Multi-agent dynamic task integrated allocation method and system based on labor division mechanism

Abstract

The invention discloses a multi-agent dynamic task integrated allocation method and system based on a labor division mechanism. The method comprises the following steps: constructing a multi-agent execution dynamic task allocation variable model by using a six-tuple; constructing a multi-agent dynamic task allocation index model; constructing a multi-agent labor division mechanism ant colony in combination with a labor division mechanism and a task allocation constraint set; determining an ant selection probability in combination with task capabilities of multiple agents, gradually selecting ants, and constructing a task allocation sequence of each agent according to a state transition rule; updating pheromones on each distribution sequence through state transition, calculating an evaluation value through an index model, if iteration is performed for the first time, taking the current evaluation value as an optimal evaluation value, otherwise, if the current evaluation value is greater than the optimal evaluation value, taking the current evaluation value as the optimal evaluation value; the ant colony is updated and iterated, the solution corresponding to the optimal evaluation value is the optimal solution, and the optimal task allocation sequence of the intelligent agent is obtained. According to the invention, the efficiency and performance of the multi-agent system are improved, and complex tasks and environments can be better coped with.

Description

Technical Field

[0001] The present invention relates to multi-agent task allocation technology, and in particular to a multi-agent dynamic task integrated allocation method and system based on a division of labor mechanism. Background Art

[0002] The multi-agent dynamic task allocation problem is a complex optimization problem, which refers to dynamically adjusting the task allocation strategy in a cluster according to the real-time changes in the system state and the dynamic nature of the task requirements to adapt to the changing environment and task conditions. This type of problem is widely present in practical applications, such as drone collaborative search, air-to-ground unmanned equipment collaborative rescue, and distributed sensor network task allocation. Different from static task allocation, dynamic task allocation needs to be based on the changes in the system state and target tasks, which places higher requirements on the running time and decision-making mechanism of the task allocation method.

[0003] With the development of multi-agent technology, the study of dynamic task allocation is of great significance for improving system efficiency, enhancing system robustness, optimizing resource utilization, coping with complex task environments, and promoting multi-agent collaboration research. Dynamic task allocation is one of the important areas of multi-agent collaboration research. By studying the problem of dynamic task allocation, agents can be allowed to adjust task allocation schemes in real time according to environmental changes and task requirements, thereby improving the overall efficiency and flexibility of the system. Resources can be used more effectively to avoid idle or reused resources, thereby improving resource utilization and overall system benefits. The task allocation strategy can be adjusted in real time according to different environmental changes, so that the system can more effectively cope with complex task requirements.

[0004] At present, there are many different methods for solving the problem of dynamic task allocation among multiple agents. For example, rule-based task allocation determines the task allocation strategy through predefined rules. Market-based allocation is similar to the trading mechanism in the market. Agents obtain tasks through bidding, auctioning, etc., and determine the task price and allocation plan according to the supply and demand relationship of resources and the competition situation. They make task selections based on local information and communicate with other agents to exchange information and coordinate actions. Task allocation based on game theory models the task allocation problem among multiple agents as a game theory model, and makes task allocation decisions through the concept of equilibrium in game theory. Task allocation based on heuristic algorithms designs corresponding strategies to adapt to task changes, such as genetic algorithms, simulated annealing algorithms, ant colony algorithms, etc.

[0005] However, in practical applications, rule-based and market-based task allocation is often used for distributed task allocation, which cannot solve large-scale allocation methods. Task allocation based on game theory has too high hardware requirements and cannot cope with complex problems. Moreover, most task allocations fail to solve the problem of collaborative task allocation among heterogeneous agents. Summary of the invention

[0006] The purpose of the present invention is to propose a multi-agent dynamic task integrated allocation method and system based on a division of labor mechanism, to solve the problem of dynamic task allocation of heterogeneous cluster agents in the background technology, and to quickly obtain a reasonable agent task allocation plan.

