Improved contract net-based multi-agent collaborative action task net reconstruction method

Abstract

The application discloses a kind of multi-agent collaborative action task network reconstruction methods based on improved contract network, comprising: if agent encounters sudden situation in task execution process, then the collaborative action task network is restructured using improved same network algorithm: the type of sudden situation of task scene is judged, and the number of tasks in the task sequence of agent is combined to determine whether to carry out single-task auction based on two contract types or multi-task auction based on concurrent transaction mechanism;The single-task auction based on two contract types is determined by the task sequence condition of bidder, including two cases of delivering sale contract tender and delivering exchange contract tender;The final bidder is determined based on time cost;In the multi-task auction based on concurrent transaction mechanism, the bidder participates in bidding by making different tenders by adding the multi-task of tenderer to its task sequence, and the tenderer matches and compares all tenders to determine the final bidder.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of multi-agent cooperation, and particularly relates to a multi-agent cooperative action task network reconstruction method based on an improved contract net. BACKGROUND

[0002] The contract net algorithm is a kind of distributed negotiation mechanism, which essentially simulates the mechanism of bidding-tender-bid in market behavior to solve action scheme decision problems. Multi-agent control has become the focus of distributed artificial intelligence. In the process of executing cooperative tasks, the task target state and the agent state are not constant. With the deepening of the enemy and us, the scene environment and action information will gradually become clear. Changes in the scene situation will cause the original pre-allocated scheme to be less feasible and less efficient.

[0003] In the process of executing action tasks, the scene situation may change. At this time, the current action scheme needs to be adjusted in real time according to the online evaluation of the change in the scene situation, so as to realize the reconstruction of the cooperative action task network and ensure the optimal overall task efficiency. The traditional pre-generation method of the cooperative action task network can obtain the optimal allocation scheme from different optimization aspects, but the algorithm complexity is high and the calculation time is long, which cannot meet the requirement of rapid response in the scene and is only suitable for task pre-allocation. SUMMARY

[0004] The purpose of the present application is to provide a multi-agent cooperative action task network reconstruction method based on an improved contract net to overcome the problems existing in the prior art.

[0005] In order to achieve the above-mentioned task, the following technical solutions are adopted in the present application:

[0006] The multi-agent cooperative action task network reconstruction method based on the improved contract net comprises:

[0007] First, the multi-agent is allocated tasks, thereby forming an initial cooperative action task network. If no sudden situation occurs in the task execution process of each agent, the allocated tasks are executed. If a sudden situation occurs in the task execution process of the agent, the improved contract net algorithm is used to reconstruct the cooperative action task network and determine a new task allocation scheme for each agent.

[0008] The improved contract net algorithm comprises:

[0009] The type of the sudden situation of the task scene is judged, the number of tasks in the task sequence of the agent is combined to determine whether to perform single-task auction based on two contract types or multi-task auction based on a concurrent transaction mechanism, and it is determined which agent acts as the tenderer.

[0010] The single-task auction based on two contract types is determined by the task sequence status of the bidder, including delivering a trading contract bid to incorporate the tenderer's task into the bidder's task sequence, delivering an exchange contract bid to exchange the tenderer's task and a certain task in the bidder's task sequence; the tenderer determines the final winning bidder based on time cost;

[0011] In the multi-task auction based on the concurrent transaction mechanism, the bidder participates in bidding by making different bids by adding the tenderer's multi-task to its own task sequence, and the tenderer matches and compares all the bids to determine the final winning bidder.

[0012] Further, the type of the sudden situation of the task scenario is judged, and the single-task auction based on two contract types or the multi-task auction based on the concurrent transaction mechanism is determined according to the number of tasks in the task sequence of the agent, including:

[0013] If the sudden situation of the task scenario is the emergence of a new threat, agent failure or damage, and there is only one task in its task sequence, the single-task auction based on two contract types is performed; if the sudden situation is the discovery of a new target, the agent is disabled, and there are multiple tasks in its task sequence, the multi-task auction based on the concurrent transaction mechanism is performed.

