and device, it is related to agent control technical field, comprising: obtaining task set and role set, task set includes multiple tasks, role set includes multiple roles;Obtain the attribute set of task set, according to attribute set, task and role are matched, obtain first mapping relationship set;Obtain the agent state vector set of multi-agent system, multi-agent system includes multiple agents;According to agent state vector set, agent and role are matched, obtain second mapping relationship set;Obtain the agent skill vector set of multi-agent system, according to attribute set, agent state vector set, agent skill vector set, first mapping relationship set and second mapping relationship set, task is assigned to agent, obtain execution agent set;Execution agent in execution agent set executes task according to execution sequence.

Description

Technical Field

[0001] This invention relates to the field of intelligent agent control technology, and in particular to a collaborative control method and apparatus for a multi-agent system. Background Technology

[0002] With the widespread application of artificial intelligence and automation systems, multi-agent systems (MAS) are widely used in scenarios such as distributed decision-making, collaborative control, and complex task processing. However, traditional MAS systems often predefine tasks and roles, lacking flexibility and autonomy, and are difficult to adapt to dynamic environments and task diversity, specifically including the following shortcomings:

[0003] 1. In some existing role-based task collaboration systems, although the mapping from role to task and from task to agent can be realized, most methods lack the agent's self-learning ability and cannot adjust skill strategies based on task execution feedback. At the same time, there is also a lack of a sound governance and negotiation mechanism, resulting in slow response and insufficient system robustness in conflicting tasks, task competition and cooperation or heterogeneous collaboration scenarios.

[0004] 2. The preset role and task allocation mode adopted by the traditional MAS system is difficult to cope with sudden tasks or environmental changes. For example, when a sudden accident in traffic scheduling causes local path failure, the traditional MAS system requires manual intervention to re-plan the control strategy of the intelligent agent, which is time-consuming and labor-intensive.

[0005] 3. Most MAS systems rely on fixed strategies and cannot optimize collaboration efficiency through task feedback. For example, during collaborative assembly by industrial robotic arms, if a link fails, the system may stall due to a lack of adaptive adjustment capabilities. Summary of the Invention

[0006] To address the above problems, this invention provides a cooperative control method for a multi-agent system, comprising the following steps:

[0007] Get the task set and the role set. The task set includes multiple tasks, and the role set includes multiple roles. Get the attribute set of the task set. Match the tasks and roles according to the attribute set to obtain the first mapping relationship set.

[0008] Obtain the set of agent state vectors for a multi-agent system, which includes multiple agents; match agents with roles based on the set of agent state vectors to obtain a second set of mapping relationships;

[0009] Obtain the agent skill vector set of the multi-agent system, and assign tasks to agents according to the attribute set, agent state vector set, agent skill vector set, first mapping relationship set and second mapping relationship set to obtain the execution agent set;

[0010] Obtain the execution sequence of the task set, and execute the tasks according to the execution sequence of the execution agents in the execution agent set.

[0011] Optionally, the step of matching tasks with roles based on the attribute set to obtain a first mapping relationship set specifically includes:

[0012] S11: Obtain the attributes of the i-th task from the attribute set. The attributes include: task skill requirement vector, task target location vector, time attribute vector, and task constraints. Select all roles that match the attributes from the role set. Construct a one-to-many mapping relationship between the i-th task and all roles that match the attributes.

[0013] S12: Repeat step S11 until all task attributes are traversed, obtain all one-to-many mapping relationships between tasks and roles, and use all one-to-many mapping relationships between tasks and roles as the first mapping relationship set.

[0014] Optionally, the step of matching agents with roles based on the agent state vector set to obtain a second mapping relationship set specifically includes:

[0015] S21: Select the k-th character R k ;

[0016] S22: Obtain the j-th agent A from the agent state vector set. j agent state vector s j According to the agent's state vector s j Agent A is obtained through deep evaluation function calculation. j With character R k The first rating between them;

[0017] S23: Repeat step S22 to calculate and obtain R for all agents and roles. k The first score between them, the agent and role R corresponding to the highest first score. k Establish a one-to-one mapping relationship between them;

[0018] S24: Repeat steps S21-S23 until all roles are traversed, obtain all one-to-one mapping relationships between agents and roles, and use all one-to-one mapping relationships between agents and roles as the second mapping relationship set.