[0007] In order to achieve the above-mentioned purpose, the present invention adopts the following technical scheme:

[0008] A multi-agent dynamic task integrated allocation method based on a division of labor mechanism, comprising:

[0009] Step 1, using six-tuples to build a variable model for multi-agent dynamic task allocation;

[0010] Step 2: Construct a multi-agent dynamic task allocation indicator model;

[0011] Step 3: Combine the division of labor mechanism and the set of task allocation constraints to build a multi-agent division of labor ant colony;

[0012] Step 4: Determine the ant selection probability based on the task capabilities of multiple agents, select ants step by step, and construct the task allocation sequence of each agent according to the state transition rule;

[0013] Step 5, update the pheromone on each allocation sequence through state transfer, calculate the evaluation value through the indicator model, if it is the first iteration, take the current evaluation value as the optimal evaluation value, otherwise, if the current evaluation value is greater than the optimal evaluation value, take the current evaluation value as the optimal evaluation value;

[0014] Step 6: Repeat steps 4 and 5 to iterate the ant colony. The solution corresponding to the optimal evaluation value is the optimal solution, that is, the optimal task allocation sequence of the agent is obtained.

[0015] Furthermore, the multi-agent dynamic task allocation variable is<E,V,T,M,F,C> Six-tuple, where E is the planning scene environment, V = {V1, V2, ..., V n} is the set of multi-agents in the cluster, n is the total number of multi-agents; T = {T1, T2, ..., T i ,…,T m} is the set of targets included in the multi-agent task execution process, and m represents the total number of targets; is the task set, N M is the total number of tasks, M k Target T i The pth task of , F is the set of obstacle entities and no-go areas in the multi-agent task execution environment in the planning scenario, and C is the set of task allocation constraints.

[0016] Furthermore, the task allocation constraint set includes task timing constraints, task capability constraints, collaborative task constraints, execution time constraints and maximum range constraints.

[0017] Furthermore, the indicator model is:

[0018] Z=ω1Z1+ω2Z2

[0019] Among them, Z represents the sum of the evaluation values ​​of the multi-agent task allocation results, ω1 and ω2 are the weight coefficients of the task execution path length and task completion degree, Z1 is the total task reward index, and Z2 is the voyage efficiency index.

[0020] Furthermore, the total task reward index Z1 is:

[0021]

[0022] In the actual operation process, the total task reward is normalized, that is:

[0023]

[0024] Among them, z i is the task reward weight of the ith agent, e i is the task completion degree of the i-th task.

[0025] Furthermore, the range performance index Z2 is:

[0026]

[0027] Among them, R i For the agent V i The estimated distance to complete all its tasks, For the agent V i Maximum range constraint.

[0028] Furthermore, the multi-agent division of labor mechanism ant colony includes ant sub-colonies and ant clusters, specifically including:

[0029] Suppose the set of ants that can perform the same task is ant subgroup AC = {AC1, AC2, ..., AC i ,…,AC n}, n is the total number of agents in the cluster, each agent V i The corresponding ant subgroup Among them, Ant (i,j) is the i-th subgroup AC i The corresponding j-th ant, m i The number of ants corresponding to the ant sub-group corresponding to the agent;

[0030] Each Ant (i,j) Starting from the initial state of the agent, the task execution sequence corresponding to each agent in the ant subgroup AC is generated by traversal. During the traversal process, the m ants that constitute the required agent set to meet the goal of the task allocation constraint set are called the ant cluster AG. j , where the minimum ant cluster is mapped to the set of task sequences of the agent set V.

[0031] Furthermore, the ant selection probability is:

[0032]

[0033] in, Ant Cluster AG j Ant (i,j) The probability of being selected when the kth state transition is performed during the cth iteration, Ant (i,j) mission capability assessment value;

[0034] Ant (i,j) Mission capability assessment value for:

[0035]

[0036]

[0037] in, Ant (i,j) Corresponding Agent V i The remaining range weight coefficient, It represents the maximum value of pheromone concentration during the transition from state a to state s, which is the prior knowledge and experience in the ant colony algorithm search process. Ant (i,j) Corresponding Agent V i Corresponding mission capability; MaxR(V i ) represents the agent V i The remaining distance in the current state; curR(V i ) represents the agent V i The distance consumed in executing the current mission; Represents the total number of tasks in the set, n all Represents the total tasks in dynamic task allocation; MaxM(V i ) represents the agent V i The total task capability value in the current state, curM(V i ) represents the agent V i The task capacity value consumed by executing the current task represents the total task in the dynamic task allocation.

