A group distributed negotiation method based on global information

Abstract

The present invention provides a group distributed negotiation method based on global information. The present invention is applied to an agent group composed of multiple homogeneous or heterogeneous agents. During the movement of a single agent, it can detect and obtain the location information of users within the detection range in real time, and exchange information with other agents in the group. In the present invention, a single agent obtains global information through an information fusion mechanism based on interaction information and detection information; each single agent uses global information through a clustering algorithm to obtain a set of target points for the next stage; the multi-agent group completes task allocation through a distributed negotiation mechanism and obtains its own target point for the next stage; the agent moves to the target point and repeats the above process to solve the multi-agent decision-making and task allocation problems.

Description

Technical Field

[0001] The present invention relates to the field of multi-agent decision-making and task allocation, and in particular to a group distributed negotiation method based on global information. Background Art

[0002] With the improvement of technology, single agents have good capabilities in many specific application scenarios and can help users achieve specified functions. As the complexity of application scenarios increases, single agents are limited to perception and task execution capabilities and can no longer meet user needs. Multi-agent joint task execution has become the best choice for solving high-complexity tasks. In the current research in the field of multi-agents, in order to ensure the completion of tasks, it is necessary to improve group decision-making and task allocation capabilities.

[0003] In the current research content of decision-making and task allocation, it can be divided into centralized decision-making and task allocation methods and distributed decision-making and task allocation methods according to the source of group decision-making and task allocation information. The characteristics of the centralized decision-making and task allocation method are: after a single intelligent agent obtains the perception information of the group, it obtains the global task allocation plan through the algorithm, and sends the plan to other intelligent agents in the group to command other intelligent agents to operate, so as to achieve the purpose of decision-making and task allocation. Its advantages are that the information is comprehensive and the task allocation results are theoretically better. Its disadvantages are that it relies too much on the central node, and the computing load of the central node is large, and practical applications are not considered.

[0004] The characteristics of distributed decision-making and task allocation methods are: a single agent makes decisions independently based on local information. Its advantage is that the group has strong anti-interference ability and the algorithm complexity is reduced compared to the centralized method. The disadvantage of current research in this direction is that it does not fully utilize the communication capabilities of the agents, lacks a decision-making negotiation mechanism, is not conducive to conflict resolution, and the decision results are also limited due to information limitations. Summary of the invention

[0005] In order to solve the problem of multi-agent decision-making and task allocation, this paper proposes a group distributed negotiation method based on global information. Under the distributed decision-making and task allocation architecture, this method makes full use of the communication capabilities between agents, resolves decision conflicts through negotiation mechanisms, improves the operation performance of multi-agent groups, and solves the problem of limitations in decision results.

[0006] In order to achieve the above object, the technical solution adopted by the present invention is:

[0007] A group distributed negotiation method based on global information, wherein the group is composed of multiple intelligent agents, comprises the following steps:

[0008] (1) A single agent detects the user's identification code in real time during movement and obtains the location information of the user within the detection range. It then exchanges the detection information consisting of the user's identification code and the corresponding location information with other agents in the group.

[0009] (2) Each agent combines the information from other agents with its own detection information to obtain global information through the information fusion mechanism;

[0010] (3) When all agents in the group have reached their own target points, each agent uses a clustering algorithm to obtain a set of target points to be selected;

[0011] (4) Multiple agents obtain their own next-stage target through a distributed negotiation mechanism;

[0012] (5) Each agent moves toward the target point and returns to step (1) to continue the task.

[0013] Furthermore, the intelligent agents have communication, detection and movement capabilities, and the intelligent agents in the group can communicate with other intelligent agents in the group, and the intelligent agents in the group have a limited detection range.

[0014] Furthermore, in step (1), the intelligent agent obtains the identification code and corresponding location information of the user within the detection range through the onboard sensor.

[0015] Furthermore, in step (1), the intelligent agent sends the obtained detection information to other intelligent agents in the group through the communication system, and the other intelligent agents receive the information.