[0019] Optionally, the step of assigning tasks to agents based on the attribute set, agent state vector set, agent skill vector set, first mapping relationship set, and second mapping relationship set to obtain an execution agent set specifically includes:

[0020] S31: Select the i-th task in the task set, obtain the task skill requirement vector and task target location vector of the i-th task from the attribute set, obtain all the roles corresponding to the i-th task according to the first mapping relationship set, and obtain the intelligent agent corresponding to each role according to the second mapping relationship set;

[0021] S32: Select the j-th agent corresponding to the i-th task, obtain the agent skill vector of the j-th agent from the agent skill vector set, obtain the agent state vector of the j-th agent from the agent state vector set, and obtain the agent position vector, current task number, maximum task capacity and historical trust degree of the j-th agent based on the agent state vector;

[0022] S33: Calculate the second score between the i-th task and the corresponding j-th agent based on the task skill requirement vector, agent skill vector, task destination location vector, agent location vector, current number of tasks, maximum task capacity, and historical trust level TS.

[0023] S34: Repeat steps S32-S33 until the second score between the i-th task and all corresponding agents is calculated. The agent with the largest second score is taken as the executing agent of the i-th task, and the i-th task is assigned to the executing agent.

[0024] S35: Repeat steps S31-S34 until all tasks are traversed, obtain all executing agents, and set all executing agents as a set of executing agents.

[0025] Optionally, the calculation process for the second score specifically includes:

[0026] Skill similarity (SM) is calculated based on the task skill requirement vector and the agent skill vector; geographical location score (LM) is calculated based on the task destination location vector and the agent location vector; and resource load score (LS) is calculated based on the current number of tasks and the maximum task capacity.

[0027] Set the first weight Second weight Third weight and the fourth weight ;

[0028] Will This serves as a second rating between the task and the corresponding agent.

[0029] Optionally, obtaining the execution sequence of the task set, and having the execution agents in the execution agent set execute tasks according to the execution sequence, specifically includes:

[0030] Obtain the urgency index and importance index of each task in the task set, calculate the priority index based on the urgency index and importance index, sort the priority indices of each task from largest to smallest to construct an execution sequence, and select the corresponding execution agents to execute the tasks in the order of priority indices in the execution sequence.

[0031] Optionally, after the executing agents in the set of executing agents perform tasks according to the execution sequence, the set of executing agents may also include:

[0032] Obtain the task completion results of each executing agent, set adjustment values ​​based on the task completion results. If the task completion result is successful, set the adjustment value to the first preset value; if the task completion result is unsuccessful, set the adjustment value to the second preset value. Update the agent skill vector of the executing agent based on the adjustment values. Calculate the task success rate based on the task completion results, and update the historical trust level of the executing agent based on the task success rate.

[0033] The present invention also provides a cooperative control device for a multi-agent system, used to implement the cooperative control method for the multi-agent system, the device comprising:

[0034] The first mapping relationship set acquisition module is used to acquire a task set and a role set. The task set includes multiple tasks, and the role set includes multiple roles. It acquires the attribute set of the task set, matches the tasks with the roles based on the attribute set, and obtains the first mapping relationship set.

[0035] The second mapping relationship set acquisition module is used to acquire the agent state vector set of a multi-agent system, which includes multiple agents; and to match agents with roles based on the agent state vector set to obtain the second mapping relationship set.

[0036] The execution agent set acquisition module is used to acquire the agent skill vector set of the multi-agent system, and assign tasks to agents according to the attribute set, agent state vector set, agent skill vector set, first mapping relationship set and second mapping relationship set to obtain the execution agent set;

[0037] The task execution module is used to obtain the execution sequence of the task set, and the execution agents in the execution agent set execute the tasks according to the execution sequence.