[0038] Furthermore, updating the pheromone on each allocation sequence through state transfer specifically includes:

[0039] Assume that in state c, task M u Move to next task M v ,v∈allow (i,j) (u) The state transition amount is s, and the transition amount s is determined as:

[0040]

[0041] Among them, τ (u,v) (c) indicates that in state c, task M u Move to next task M v The amount of pheromone on the path is the average of the total amount of pheromone left by the ants before; α and β are the importance of pheromone and heuristic factor respectively; For the current task M u Move to next task M v The inspiration factor;

[0042] After the state transfer is performed based on the transfer amount s, the pheromones on each path of the ant colony in the t+1 state are updated as follows:

[0043]

[0044] Among them, ρ local ∈[0,1] represents the local pheromone volatility coefficient, Indicates pheromone concentration.

[0045] A multi-agent dynamic task integrated allocation system based on division of labor mechanism, comprising:

[0046] Input unit, used to input six-tuples to construct multi-agent execution dynamic task allocation variables and ant colony algorithm iteration variables

[0047] A multi-agent dynamic task allocation indicator model construction unit, used to construct a multi-agent dynamic task allocation indicator model;

[0048] The multi-agent division of labor mechanism ant colony construction unit combines the division of labor mechanism and the task allocation constraint set to construct a multi-agent division of labor mechanism ant colony;

[0049] The iterative solution unit determines the ant selection probability in combination with the task capabilities of multiple agents, selects ants step by step, and constructs the task allocation sequence of each agent according to the state transition rule; updates the pheromone on each allocation sequence through state transition, and calculates the evaluation value through the indicator model. If it is the first iteration, the current evaluation value is used as the optimal evaluation value. Otherwise, if the current evaluation value is greater than the optimal evaluation value, the current evaluation value is used as the optimal evaluation value, and the ant colony is updated and iterated. The solution corresponding to the optimal evaluation value is the optimal solution, that is, the optimal task allocation sequence of the agent is obtained.

[0050] Compared with the prior art, the present invention has the following beneficial effects:

[0051] 1. The present invention starts from the perspective of multi-agent cluster collaborative planning, comprehensively considers the real-time position of each agent, the nature of the load and the target, the timing, etc. to carry out an integrated task allocation process, and can support collaborative planning of multiple agents with different loads;

[0052] 2. The present invention combines the division of labor mechanism to allocate dynamic tasks, and can timely and reasonably combine the characteristics of different agents to generate a multi-agent cluster task allocation plan;

[0053] 3. The present invention is intelligent and parallel. During the calculation process, it can perform parallel calculations according to computer performance, greatly improving the planning speed, effectively improving the optimization ability of the algorithm, obtaining a better task allocation sequence within a limited number of iterations, and improving the efficiency and performance of multi-agent collaborative applications. It can better cope with larger-scale scenarios such as large numbers of clusters and complex planning environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 This is a flowchart of the overall framework of the multi-agent dynamic task integration allocation method based on the division of labor mechanism.

[0055] Figure 2 This is a mapping diagram of the ant colony structure between intelligent agents and division of labor mechanism. DETAILED DESCRIPTION

[0056] This embodiment provides a method for integrated dynamic task allocation of multiple agents based on a division of labor mechanism. This method starts from the perspective of multi-agent cluster collaborative control, comprehensively considers the position of each agent, the nature of the load and target, the timing, etc. to perform integrated task allocation, and has the characteristics of intelligence and parallelism. Combined with the division of labor mechanism, it can effectively improve the optimization ability of the algorithm, obtain a better task allocation sequence within a limited number of iterations, improve the efficiency and performance of the multi-agent system, and better cope with complex tasks and environments. Figure 1 The method specifically comprises the following steps:

[0057] Step 1: Use sextuples to define variables for dynamic task allocation for multi-agent execution.

[0058] Multi-agent task allocation variables<E,V,T,M,F,C> Six-tuple is used for unified quantification, where E is the planning scene environment, V = {V1, V2, …, V n} is the set of multi-agents in the cluster, n is the total number of multi-agents; T = {T1, T2, ..., T i ,…,T m} is the set of targets included in the multi-agent task execution process, and m represents the total number of targets; is the task set, N M is the total number of tasks, M k Target T i The pth task of The task type needs to be determined in combination with the task scenario, including but not limited to: exploration, execution, standby and evaluation. The timing relationship contained in the task type is defined in the constraint model in step 2. Type i Indicates the target T i The set of tasks to be executed, so task M k It can be expressed as:

[0059]

[0060] F is the set of obstacle entities and no-go areas in the multi-agent task execution environment in the planning scenario, and C is the set of constraints for task allocation.