[0016] Furthermore, each intelligent agent maintains a global situation table, which contains the user's identification code and corresponding location information; in step (2), each intelligent agent dynamically updates the global situation table of the agent based on the detection information of the agent and the detection information sent by other intelligent agents in the group.

[0017] Furthermore, the clustering algorithm in step (3) is a K-means clustering algorithm, in which the initial cluster center position is randomly selected, the number of cluster centers is the number of agents in the group, and the output is a cluster center point set, i.e., a set of target points to be selected.

[0018] Furthermore, the specific method of the distributed negotiation mechanism in step (4) is as follows: the intelligent agents in the group select the best position in the subject's target point set in turn based on the asynchronous communication mechanism and their own target point set, and send the selection information to other intelligent agents in the group; other intelligent agents delete the target points related to the received information from their respective target point sets based on the conflict resolution mechanism until all intelligent agents complete the target point selection.

[0019] Compared with the prior art, the present invention has the following beneficial effects:

[0020] 1. In view of the problem that the existing distributed decision-making and task allocation methods do not fully utilize the inter-machine communication capabilities, the present invention uses an information interaction mechanism to enable the intelligent agents in the group to send and receive detection information, thereby improving the comprehensiveness of information acquisition of a single intelligent agent.

[0021] 2. In view of the problem of resolving decision conflicts in existing distributed decision-making and task allocation methods, the present invention designs a distributed negotiation mechanism to solve the problem of decision result conflicts faced by distributed decision-making, and reasonably limits the total amount of information transmission in distributed decision-making, thereby improving the performance of distributed decision-making and task allocation. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 This is a flow chart of a group distributed negotiation method based on global information provided by an embodiment of the present invention.

[0023] Figure 2 This is a schematic diagram of the distributed negotiation global situation representation provided by an embodiment of the present invention.

[0024] Figure 3 It is a schematic diagram of a distributed negotiation mechanism provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0025] The following will be combined with the embodiments to clearly and completely describe the concept, technical solution advantages and technical effects of the present invention, so as to fully understand the purpose, characteristics and effects of the present invention. It should be noted that the specific embodiments described herein are only used to explain the present invention and do not limit the present invention.

[0026] A group distributed negotiation method based on global information, see Figure 1 First, a multi-agent group model is constructed. The agents in the group perform detection tasks while moving towards the target point, and exchange the detected information with other agents. After that, the single agent in the group establishes and updates the global situation table based on the received information and the body detection information, performs the information fusion process, and obtains the global information representation. After that, when all agents reach the target point of the body, the target point candidate set is obtained through the clustering method. Finally, the distributed negotiation process is executed to select the target point of each next stage, and the above process is repeated.

[0027] The specific steps of this method are as follows:

[0028] S1: Modeling for scenarios.

[0029] Multi-agent group A = {A1, A2, ..., A n}It is composed of a single agent with communication, detection, and mobility capabilities. i Detection range R i limited.

[0030] S2: Information interaction among agents within the group.

[0031] Single Agent A i Towards its own target point T i During the movement, the detection range R can be obtained in real time i Internal user identification code and the corresponding location information And the ID i Information and P i Information is sent to other drones in the group through the information exchange mechanism. -i ={A1, ..., A i-1 , A i+1 …, A n}.

[0032] S3: Establish a global situation table and execute the information fusion process.

[0033] Single Agent A i It has a global situation table that stores the detected users and their corresponding location information, namely S i ={(ID i1 , P i1 ), (ID i2 , P i2 ),…,(ID iall , P iall )}. i Ontology exploration information and A -i Pass the received (i-1) messages Execute the information fusion process and update S i Due to the existence of the communication mechanism, all agents can reach The global situation table and update process are as follows Figure 2 shown.

[0034] S4: Based on the clustering algorithm, a set of target points to be selected is obtained.

[0035] When the agent A in the group i All reach the target point T i After that, single agent A i Through K-means clustering algorithm, based on S iData, input random n cluster center initial positions, calculate n cluster centers C i ={C i1 , …, C in}, the composition belongs to A i The set of target points to be selected, n is equal to the number of agents in the group. Since the initial position is randomly generated, there is

[0036] S5: Based on the distributed negotiation mechanism, the target point of the next stage is calculated through global information, and then the process returns to step S2 to repeat the above steps.