[0038] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the cooperative control method of the multi-agent system.

[0039] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the cooperative control method of the multi-agent system.

[0040] The present invention has the following beneficial effects:

[0041] 1. Based on the attribute set of the task set, tasks are matched with roles to obtain a first mapping relationship set. The first mapping relationship set fully considers the task skill requirement vector, task target location vector, time attribute vector, and task constraints, which are parameters that can be learned and updated. Based on the agent state vector set, agents are matched with roles to obtain a second mapping relationship set. The second mapping relationship set fully considers the agent's state vector, which can improve the agent's perception of the current environment. Based on the attribute set, agent state vector set, agent skill vector set, first mapping relationship set, and second mapping relationship set, tasks are assigned to agents, allowing agents to continuously optimize task execution through self-learning and improve task success rate.

[0042] 2. After the agent executes the task according to the execution sequence, the agent's skill vector and historical trust level are updated according to the task completion result, so that the agent's parameters are continuously adjusted in the direction of improving the task success rate, thereby improving the adaptive adjustment capability of the multi-agent system. Attached Figure Description

[0043] Figure 1 This is a flowchart of a method according to an embodiment of the present invention;

[0044] Figure 2 This is a structural diagram of the device according to an embodiment of the present invention;

[0045] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0046] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the embodiments of this application. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0047] Reference Figure 1 This invention provides a cooperative control method for a multi-agent system, comprising the following steps:

[0048] Get the task set and the role set. The task set includes multiple tasks, and the role set includes multiple roles. Get the attribute set of the task set. Match the tasks and roles according to the attribute set to obtain the first mapping relationship set.

[0049] In some embodiments, the step of matching tasks with roles based on an attribute set to obtain a first mapping relationship set specifically includes:

[0050] S11: Obtain the attributes of the i-th task from the attribute set. The attributes include: task skill requirement vector, task target location vector, time attribute vector, and task constraints. Select all roles that match the attributes from the role set. Construct a one-to-many mapping relationship between the i-th task and all roles that match the attributes.

[0051] In some embodiments, each task is bound to a task skill requirement vector, where the task skill requirement vector req for the i-th task is... i The expression is as follows:

[0052]

[0053] Where r k Let k represent the k-th skill. Taking the spraying task as an example, the task skill requirement vector for the spraying task is [navigation (0.8), dosage control (0.9), spraying angle (0.6)].

[0054] The time attribute vector includes the latest task completion time and estimated execution time. Task constraints include spatial location constraints, whether it can only be completed by a specific role, whether collaboration is required, etc., which are set according to the specific task.

[0055] For each task, first query the role template library based on the task's attributes and select all roles that match the attributes. For example, roles that match the wine spraying task include flying wine sprayers and low-altitude identification personnel.

[0056] S12: Repeat step S11 until all task attributes are traversed, obtain all one-to-many mapping relationships between tasks and roles, and use all one-to-many mapping relationships between tasks and roles as the first mapping relationship set.

[0057] Obtain the set of agent state vectors for a multi-agent system, which includes multiple agents; match agents with roles based on the set of agent state vectors to obtain a second set of mapping relationships;

[0058] In some embodiments, the step of matching agents with roles based on the agent state vector set to obtain a second mapping relationship set specifically includes:

[0059] S21: Select the k-th character Rk ;

[0060] S22: Obtain the j-th agent A from the agent state vector set. j agent state vector s j According to the agent's state vector s j Agent A is obtained through deep evaluation function calculation. j With character R k The first rating between them;

[0061] In some embodiments, the agent state vector s j Represents a certain intelligent agent The current capability status, behavioral trends, and perception results, including parameters such as the agent's agent position vector, current number of tasks, maximum task capacity, and historical trust level;

[0062] First rating The expression is as follows:

[0063]

[0064] Here, MLP() represents a deep evaluation function built on a multilayer perceptron neural network. The deep evaluation function includes an input layer, hidden layers, and an output layer. The input layer receives the agent's state vector and role ID, with each neuron corresponding to a feature dimension. The hidden layers are multilayer fully connected structures, with each neuron connected to the output of the previous layer via weights. Non-linear activation functions (such as ReLU) are used to introduce non-linearity, addressing complex relationships that linear models cannot handle. The output layer consists of a single neuron and a linear activation function, used to output the first score.