[0061] Step 2: Construct a multi-agent dynamic task allocation constraint model and indicator model.

[0062] The constraint model includes task timing constraints, task capability constraints, collaborative task constraints, execution time constraints, maximum range constraints, etc. Specifically, it includes:

[0063] Task timing constraints: There is a mutual order relationship between the detection, execution and evaluation of the same target. For example, the execution of the execution task must be based on the confirmation of the detection result, and the evaluation task can only be executed after the execution task. i and M j There is a specific sequence between them, which is called task timing constraint.

[0064] Task capability constraints: In the process of dynamic task allocation, we first need to consider the different flight dynamics models between different agents, such as minimum turning radius, average speed, range, etc. Therefore, in task allocation, the individual capabilities of different agents are limited and they can only perform tasks within their own task range. Tasks beyond their own capabilities must be performed by other agents.i The types of tasks that can be executed are Capacity (V i ), then V i The tasks performed M k The task types in need to meet

[0065]

[0066] Collaborative task constraints: When searching or executing a task for a target or area, it is necessary to specify a combination of multiple agents to complete the task collaboratively, using V need = {V1, V2, …, V j}express.

[0067] Task execution time constraints: Some special tasks often require task execution time constraints. For example, to ensure accurate positioning of a dynamic target, the scheduled detection task time must not be less than a certain time threshold.

[0068] Maximum range constraint: The estimated range of the agent to complete all its tasks must not exceed the agent’s maximum range.

[0069] Index model: In the process of dynamic task allocation, the indicators for evaluating system performance are measured by two types of indicators: the total path length of tasks completed by multiple agents and the total number of tasks completed. The calculation method is as follows:

[0070] Z=ω1Z1+ω2Z2

[0071] Among them, Z represents the sum of the evaluation values ​​of the multi-agent task allocation results. ω1 and ω2 are the weight coefficients of the task execution path length and task completion, respectively. Different task execution effects can be obtained by adjusting different weight coefficients. Z1 is the total task reward index, and Z2 is the voyage efficiency index. The value of the covering task reward in the cluster task allocation process directly reflects the completion of the task. Z1 is the total task reward index of the multi-agent:

[0072]

[0073] z i is the task reward weight of the ith agent (i.e., the importance of the ith task), e i is the task completion degree of the ith task (related to the number of agents executing, e.g., if three agents are required and two have executed, then it is 2 / 3).

[0074] In the actual operation process, the total task reward is normalized, that is:

[0075]

[0076] Assume R i For the agent Vi The estimated distance to complete all its tasks, For the agent V i The maximum range constraint, the normalized range performance index Z2 is:

[0077]

[0078] Step 3: Combine the labor division mechanism and task allocation constraint model to construct a multi-agent labor division mechanism ant colony, including the generation method of ant sub-groups and ant clusters:

[0079] In the dynamic task allocation solution, since the tasks and order undertaken by each agent are not exactly the same, each agent is mapped to an independent ant subgroup, each subgroup corresponds to its task allocation execution plan, and they cooperate with each other based on the information of pheromone distribution.

[0080] Suppose the set of ants that can perform the same task is ant subgroup AC = {AC1, AC2, ..., AC i ,…,AC n}, n is the total number of agents in the cluster, each agent V i The corresponding ant subgroup Among them, Ant (i,j) is the i-th subgroup AC i The corresponding j-th ant, m i is the number of ants corresponding to the ant sub-colony corresponding to the agent.

[0081] Each Ant (i,j) Starting from the initial state of the agent, the exhaustive method or greedy algorithm is used to traverse and generate the task execution sequence corresponding to each agent in the ant subgroup AC. During the traversal process, the task constraint model needs to be checked and the task execution sequence that does not meet the task constraint is pruned. The m ants that meet various constraints and constitute the set of agents required for the target are called ant cluster AG. j ,j=1,2,…,m. The mapping diagram of the intelligent agent and the division of labor mechanism ant colony structure is as follows Figure 2 As shown, AG j The minimum ant cluster in the ant colony that constructs a complete task allocation plan for a multi-agent set. Therefore, the task sequence set of the multi-agent set V is mapped to the ant cluster AG j The set of feasible solutions constructed.