[0014] Further, it is determined which agent will act as the tenderer, including:

[0015] Among them, if the agent encountering the sudden situation has the tendering ability, it will act as the tenderer; if it does not have the tendering ability, the agent closest to it and having the tendering ability will act as the tenderer to bid for the to-be-executed task of the agent encountering the sudden situation; if two or more agents discover the same sudden situation at the same time, the closest one will be the tenderer.

[0016] Further, the single-task auction based on two contract types specifically includes:

[0017] The tenderer U i puts the task T h to be auctioned for sale; i publishes it as tender information;

[0018] After receiving the tender information, each agent determines whether it can participate in bidding as a bidder according to its own initial state: when the agent U j has not completed its current task sequence and has the ability to execute the task , it will execute a) as a bidder; when the agent U j has completed its current task sequence and has the ability to execute the task , it will execute b) as a bidder;

[0019] a) bidder U j will tender the bidding U i auction task into its own task sequence to complete non-deadlock sorting; calculate the time consumption of task execution tender the buy-sell contract bid

[0020] b) bidder U j will tender the bidding U i auction task into its own task sequence to complete non-deadlock sorting; through calculation comparison, if the task in the task sequence S j of U j can make the total time consumption of the entire task sequence minimum after exchange with the auction task , U j will tender the exchange contract bid to the bidding U i

[0021] calculate the bidding result: use the formula to calculate the bid price, n represents the number of bidders;

[0022] bidding U i will evaluate and compare all the buy-sell contract bids and exchange contract bids received within the preset time, finally select the bid with the lowest bid price among all the bidders in this round, that is, the bidder with the smallest action task time wins the bid and sends the bid information, and sends the bid information to other bidders.

[0023] Further, the initialization state includes the current position, task sequence, task execution ability, and remaining flight time.

[0024] Further, if the bid with the lowest bid price is an exchange contract bid, the task T j exchanged by the bidder U l of the exchange contract bid j will be included in the bidding task of the bidding U i , and the bidding U i will initiate the next bidding for this task and wait for other bidders to bid.

[0025] Further, the multi-task auction based on the concurrent transaction mechanism specifically includes:

[0026] bidding U i will publish the time sequence multi-task as bidding information, wherein need to be in​​​ Execute before;

[0027] After receiving the bidding information, each intelligent agent determines whether the task... In agent U j Within its capabilities, then the intelligent agent U j As a bidder, Add to its own task sequence S j middle;

[0028] Assume bidder U j Task sequence S j There is m j One task, S can be added j m j +1 different positions; if all these new task sequences can be executed, then m can be obtained. j +1 more options, at this point U j against propose m j +1 type of tender document l represents Insert into S j The l-th position;

[0029] Tenderer U i All received tenders will be processed within the preset time. To perform bid matching, if there is a matching set of bids: Bidder U j1 against Tender documents were submitted Bidder U j2 against Tender documents were submitted Evaluate the matching tender and calculate the action time for its corresponding action plan;

[0030] The tenderer selects the matching bids with the shortest action time from all matching bids, chooses the bidder who submitted the bid as the successful bidder, sends the winning bid information, and sends the unsuccessful bid information to the other bidders.

[0031] Furthermore, the action time refers to the time of action of bidder U. j1 Execute the task first Afterwards, bidder U j2 Execute the task again Total time consumed.

[0032] Furthermore, if no matching tender can be obtained, meaning no agent has a match... If a certain task is bid on, the current intelligent agent system will abandon handling the unexpected situation.

[0033] Further, after receiving the winning bid message, the bidder sends confirmation information to the tenderer, both parties sign a contract and update their respective task states, and the winning bidder updates the task sequence and continues to perform the task according to the new task sequence.

[0034] A terminal device comprises a processor, a memory and a computer program stored in the memory; when the processor executes the computer program, the improved contract net-based multi-agent collaborative action task net reconstruction method is realized.

[0035] A computer readable storage medium, the medium stores a computer program; when the computer program is executed by the processor, the improved contract net-based multi-agent collaborative action task net reconstruction method is realized.