[0065] S23: Repeat step S22 to calculate and obtain R for all agents and roles. k The first score between them, the agent and role R corresponding to the highest first score. k Establish a one-to-one mapping relationship between them;

[0066] S24: Repeat steps S21-S23 until all roles are traversed, obtain all one-to-one mapping relationships between agents and roles, and use all one-to-one mapping relationships between agents and roles as the second mapping relationship set.

[0067] Obtain the agent skill vector set of the multi-agent system, and assign tasks to agents according to the attribute set, agent state vector set, agent skill vector set, first mapping relationship set and second mapping relationship set to obtain the execution agent set;

[0068] In some embodiments, the step of assigning tasks to agents based on an attribute set, an agent state vector set, an agent skill vector set, a first mapping relationship set, and a second mapping relationship set to obtain an execution agent set specifically includes:

[0069] S31: Select the i-th task in the task set, obtain the task skill requirement vector and task target location vector of the i-th task from the attribute set, obtain all the roles corresponding to the i-th task according to the first mapping relationship set, and obtain the intelligent agent corresponding to each role according to the second mapping relationship set;

[0070] S32: Select the j-th agent corresponding to the i-th task, obtain the agent skill vector of the j-th agent from the agent skill vector set, obtain the agent state vector of the j-th agent from the agent state vector set, and obtain the agent position vector, current task number, maximum task capacity and historical trust degree of the j-th agent based on the agent state vector;

[0071] In some embodiments, each agent has an agent skill vector, which consists of multiple skill parameters representing its capability state under multiple skill inferences. Taking a spraying task as an example, the agent skill vector includes at least the following skill parameters: pathfinding, navigation precision, vision recognition, spray control, area coverage, task timing, and communication stability. Each skill parameter's success rate or performance estimate is calculated using a sliding window statistical or learning algorithm, and the resulting agent skill vector is used for subsequent role matching and task scheduling.

[0072] S33: Calculate the second score between the i-th task and the corresponding j-th agent based on the task skill requirement vector, agent skill vector, task destination location vector, agent location vector, current number of tasks, maximum task capacity, and historical trust level TS.

[0073] In some embodiments, the calculation process of the second score specifically includes:

[0074] Skill similarity (SM) is calculated based on the task skill requirement vector and the agent skill vector; geographical location score (LM) is calculated based on the task destination location vector and the agent location vector; and resource load score (LS) is calculated based on the current number of tasks and the maximum task capacity.

[0075] Set the first weight Second weight Third weight and the fourth weight ;

[0076] Will This serves as a second rating between the task and the corresponding agent.

[0077] In some embodiments, cosine similarity is used to calculate the skill similarity SM (SkilMatch) between the task skill requirement vector and the agent skill vector, and the calculation formula is as follows:

[0078]

[0079] in, This represents the task skill requirement vector for the i-th task. Let SM represent the agent skill vector of the j-th agent, where || represents the magnitude of the vector; SM is 1 if there is a perfect match, and SM is 0 if there is no perfect match.

[0080] The coordinates of the task target location are obtained through the task target location vector, and the coordinates of the agent's location are obtained through the agent's location vector. The location score LM (LocationMatch) is calculated based on the Euclidean distance between the task target location coordinates and the agent's location coordinates. The calculation formula is as follows:

[0081]

[0082] in, This represents the Euclidean distance between the target coordinates of the i-th task and the agent position coordinates of the j-th agent. Indicates the maximum tolerable distance within the mission area;

[0083] The LoadScore (LS) represents the current idle level of the agent's task load, and is calculated using the following formula:

[0084]

[0085] in, This represents the current task number of the j-th agent. LS represents the maximum task capacity of the j-th agent; LS is 1 when completely idle and 0 when the task is fully loaded.