[0082] Step 4: Combine the task capability characteristics of multiple agents, construct the ant colony task capability calculation formula, and generate the problem solution.

[0083] In the process of solving the multi-agent cluster task allocation, the basic element of the multi-agent set V constructed by the ant colony algorithm based on the division of labor mechanism is the ant cluster, that is, each ant cluster corresponds to a feasible solution that satisfies the task constraint model in the problem solving process;

[0084] Ant Cluster AG j In the process of constructing the solution, one ant is selected each time. (i,j) The state is transferred to determine the subsequent task. In the process of selecting ants, the task capability of the agent is mapped to the ant colony, and the selection is made according to the individual capability of the ants. The probability can be expressed as:

[0085]

[0086] in, Ant Cluster AG j Ant (i,j) The probability of being selected when performing the kth state transition during the cth iteration. Ant (i,j) mission capability assessment value; Ant Cluster AG j The sum of the task capabilities of all ants in the . The calculation method is:

[0087]

[0088] in, Ant (i,j) Corresponding Agent V i The remaining range weight coefficient. It represents the maximum value of pheromone concentration during the transition from state a to state s, which is the prior knowledge and experience in the ant colony algorithm search process. Ant (i,j) Corresponding Agent V i Corresponding mission capabilities.

[0089] The calculation method of the range weight coefficient is:

[0090]

[0091] MaxR(V i ) represents the agent V i The remaining distance in the current state; curR(V i ) represents the agent V i The flight distance consumed in executing the current mission.

[0092] The task capability calculation method is:

[0093]

[0094] Represents the total number of tasks in the set, n all Represents the total number of tasks in dynamic task allocation. MaxM(V i ) represents the agent V i The total task capability value in the current state, curM(V i ) represents the agent V i The task capability value consumed to execute the current task.

[0095] Step 5: Update and iterate the task execution sequence through state transfer calculation and indicator model calculation to obtain the optimal task allocation sequence of the agent.

[0096] Step 5.1 updates the local pheromone through state transfer calculation.

[0097] Assume that in state c, task M u Move to next task M v ,v∈allow (i,j) (u) The state transition amount is s. In the initial state, the transition amount of s is randomly generated based on the probability distribution. When there is prior knowledge, the calculation method of the transition amount s is:

[0098]

[0099] Among them, τ (u,v) (v) indicates that in state c, task M u Move to next task M v The amount of pheromone on the path is the average of the total amount of pheromone left by the ants before; α and β are the importance of pheromone and heuristic factor respectively. For the current task M u Move to next task M v This factor represents the inspiration factor of the completed M u Then execute M v The difficulty of the mission can be calculated based on the sequence of tasks or the distance between two tasks.

[0100] After the state transfer, the method of updating the pheromone on each path is the same as the traditional ant colony algorithm pheromone update method. j=1,2,…,m, at the tth iteration:

[0101]

[0102] Among them, ρ local ∈[0,1] represents the local pheromone volatilization coefficient, 1-ρ is the pheromone residual coefficient, Indicates pheromone concentration.

[0103] Step 5.2: Calculate the total path length of the current ant colony and the total number of completed tasks, obtain the solution evaluation value of the current iteration number according to the indicator model formula, and update the optimal solution. After the iteration is completed, the optimal ant colony solution is the optimal task allocation sequence of the agent. The optimal solution is recorded in each iteration. After reaching the maximum number of iterations, the system outputs the optimal task allocation sequence of the agent.

[0104] A multi-agent dynamic task integrated allocation system, comprising:

[0105] Input unit, used to input six-tuples to construct multi-agent execution dynamic task allocation variables and ant colony algorithm iteration variables

[0106] A multi-agent dynamic task allocation indicator model construction unit, used to construct a multi-agent dynamic task allocation indicator model;

[0107] The multi-agent division of labor mechanism ant colony construction unit combines the division of labor mechanism and the task allocation constraint set to construct a multi-agent division of labor mechanism ant colony;

[0108] The iterative solution unit determines the ant selection probability in combination with the task capabilities of multiple agents, selects ants step by step, and constructs the task allocation sequence of each agent according to the state transition rule; updates the pheromone on each allocation sequence through state transition, and calculates the evaluation value through the indicator model. If it is the first iteration, the current evaluation value is used as the optimal evaluation value. Otherwise, if the current evaluation value is greater than the optimal evaluation value, the current evaluation value is used as the optimal evaluation value, and the ant colony is updated and iterated. The solution corresponding to the optimal evaluation value is the optimal solution, that is, the optimal task allocation sequence of the agent is obtained.