[0036] Compared with the prior art, the present application has the following technical features:

[0037] In the cases of the failure of the agent of our side and the discovery of the new target of the opponent in the collaborative task, the method can quickly and effectively reconstruct the dynamic collaborative action task net, ensuring the normal progress of the collaborative task. In the case of discovering the new target of the opponent, the negotiation times and the algorithm time of the improved contract net are significantly lower than those of the traditional contract net algorithm, which can better adapt to the collaborative action task reconstruction under the scene of sudden situation. The method also provides valuable reference for the collaborative task reconstruction method in other complex environments, and has strong robustness and fault tolerance. BRIEF DESCRIPTION OF DRAWINGS

[0038] Figure 1 The flowchart of the method of the present application is shown in the figure;

[0039] Figure 2 The fault diagram of the agent 3 in case one of the embodiment of the present application is shown in the figure;

[0040] Figure 3 The schematic diagram of the new target unit of the opponent in case two of the embodiment of the present application is shown in the figure;

[0041] Figure 4 The corresponding result after redistribution in case one of the embodiment of the present application is shown in the figure;

[0042] Figure 5 The corresponding result after redistribution in case two of the embodiment of the present application is shown in the figure;

[0043] Figure 6 The comparison of the negotiation times of the new target between the present method and the traditional method in the embodiment of the present application is shown in the figure;

[0044] Figure 7 The comparison of the algorithm time of the new target between the present method and the traditional method in the embodiment of the present application is shown in the figure;

[0045] Figure 8 This is a comparison chart of the overall performance of the allocation schemes in the embodiments of the present invention. Detailed Implementation

[0046] This invention provides a method for reconstructing a multi-agent cooperative action task network based on an improved contract network. First, tasks are assigned to the multiple agents to form an initial cooperative action task network. If no unforeseen circumstances occur during task execution, the agents execute their assigned tasks. If unforeseen circumstances arise during task execution, the cooperative action task network is reconstructed using an improved contract network algorithm to establish a new task allocation scheme for each agent. These unforeseen circumstances include the discovery of new threats or targets, agent malfunctions, damage, or incapacitation during task execution.

[0047] 1. Contract Network Algorithm.

[0048] The bidding methods in the contract network algorithm mainly include sales contracts and exchange contracts:

[0049] (1) Sales Contract

[0050] A sales contract can be described as a quadruple. U i It is the tenderer, U j T is the bidder. i,j It's U i Migrate to U j The task This refers to the change in the overall system performance after migration; performance in this solution is evaluated using task completion time; in this solution, both the tenderer and the bidder are intelligent agents, and the system refers to a multi-agent system composed of all intelligent agents.

[0051] (2) Exchange Contract

[0052] An exchange contract can be described as a quintuple. Where T j,i It's U j Used with U i The task of exchange It refers to the change in the overall system performance after the two tasks are swapped.

[0053] 2. A collaborative action task network reconstruction method based on an improved contract network algorithm.

[0054] In the process of multi-agent collaborative action, when a sudden situation occurs in the task scene, a concurrent transaction mechanism is adopted, and the idle agents with task execution ability act as bidders to bid for the multiple time sequence tasks of the tenderer; since a single bidder cannot know the complete action plan and cannot calculate the entire task engagement time, the bidding document no longer contains the evaluation information of the engagement time at this time. Since each bidding unit has its own task sequence, the order of the agent executing the task should be given in the document, so the bidding document in the single-task auction becomes the bidding document in the multi-task auction wherein is the task to be auctioned, is the time consumption of the action task, is the kth new task entering the auction sequence, and I is the position of the new task inserted into the original task sequence of the agent.

[0055] Then the tenderer matches all the documents submitted by the bidders to obtain multiple matched documents and multiple feasible action plans, and then evaluates and compares the entire engagement time of these plans to obtain the reconstructed collaborative action task network.

[0056] During the execution of the initial collaborative action task network by each agent:

[0057] (1) A sudden situation occurs in the scene: if the sudden situation of the task scene is the emergence of a new threat, agent failure or damage, and there is only one task in its task sequence, go to step (2) to conduct a single-task auction based on two types of contracts; if the sudden situation is the discovery of a new target, the agent is disabled, and there are multiple time sequence tasks in its task sequence, go to step (5) to conduct a multi-task auction based on the concurrent transaction mechanism.