[0086] The historical trust score (TS) is calculated based on the success rate or failure rate of tasks within a sliding window. For example, if 8 out of the last 10 tasks are successful, the historical trust score is 0.8.

[0087] The formula for calculating the second score is as follows:

[0088]

[0089] Among them, each indicator is normalized to the range of [0,1], and each weight can be configured according to application requirements. The system uses the candidate role with the highest score as the allocation result. This calculation method is simple and efficient and is suitable for intelligent agent scheduling scenarios with limited resources or high real-time decision requirements.

[0090] S34: Repeat steps S32-S33 until the second score between the i-th task and all corresponding agents is calculated. The agent with the largest second score is taken as the executing agent of the i-th task, and the i-th task is assigned to the executing agent.

[0091] S35: Repeat steps S31-S34 until all tasks are traversed, obtain all executing agents, and set all executing agents as a set of executing agents.

[0092] Obtain the execution sequence of the task set, and execute the tasks according to the execution sequence of the execution agents in the execution agent set.

[0093] In some embodiments, obtaining the execution sequence of the task set, and having the execution agents in the execution agent set execute tasks according to the execution sequence, specifically includes:

[0094] Obtain the urgency index and importance index of each task in the task set, calculate the priority index based on the urgency index and importance index, sort the priority indices of each task from largest to smallest to construct an execution sequence, and select the corresponding execution agents to execute the tasks in the order of priority indices in the execution sequence.

[0095] In some embodiments, an execution sequence refers to queuing tasks to be executed according to their priority indices so that the system or agent can execute tasks in order of highest priority. For example, in dynamic scenarios such as smart agriculture, the urgency index, importance index, scope of impact, allocation probability, and failure risk of each task are obtained. Based on these indices, a priority index is calculated, and tasks with higher priority indices are executed first.

[0096] In some embodiments, after the execution agents in the set of execution agents execute tasks according to the execution sequence, the set of execution agents further includes:

[0097] Obtain the task completion results of each executing agent, set adjustment values ​​based on the task completion results. If the task completion result is successful, set the adjustment value to the first preset value; if the task completion result is unsuccessful, set the adjustment value to the second preset value. Update the agent skill vector of the executing agent based on the adjustment values. Calculate the task success rate based on the task completion results, and update the historical trust level of the executing agent based on the task success rate.

[0098] In some embodiments, each skill parameter in the agent's skill vector has a capability estimation parameter. The capability estimation parameter is updated based on the actual feedback from task execution, making it increasingly closer to the agent's true skill performance, thereby optimizing matching and task success rate. The update formula for the capability estimation parameter is as follows:

[0099]

[0100] in, Let u be the skill parameter of the j-th agent. This represents the learning rate, used to control the adjustment range; if the task completion result is successful, the adjustment value is set to 1, and if the task completion result is unsuccessful, the adjustment value is set to 0. This refers to the u-th skill parameter of the j-th agent after the update.

[0101] Historical trust score is updated based on task success rate. For example, if 9 out of the last 10 tasks are successful, the historical trust score will be updated to 0.9.

[0102] refer to Figure 2 The present invention also provides a cooperative control device 20 for a multi-agent system, used to implement the cooperative control method for the multi-agent system, the device comprising:

[0103] The first mapping relationship set acquisition module 21 is used to acquire a task set and a role set. The task set includes multiple tasks, and the role set includes multiple roles. It acquires the attribute set of the task set, matches the tasks with the roles according to the attribute set, and obtains the first mapping relationship set.

[0104] The second mapping relationship set acquisition module 22 is used to acquire the agent state vector set of a multi-agent system, which includes multiple agents; and to match agents with roles based on the agent state vector set to obtain the second mapping relationship set.

[0105] The execution agent set acquisition module 23 is used to acquire the agent skill vector set of the multi-agent system, and assign tasks to agents according to the attribute set, agent state vector set, agent skill vector set, first mapping relationship set and second mapping relationship set to obtain the execution agent set;

[0106] The task execution module 24 is used to obtain the execution sequence of the task set, and the execution agents in the execution agent set execute the tasks according to the execution sequence.