[0109] The code execution process based on the system is:

[0110] Input: multi-agent set, target set, task set, state transition parameter, number of iterations, importance of heuristic factor, pheromone volatility coefficient.

[0111] Output: Task execution sequence of multi-agent set

[0112] Step 1: Initialize the size of the ant colony, establish ant sub-colonies and ant clusters, and initialize the iteration counter.

[0113] Step 2: Perform the following operations on the ant cluster:

[0114] Step 2.1: Select ants step by step according to the ant selection probability; and construct the task assignment sequence of each agent according to the state transition rule.

[0115] Step 2.2: After generating each task assignment sequence, update the local pheromone on each sequence;

[0116] Step 2.3: Calculate the evaluation value of the solution according to the indicator model and update the current optimal solution.

[0117] Step 3: When the counter reaches the maximum number of iterations, output the optimal solution, otherwise continue the algorithm calculation.

[0118] Step 4: The system outputs the optimal task assignment sequence for the agent.

[0119] In summary, this solution can cope with and handle the problem of dynamic task allocation in heterogeneous multi-agents, obtain an efficient multi-agent dynamic task allocation solution, obtain a better task allocation sequence within a limited number of iterations, improve the efficiency and performance of multi-agent collaborative applications, and better cope with larger-scale scenarios such as large number of clusters and complex planning environments.

[0120] The above description is only a preferred embodiment of the present embodiment and is not intended to limit the present embodiment. For those skilled in the art, the present embodiment may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present embodiment shall be included in the protection scope of the present embodiment.

Claims

1. A multi-agent dynamic task integration allocation method based on a division of labor mechanism, characterized in that: include: Step 1, using six-tuples to build a variable model for multi-agent dynamic task allocation; Step 2: Construct a multi-agent dynamic task allocation indicator model; Step 3: Combine the division of labor mechanism and the set of task allocation constraints to build a multi-agent division of labor ant colony; Step 4: Determine the ant selection probability based on the task capabilities of multiple agents, select ants step by step, and construct the task allocation sequence of each agent according to the state transition rule; Step 5, update the pheromone on each allocation sequence through state transfer, calculate the evaluation value through the indicator model, if it is the first iteration, take the current evaluation value as the optimal evaluation value, otherwise, if the current evaluation value is greater than the optimal evaluation value, take the current evaluation value as the optimal evaluation value; Step 6: Repeat steps 4 and 5 to iterate the ant colony. The solution corresponding to the optimal evaluation value is the optimal solution, that is, the optimal task allocation sequence of the agent is obtained.

2. According to the multi-agent dynamic task integration allocation method based on the division of labor mechanism according to claim 1, it is characterized in that: The multi-agent dynamic task allocation variable is<E,V,T,M,F,C> Six-tuple, where E is the planning scene environment, V = {V1, V2, ..., V n } is the set of multi-agents in the cluster, n is the total number of multi-agents; T = {T1, T2, ..., T i ,…,T m } is the set of targets included in the multi-agent task execution process, and m represents the total number of targets; is the task set, N M is the total number of tasks, M k Target T i The pth task of , F is the set of obstacle entities and no-go areas in the multi-agent task execution environment in the planning scenario, and C is the set of task allocation constraints.

3. According to the multi-agent dynamic task integration allocation method based on the division of labor mechanism of claim 2, it is characterized in that: The task allocation constraint set includes task timing constraint, task capability constraint, collaborative task constraint, execution time constraint and maximum range constraint.

4. According to the multi-agent dynamic task integration allocation method based on the division of labor mechanism according to claim 2, it is characterized in that: The indicator model is: Z=ω1Z1+ω2Z2 Among them, Z represents the sum of the evaluation values ​​of the multi-agent task allocation results, ω1 and ω2 are the weight coefficients of the task execution path length and task completion degree, Z1 is the total task reward index, and Z2 is the voyage efficiency index.

5. According to the multi-agent dynamic task integrated allocation method based on the division of labor mechanism of claim 4, it is characterized in that: The total task reward index Z1 is: In the actual operation process, the total task reward is normalized, that is: Among them, z i is the task reward weight of the ith agent, e i is the task completion degree of the i-th task.