[0058] Among them, if the agent encountering a sudden situation has the ability to bid (computing power, energy, communication ability, etc. meet the bidding demand), it will be the tenderer; if it does not have the ability to bid (such as failure or damage), the agent closest to it and having the ability to bid will be the tenderer, and the agent encountering a sudden situation will be bid for the task to be executed; if two or more agents discover the same sudden situation at the same time, the nearest one will be the tenderer.

[0059] (2) Publish the bidding information: the tenderer U i publishes the task T h i to be auctioned as the bidding information; wherein the task to be auctioned is the task in its task sequence; due to the sudden situation, the agent cannot continue to complete the task, and needs to conduct a task auction in the form of bidding to enable other agents with the ability to execute the task to complete the task.

[0060] (3) Determine whether to submit a bid: In addition to the tenderer U i After receiving the bidding information, the other intelligent agents, excluding U, determine whether they can participate in the bidding as bidders based on their initial state, such as current position, task sequence, task execution capability, and remaining flight time: when intelligent agent U... j The current task sequence is not full and there are tasks to execute. When the capability is met, it is used as the bidder to execute a); when the intelligent agent U j The current task sequence is full and there are tasks to execute. When considering the ability, it will be used as the bidder to execute (b).

[0061] a) Bidder U j Tenderer U i Auction task Incorporate tasks into their own task sequence to achieve non-deadlock sorting; compute and execute tasks. Time consumption Submitting the tender document for the sale and purchase contract That is, the tender document Includes tasks Time consumption Two parts.

[0062] b) Bidder U j Tenderer U i Auction task Incorporate it into its own task sequence to complete a deadlock-free sorting; through calculation and comparison, if U... j Task sequence S j Tasks in With the task of auction The swap will reduce the total time consumption of the entire task sequence. If U is the smallest, then j To the tenderer U i Submission and exchange of contract tender documents

[0063] (4) Calculate the bidding results: using the formula Calculate the tender document price; where Representing the j-th bidder U j Execute the task Time consumption, Representing bidder U j The total time consumed for the entire task sequence after the task exchange; n represents the number of bidders.

[0064] Tenderer U iWithin a preset time, all received sales contract bids and exchange contract bids are evaluated and compared. Finally, the bid with the lowest price is selected from all bidders in this round, and the bidder with the shortest action time is selected and the bid information is sent. The bid information is sent to other bidders, and the bid rejection information is sent. Then, the transition step (8) is initiated.

[0065] If the lowest bid price is for an exchange contract, then the bid from bidder U of the exchange contract will be awarded. j The exchanged task T l j Including bidder U i In the bidding process, the tenderer U i The next tender will be launched for this task, and other bidders will be invited to submit their bids; that is, the tenderer is equivalent to a task recycling and allocation platform.

[0066] (5) Issuing a tender: Tenderer U i Timing multitasking As part of the public release of bidding information, Need to be The tasks are carried out in advance; for example, the former is a reconnaissance mission, and the latter is a strike mission.

[0067] (6) Determine whether to bid: After receiving the bidding information, each agent determines whether it can participate in the bidding based on its current position, task sequence, task execution capability, remaining flight time, and other initialization status. In agent U j Within its capabilities, then the intelligent agent U j As a bidder, Add to its own task sequence S j middle;

[0068] Assume bidder U j Task sequence S j There is m j For each task, then obviously S can be added j m j +1 different positions; if all these new task sequences can be executed, then m can be obtained. j +1 more options, at this point U j against m can be proposed j +1 type of tender document l represents Insert into S j The l-th position.

[0069] (7) Bidding Matching: Bidder U i All received tenders will be processed within the preset time. To perform bid matching, if there is a matching set of bids: Bidder U j1 against Tender documents were submitted Bidder U j2 against Tender documents were submitted Evaluate this matching tender and calculate the action time for its corresponding action plan. The action time mentioned above refers to the time of bidder U. j1 Execute the task first Afterwards, bidder U j2 Execute the task again Total time consumed.

[0070] Assuming there are r sets of mutually matching bids, perform the same operation on each of these matching bids and compare the calculation results; select the set of matching bids with the shortest action time among all the calculation results, select the bidder who submitted the bid as the successful bidder and send the winning bid information, and send the unsuccessful bid information to the other bidders.