[0107] This application provides an electronic device, including a processor and a memory; the memory stores a computer program, wherein the computer program, when executed by the processor, implements a cooperative control method for a multi-agent system according to any of the above-described schemes.

[0108] Specifically, the processor may include, for example, a general-purpose microprocessor, an instruction set processor and / or an associated chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor may also include onboard memory for caching purposes. The processor may be a single processing unit or multiple processing units for performing different actions of the method flow according to embodiments of this application.

[0109] Memory can be any medium capable of containing, storing, transmitting, propagating, or transmitting instructions. For example, memory can include, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, instruments, or propagation media. Specific examples of memory include: magnetic storage devices such as magnetic tape or hard disk drives (HDDs); optical storage devices such as optical discs (CD-ROMs); and also random access memory (RAM) or flash memory; and / or wired / wireless communication links.

[0110] This application also provides a computer-readable medium storing a computer program thereon, which, when executed by a processor, implements a cooperative control method for a multi-agent system according to any of the above-described schemes. This computer-readable medium may be included in the device / apparatus / system described in the above embodiments; or it may exist independently and not assembled into the device / apparatus / system. The aforementioned computer-readable medium carries one or more programs, which, when executed, implement the method as described in the embodiments of this application.

[0111] According to embodiments of this application, a computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this application, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media can also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wireless, wired, optical fiber, radio frequency signals, etc., or any suitable combination thereof.

[0112] Those skilled in the art will understand that the features described in the various embodiments and / or claims of this application can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in this application. In particular, the features described in the various embodiments and / or claims of this application can be combined and / or combined in various ways without departing from the spirit and teachings of this application. All such combinations and / or combinations fall within the scope of this application. Therefore, the scope of this application should not be limited to the above embodiments, but should be defined not only by the appended claims, but also by their equivalents. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A cooperative control method for a multi-agent system, characterized in that, Including the following steps: Get the task set and the role set. The task set includes multiple tasks, and the role set includes multiple roles. Get the attribute set of the task set. Match the tasks and roles according to the attribute set to obtain the first mapping relationship set. Obtain the set of agent state vectors for a multi-agent system, which includes multiple agents; match agents with roles based on the set of agent state vectors to obtain a second set of mapping relationships; Obtain the agent skill vector set of the multi-agent system, and assign tasks to agents according to the attribute set, agent state vector set, agent skill vector set, first mapping relationship set and second mapping relationship set to obtain the execution agent set; Obtain the execution sequence of the task set, and execute the tasks according to the execution sequence of the execution agents in the execution agent set; The process of assigning tasks to agents based on the attribute set, agent state vector set, agent skill vector set, first mapping relationship set, and second mapping relationship set to obtain an execution agent set specifically includes: S31: Select the i-th task in the task set, obtain the task skill requirement vector and task target location vector of the i-th task from the attribute set, obtain all the roles corresponding to the i-th task according to the first mapping relationship set, and obtain the intelligent agent corresponding to each role according to the second mapping relationship set; S32: Select the j-th agent corresponding to the i-th task, obtain the agent skill vector of the j-th agent from the agent skill vector set, obtain the agent state vector of the j-th agent from the agent state vector set, and obtain the agent position vector, current task number, maximum task capacity and historical trust degree of the j-th agent based on the agent state vector; S33: Calculate the second score between the i-th task and the corresponding j-th agent based on the task skill requirement vector, agent skill vector, task destination location vector, agent location vector, current number of tasks, maximum task capacity, and historical trust level TS. S34: Repeat steps S32-S33 until the second score between the i-th task and all corresponding agents is calculated. The agent with the largest second score is taken as the executing agent of the i-th task, and the i-th task is assigned to the executing agent. S35: Repeat steps S31-S34 until all tasks are traversed, obtain all executing agents, and set all executing agents as a set of executing agents.