6. According to the multi-agent dynamic task integrated allocation method based on the division of labor mechanism of claim 4, it is characterized in that: The range performance index Z2 is: Among them, R i For the agent V i The estimated distance to complete all its tasks, For the agent V i Maximum range constraint.

7. According to the multi-agent dynamic task integrated allocation method based on the division of labor mechanism as described in claim 1, it is characterized in that: The multi-agent division of labor mechanism ant colony includes ant sub-colonies and ant clusters, specifically including: Suppose the set of ants that can perform the same task is ant subgroup AC = {AC1, AC2, ..., AC i ,…,AC n }, n is the total number of agents in the cluster, and each agent V i The corresponding ant subgroup Among them, Ant (i,j) is the i-th subgroup AC i The corresponding j-th ant, m i The number of ants corresponding to the ant sub-group corresponding to the agent; Each Ant (i,j) Starting from the initial state of the agent, the task execution sequence corresponding to each agent in the ant subgroup AC is generated by traversal. During the traversal process, the m ants that constitute the required agent set to meet the goal of the task allocation constraint set are called the ant cluster AG. j , where the minimum ant cluster is mapped to the set of task sequences of the agent set V.

8. The method for multi-agent dynamic task integration allocation based on the division of labor mechanism according to claim 1 is characterized in that: The ant selection probability is: in, Ant Cluster AG j Ant (i,j) The probability of being selected when the kth state transition is performed during the cth iteration, Ant (i,j) mission capability assessment value; Ant (i,j) Mission capability assessment value for: in, Ant (i,j) Corresponding Agent V i The remaining range weight coefficient, It represents the maximum value of pheromone concentration during the transition from state a to state s, which is the prior knowledge and experience in the ant colony algorithm search process. Ant (i,j) Corresponding Agent V i Corresponding mission capability; MaxR(V i ) represents the agent V i The remaining distance in the current state; curR(V i ) represents the agent V i The distance consumed in executing the current mission; Represents the total number of tasks in the set, n all Represents the total tasks in dynamic task allocation; MaxM(V i ) represents the agent V i The total task capability value in the current state, curM(V i ) represents the agent V i The task capacity value consumed by executing the current task represents the total task in the dynamic task allocation.

9. The method for allocating multi-agent dynamic tasks based on a division of labor mechanism according to claim 1 is characterized in that: Updating the pheromone on each allocation sequence through state transfer specifically includes: Assume that in state c, task M u Move to next task M v ,v∈allow (i,j) (u) The state transition amount is s, and the transition amount s is determined as: Among them, τ (u,v) (c) indicates that in state c, task M u Move to next task M v The amount of pheromone on the path is the average of the total amount of pheromone left by the ants before; β and β are the importance of pheromone and heuristic factor respectively; For the current task M u Move to next task M v The inspiration factor; After the state transfer is performed based on the transfer amount s, the pheromones on each path of the ant colony in the t+1 state are updated as follows: Among them, ρ local ∈[0,1] represents the local pheromone volatility coefficient, Indicates pheromone concentration.

10. A multi-agent dynamic task integrated allocation system implementing any of the methods described in claims 1-9, characterized in that: include: Input unit, used to input six-tuples to construct multi-agent execution dynamic task allocation variables and ant colony algorithm iteration variables A multi-agent dynamic task allocation indicator model construction unit, used to construct a multi-agent dynamic task allocation indicator model; The multi-agent division of labor mechanism ant colony construction unit combines the division of labor mechanism and the task allocation constraint set to construct a multi-agent division of labor mechanism ant colony; The iterative solution unit determines the ant selection probability based on the task capabilities of multiple agents, selects ants step by step, and constructs the task allocation sequence of each agent according to the state transition rules; The pheromones on each allocation sequence are updated through state transfer, and the evaluation value is calculated through the indicator model. If it is the first iteration, the current evaluation value is taken as the optimal evaluation value. Otherwise, if the current evaluation value is greater than the optimal evaluation value, the current evaluation value is taken as the optimal evaluation value, and the ant colony is updated and iterated. The solution corresponding to the optimal evaluation value is the optimal solution, that is, the optimal task allocation sequence of the intelligent agent is obtained.

Images