[0071] Furthermore, if no matching tender can be obtained, meaning no agent has a suitable tender... If a task is bid on in k=1,2, the current intelligent agent system will abandon handling the unexpected situation.

[0072] (8) Result announcement: After receiving the winning bid message, the bidder sends a confirmation message to the tenderer. Both parties sign the contract and update their respective task statuses, and announce the winning bid results to other intelligent agents. After the winning bidder updates the task sequence, it continues to execute the task according to the new task sequence.

[0073] If an agent encounters an emergency and is unable to make a bidding process, and another agent makes the bidding on its behalf, then the agent that makes the bidding will continue to execute its task sequence after completing the bidding.

[0074] Example:

[0075] 1. Experiment setup.

[0076] The cooperative action task network reconstruction algorithm proposed in this project does not require reallocating task schemes to all agents, but rather performs rapid local adjustments based on the pre-assigned schemes. The following simulation verification is conducted using the example of a new target appearing in the scenario.

[0077] 1.1 Initial setup for Case 1

[0078] Case 1 (Node Failure): At t=60s, agent 3 (reconnaissance and strike drone) fails. The nearest adjacent control node performs a replanning task. The agent states are shown in Tables 1 and 2, where nodes 1 to 12 represent our reconnaissance and strike drone group, and nodes 13 to 20 represent the enemy target units. The scenario is as follows... Figure 2 As shown in the diagram. Blue nodes represent our reconnaissance and strike drones, orange nodes represent enemy target units, blue arrows represent tracking and positioning tasks, red arrows represent attack tasks, and control nodes are not shown in the diagram.

[0079] Table 1. Status Settings of Our Agent

[0080]

[0081] Table 2. Opponent Agent State Settings Table

[0082]

[0083] 1.2 Initial setup for Case 2.

[0084] Case 2 (New Target Discovery): At t=60s, a new target 21 appears at a distance of 15km and 85km within the scenario area. A bidding contract is issued by the neighboring intelligent agent, reconnaissance and strike drone 7, to reconstruct the collaborative action task network. The agent states are shown in Tables 3 and 4, where nodes 1 to 12 represent our own reconnaissance and strike drone group, and nodes 13 to 21 represent the opposing target unit. The scenario is as follows... Figure 3 As shown in the diagram. Blue nodes represent our reconnaissance and strike drones, orange nodes represent target units, blue arrows represent tracking and positioning tasks, and red arrows represent attack tasks. Control nodes are not shown in the diagram.

[0085] Table 3. Status Settings of Our Agent

[0086]

[0087]

[0088] Table 4. Opponent Agent State Settings Table

[0089]

[0090] 2. Analysis of experimental results.

[0091] An improved contract network algorithm was used to address node failures encountered during collaborative tasks (Case 1), resulting in the new agent task allocation results shown in Table 5. Here, S, I, and T represent the reconnaissance node, attack node, and target node, respectively, with subscript numbers indicating node numbers. The order of collaborative task completion and the collaborative action task network status are as follows: Figure 4 As shown.

[0092] Table 5. Results of Collaborative Task Reassignment in Case 1

[0093]

[0094]

[0095] An improved contract net algorithm was used to address the situation where new targets are discovered during collaborative tasks (Case 2). The new agent task allocation results are shown in Table 6. The order of completion of collaborative tasks and the state of the collaborative action task net are as follows: Figure 5 As shown.

[0096] Table 6 Results of Collaborative Task Reassignment in Case 2

[0097]

[0098] Simulation analysis of the collaborative action task network reconstruction in the two cases above demonstrates that the proposed dynamic collaborative action task network reconstruction method based on the improved contract network is fast and effective in addressing situations such as agent dysfunction and the discovery of new targets during collaborative tasks. It can promptly provide collaborative task solutions in unexpected scenarios while meeting task timing requirements, ensuring the normal progress of collaborative tasks. Furthermore, the simulation times for the collaborative action task network reconstruction algorithms are 2.77s and 2.95s, respectively, which is significantly shorter than the more than 5s required by most current centralized optimization algorithms, thus better adapting to unexpected scenario conditions.