2. The cooperative control method for a multi-agent system according to claim 1, characterized in that, The process of matching tasks with roles based on attribute sets to obtain a first mapping relationship set specifically includes: S11: Obtain the attributes of the i-th task from the attribute set. The attributes include: task skill requirement vector, task target location vector, time attribute vector, and task constraints. Select all roles that match the attributes from the role set. Construct a one-to-many mapping relationship between the i-th task and all roles that match the attributes. S12: Repeat step S11 until all task attributes are traversed, obtain all one-to-many mapping relationships between tasks and roles, and use all one-to-many mapping relationships between tasks and roles as the first mapping relationship set.

3. The cooperative control method for a multi-agent system according to claim 1, characterized in that, The step of matching agents with roles based on the agent state vector set to obtain a second mapping relationship set specifically includes: S21: Select the k-th character R k ; S22: Obtain the j-th agent A from the agent state vector set. j agent state vector s j According to the agent's state vector s j Agent A is obtained through deep evaluation function calculation. j With character R k The first rating between them; S23: Repeat step S22 to calculate and obtain R for all agents and roles. k The first score between them, the agent and role R corresponding to the highest first score. k Establish a one-to-one mapping relationship between them; S24: Repeat steps S21-S23 until all roles are traversed, obtain all one-to-one mapping relationships between agents and roles, and use all one-to-one mapping relationships between agents and roles as the second mapping relationship set.

4. The cooperative control method for a multi-agent system according to claim 1, characterized in that, The calculation process for the second score specifically includes: Skill similarity (SM) is calculated based on the task skill requirement vector and the agent skill vector; geographical location score (LM) is calculated based on the task destination location vector and the agent location vector; and resource load score (LS) is calculated based on the current number of tasks and the maximum task capacity. Set the first weight Second weight Third weight and the fourth weight ; Will This serves as a second rating between the task and the corresponding agent.

5. The cooperative control method for a multi-agent system according to claim 1, characterized in that, The process of obtaining the execution sequence of the task set, and the execution agents in the execution agent set executing the tasks according to the execution sequence, specifically includes: Obtain the urgency index and importance index of each task in the task set, calculate the priority index based on the urgency index and importance index, sort the priority indices of each task from largest to smallest to construct an execution sequence, and select the corresponding execution agents to execute the tasks in the order of priority indices in the execution sequence.

6. The cooperative control method for a multi-agent system according to claim 1, characterized in that, After the executing agents in the set of executing agents have executed the tasks according to the execution sequence, the set of executing agents also includes: Obtain the task completion results of each executing agent, set adjustment values ​​based on the task completion results. If the task completion result is successful, set the adjustment value to the first preset value; if the task completion result is unsuccessful, set the adjustment value to the second preset value. Update the agent skill vector of the executing agent based on the adjustment values. Calculate the task success rate based on the task completion results, and update the historical trust level of the executing agent based on the task success rate.

7. A cooperative control device for a multi-agent system, used to implement the cooperative control method for a multi-agent system as described in any one of claims 1 to 6, characterized in that, The device includes: The first mapping relationship set acquisition module is used to acquire a task set and a role set. The task set includes multiple tasks, and the role set includes multiple roles. It acquires the attribute set of the task set, matches the tasks with the roles based on the attribute set, and obtains the first mapping relationship set. The second mapping relationship set acquisition module is used to acquire the agent state vector set of a multi-agent system, which includes multiple agents; and to match agents with roles based on the agent state vector set to obtain the second mapping relationship set. The execution agent set acquisition module is used to acquire the agent skill vector set of the multi-agent system, and assign tasks to agents according to the attribute set, agent state vector set, agent skill vector set, first mapping relationship set and second mapping relationship set to obtain the execution agent set; The task execution module is used to obtain the execution sequence of the task set, and the execution agents in the execution agent set execute the tasks according to the execution sequence.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the cooperative control method for a multi-agent system as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the cooperative control method for a multi-agent system as described in any one of claims 1 to 6.

Images