[0037] The agents in the group complete the target point selection based on the distributed negotiation mechanism. The schematic diagram of the distributed negotiation mechanism is as follows: Figure 3 That is, through the asynchronous communication mechanism, the target point closest to the main body is selected in turn as the target point of the next stage, for example, A i In C i , select And send the selection information to other agents A in the group -i After receiving the information, other agents delete the target point closest to the selected position from the candidate target point set of the agent based on the conflict resolution mechanism, such as A j (i≠j, j∈[1,n]) received a i Select Message Then A j In the ontology target point set C j ={C j1 , …, C jn}Delete distance The nearest point, assumed to be C jd , then A j Change of the set of points to be selected C j ={C j1 , …, C jd-1 , C jd+1 , …, C jn}. This process will continue until all agents complete the selection of the next stage target point. Agent set A = {A1, A2, ..., A n Through the distributed negotiation mechanism, the next stage target point set T = {T1, T2, ..., T n}, each agent moves to the local target point and returns to step S2 to continue the task.

[0038] In summary, the present invention is applied to an agent group composed of multiple homogeneous or heterogeneous agents, which is based on a distributed negotiation mechanism and fully utilizes the communication capabilities between agents. In the present invention, a single agent can detect and obtain the location information of users within the detection range in real time during movement, and exchange information with other agents in the group. Based on the interaction information and detection information, the single agent obtains global information through an information fusion mechanism; each single agent uses the global information through a clustering algorithm to obtain a set of target points for the next stage; the multi-agent group completes task allocation through a distributed negotiation mechanism and obtains its own target point for the next stage; the agent moves to the target point and repeats the above process to solve the problem of multi-agent decision-making and task allocation.

Claims

1. A group distributed negotiation method based on global information, characterized in that: The group is composed of multiple intelligent agents and includes the following steps: (1) A single agent detects the user's identification code in real time during movement and obtains the user's location information within the detection range. It then exchanges the detection information consisting of the user's identification code and corresponding location information with other agents in the group. (2) Each agent combines the information from other agents with its own detection information to obtain global information through the information fusion mechanism; (3) When all agents in the group have reached their own target points, each agent uses a clustering algorithm to obtain a set of target points to be selected; the clustering algorithm is the K-means clustering algorithm. The initial cluster center position is randomly selected in the clustering algorithm. The number of cluster centers is the number of agents in the group. The output is the set of cluster center points, that is, the set of target points to be selected; (4) Multiple agents obtain their own target points for the next stage through a distributed negotiation mechanism. The specific method of the distributed negotiation mechanism is as follows: the agents in the group use an asynchronous communication mechanism to combine their own candidate target point set, select the best position in the candidate target point set in turn, and send the selection information to other agents in the group; other agents use a conflict resolution mechanism to delete the target points related to the received information from their respective candidate target point sets until all agents complete the target point selection; (5) Each agent moves toward the target point and returns to step (1) to continue the task.

2. A group distributed negotiation method based on global information according to claim 1, characterized in that: The intelligent agents have communication, detection and movement capabilities. All intelligent agents in the group can communicate with other intelligent agents in the group. All intelligent agents in the group have limited detection ranges.

3. The group distributed negotiation method based on global information according to claim 1, characterized in that: In step (1), the intelligent agent obtains the identification code and corresponding location information of the user within the detection range through the onboard sensors.

4. The group distributed negotiation method based on global information according to claim 1, characterized in that: In step (1), the intelligent agent sends the obtained detection information to other intelligent agents in the group through the communication system, and other intelligent agents receive the information.

5. The group distributed negotiation method based on global information according to claim 1, characterized in that: Each intelligent agent maintains a global situation table, which contains the user's identification code and corresponding location information. In step (2), each intelligent agent dynamically updates the global situation table of the agent based on the detection information of the agent and the detection information sent by other intelligent agents in the group.

Images