[0099] 3. Experimental comparison and analysis.

[0100] The analysis experiment, with a fixed agent configuration, compared and analyzed the optimization speed of two algorithms—the classic ContractNet protocol and the improved ContractNet protocol—by varying the total number of newly added targets. In the traditional ContractNet (CN) algorithm, the tenderer selects the agent with the longest task execution time each time, the tender task selects the target with the highest task importance, and the bidders are the remaining agents. Two types of tender documents are used: buying and selling, and exchanging. The winning bid selection method is the same as the improved ContractNet algorithm. In Case 2 (adding targets), to verify the effectiveness of the algorithm with multiple UAVs of different scales, the experiment added 1 to 10 targets to the set scenario, with randomly generated locations. To improve the realism of the comparison results, each group of experiments was run 20 times, and the average computation time and communication volume were compared. The average number of negotiations and algorithm time results for the above cases are shown below. Figure 6 , Figure 7 As shown.

[0101] As clearly shown in the figure, the improved contract network algorithm has significantly fewer negotiation attempts and shorter processing time compared to the traditional contract network algorithm. This is because the improved algorithm merges multiple tasks into a single tender document, allowing multiple bidders to submit tenders simultaneously. Furthermore, it reduces the number of invalid bids that do not meet timing constraints, resulting in a substantial decrease in the number of negotiations. This effectively reduces the amount of communication between agents, lowers the system's communication costs and negotiation time, and improves the system's negotiation efficiency.

[0102] The overall performance comparison chart of the allocation schemes obtained from the above cases is shown below. Figure 8 As shown, the numbers at the bottom represent the number of bidding rounds. It can be seen that in each bidding round, the improved contract network algorithm achieves a higher overall performance value than the traditional contract network algorithm. The improved contract network algorithm is better able to allocate tasks and reconstruct the collaborative action task network.

[0103] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for reconstructing a multi-agent cooperative action task network based on an improved contract network, characterized in that: include: First, tasks are assigned to multiple agents to form an initial collaborative action task network; If no unforeseen circumstances occur during the execution of the tasks by the agents, the assigned tasks will be executed accordingly. If an agent encounters an unexpected situation during task execution, the improved network algorithm is used to reconstruct the collaborative action task network and establish a new task allocation scheme for each agent. The improved contract network algorithm mentioned above includes: Determine the type of emergency situation in the task scenario, and combine the number of tasks in the agent's task sequence to determine whether to conduct a single-task auction based on two contract types or a multi-task auction based on a concurrent transaction mechanism, and determine which agent should act as the bidder. The single-task auction based on the two contract types is determined by the task sequence status of the bidders, including two scenarios: submitting a sales contract bid to include the tenderer's task in the bidder's task sequence, and submitting an exchange contract bid to exchange the tenderer's task and a task in the bidder's task sequence; the tenderer determines the final successful bidder based on the time cost. In the multi-task auction based on the concurrent transaction mechanism, bidders create different bids by adding the bidder's multiple tasks to their own task sequence. The bidder then matches and compares all bids to determine the final winner.

2. The method for reconstructing a multi-agent cooperative action task network based on an improved contract network according to claim 1, characterized in that, Determine the type of unexpected situation in the task scenario, and based on the number of tasks in the agent's task sequence, determine whether to conduct a single-task auction based on two contract types or a multi-task auction based on a concurrent transaction mechanism, including: If the unexpected situation in the task scenario is the emergence of a new threat, the failure or damage of the agent, and there is only one task in its task sequence, a single-task auction based on the two contract types will be conducted; if the unexpected situation is the discovery of a new target, the incapacitation of the agent, and there are multiple time-series tasks in its task sequence, a multi-task auction based on the concurrent transaction mechanism will be conducted.

3. The method for reconstructing a multi-agent cooperative action task network based on an improved contract network according to claim 1, characterized in that, Determine which intelligent agent will act as the tenderer. include: Specifically, if an agent encountering an emergency has the ability to bid, it will be designated as the bidder; if it does not have the ability to bid, the nearest agent with the ability to bid will be designated as the bidder to bid for the tasks to be performed by the agent encountering the emergency; if two or more agents discover the same emergency at the same time, the one closest to it will be designated as the bidder.

4. The method for reconstructing a multi-agent cooperative action task network based on an improved contract network according to claim 1, characterized in that, Single-task auctions based on two contract types specifically include: Tenderer U i The task to be auctioned This information was released to the public as part of a bidding process. After receiving the bidding information, each agent determines whether it can participate in the bidding as a bidder based on its own initialization state: when agent U... j The current task sequence is not full and there are tasks to execute. When the capability is met, it is used as the bidder to execute a); when the intelligent agent U j The current task sequence is full and there are tasks to execute. When considering the capabilities, use them as bidders to execute (b); a) Bidder U j Tenderer U i Auction task Incorporate tasks into their own task sequence to achieve non-deadlock sorting; compute and execute tasks. Time consumption Submitting the tender document for the sale and purchase contract b) Bidder U j Tenderer U i Auction Task T h i Incorporate it into its own task sequence to complete a deadlock-free sorting; through calculation and comparison, if U... j Task sequence S j Task T in l j With the task of auction The swap will reduce the total time consumption of the entire task sequence. If U is the smallest, then j To the tenderer U i Submission and exchange of contract tender documents Calculate the bidding results: using the formula Calculate the tender price, where n represents the number of bidders; Tenderer U i Within a preset timeframe, all received sales contracts and exchange contracts are evaluated and compared. The bidder with the lowest price, which is also the bidder with the shortest execution time, is selected as the winner of this round of bidding. The winning bid is then sent to the other bidders.

5. The method for reconstructing a multi-agent cooperative action task network based on an improved contract network according to claim 4, characterized in that, If the lowest bid price is for an exchange contract, then the bid from bidder U of the exchange contract will be awarded. j The exchanged task T l j Including bidder U i In the bidding process, the tenderer U i A new tender will be issued for this task, and other bidders will be invited to submit their bids.

6. The method for reconstructing a multi-agent cooperative action task network based on an improved contract network according to claim 1, characterized in that, Multi-task auctions based on concurrent transaction mechanisms specifically include: Tenderer U i Timing multitasking As part of the public release of bidding information, Need to be Execute before; After receiving the bidding information, each intelligent agent determines whether the task... k=1,2 in agent U j Within its capabilities, then the intelligent agent U j As a bidder, Add to its own task sequence S j middle; Assume bidder U j Task sequence S j There is m j One task, S can be added j m j +1 different positions; if all these new task sequences can be executed, then m can be obtained. j +1 more options, at this point U j against propose m j +1 type of tender document l represents Insert into S j The l-th position; Tenderer U i All received tenders will be processed within the preset time. To perform bid matching, if there is a matching set of bids: Bidder U j1 against Tender documents were submitted Bidder U j2 against Tender documents were submitted Evaluate the matching tender and calculate the action time for its corresponding action plan; The tenderer selects the matching bids with the shortest action time from all matching bids, chooses the bidder who submitted the bid as the successful bidder, sends the winning bid information, and sends the unsuccessful bid information to the other bidders.

7. The method for reconstructing a multi-agent cooperative action task network based on an improved contract network according to claim 6, characterized in that, The action time refers to the time of action for bidder U. j1 Execute the task first Afterwards, bidder U j2 Execute the task again Total time consumed.

8. The method for reconstructing a multi-agent cooperative action task network based on an improved contract network according to claim 6, characterized in that, If no matching bids can be obtained, meaning no agent has a suitable bid... If a task is bid on in k=1,2, the current intelligent agent system will abandon handling the unexpected situation.

9. The method for reconstructing a multi-agent cooperative action task network based on an improved contract network according to claim 1, characterized in that, After receiving the winning bid notification, the bidder sends a confirmation message to the tenderer. Both parties sign the contract, update their respective task statuses, and announce the winning bid to other intelligent agents. After updating the task sequence, the winning bidder continues to execute the task according to the new task sequence.

10. A computer-readable storage medium storing a computer program; characterized in that, When the computer program is executed by the processor, it implements the multi-agent cooperative action task network reconstruction method based on the improved contract network as described in any one of claims 1-9.

Images