Multi-agent system consistency method under switching topology based on TOPSIS comprehensive decision

Abstract

The invention designs a multi-agent system consistency method under switching topology based on TOPSIS comprehensive decision. The method comprises an important agent evaluation strategy based on a TOPSIS model and a multi-agent distributed consistency protocol for important agents. According to the method, a topological structure formed by an intelligent agent at each moment is mined, useful attribute information is extracted, a comprehensive decision is made based on a TOPSIS model, and the importance of the intelligent agent is quantified; meanwhile, based on the screened important agents, a consistency protocol is redesigned, information of the important agents is effectively utilized, and convergence is promoted. The whole method can adapt to multi-agent systems of different magnitudes, solves the problem that convergence is too slow or even cannot be achieved especially under a large-scale high-density structure, and accelerates the consistency of the system.

Description

Technical Field

[0001] The present invention relates to the field of multi-agent system consistency, and in particular to a multi-agent system consistency algorithm based on TOPSIS comprehensive decision-making under switching topology conditions. Technical Background

[0002] Many creatures in nature, such as ants, birds, and fish, complete complex tasks through group behavior and information exchange. This group collaboration mechanism has inspired the design of multi-agent systems. Multi-agent systems are distributed control systems in which agents can communicate with neighboring agents through simple local interaction rules and autonomously adjust their own states, so that an orderly coordinated movement and intelligent behavior are presented at the overall level.

[0003] The consistency problem is the key in multi-agent systems. It focuses on how to make the states (such as position, speed, etc.) of all agents consistent through distributed control algorithms. The consistency problem is solved by designing a reasonable distributed control algorithm (i.e., consistency protocol). A good algorithm can promote effective communication and cooperation between agents, and make the system state tend to be unified even when the environment changes.

[0004] Network topology is divided into two categories: fixed topology and switching topology. Fixed topology means that the network topology of the multi-agent system remains unchanged as the environment and time change; switching topology means that the network topology of the multi-agent system changes as the environment changes and time passes. Usually, the agent communicates with neighboring agents based on the consensus protocol. When the position state of the agent changes or the agent fails, the network topology of the system will also change accordingly.

[0005] In existing distributed systems, the discrete-time model has uncertainty and time-varying nature in the initial number of agents and topological structure, which leads to slow convergence under the existing consistency protocol. Therefore, the key issue is how to design a distributed consistency protocol based on switching topology in the discrete-time model, so as to effectively improve the convergence speed while maintaining system stability.

[0006] The present invention proposes a multi-agent system consistency method under switching topology based on TOPSIS comprehensive decision. By extracting many attribute information of the topological network, the intrinsic connection between agents can be known; through the TOPSIS comprehensive decision method, agents can be comprehensively and accurately evaluated, important agents that are conducive to convergence can be screened out, and consistency protocols can be designed based on this to promote system convergence. Summary of the invention

[0007] The purpose of the present invention is to solve the problem that the existing consensus algorithm converges too slowly or even fails to converge in a large-scale dense multi-agent system structure. It provides an important agent evaluation strategy based on the TOPSIS model to screen out more valuable neighboring agents for communication and reduce a large amount of communication costs. At the same time, a new consistency protocol is designed to improve the convergence speed.

[0008] The TOPSIS comprehensive decision-making multi-agent system consistency method proposed in this invention updates and evolves the agent state by analyzing the topology at each moment in discrete time. The specific steps are as follows:

[0009] S100: Distributed discrete-time model multi-agent system under switching topology, At this moment, all agents Sense and store all neighboring agent information within the communication radius ;

[0010] S200: Combined with the topological structure formed by the agent information, a TOPSIS model-based important agent evaluation and screening strategy is designed to screen out important agents and make the agents Interact only with important agents, rather than all agents in the neighborhood;

[0011] S300: Design a consensus protocol and calculate the candidate states of the agents at the next moment based on the selected important agents;

[0012] S400: All agents calculate their own state at the next moment according to the consistency protocol and evolve, thereby obtaining the state of the entire system agent at the next moment, and performing consistency judgment on the multi-agent system.

[0013] Furthermore, the neighbor agent information described in step S100 , mainly includes two aspects:

[0014] (1) Neighbor agent status information: ;

[0015] (2) Topological structure attribute information: the number of neighbors of a neighbor, denoted as , the number of common neighbors with all neighbors, denoted as , and the state distance with all neighbors, denoted as The number of common neighbors with all neighbors is defined as follows:

[0016]

[0017] in, Representing a collection The modulus of ; the state distance definition expression with all neighbors is as follows:

[0018]

[0019] in, represents the L2 norm.

[0020] Furthermore, the step S200 of designing an important agent evaluation and screening strategy based on the TOPSIS model comprises the following steps:

[0021] S210: Decision Matrix Construction: Assume Agent exist At this moment neighbor agents, whose corresponding set is , the attribute information set is (corresponding to the three attributes of the number of neighbors of the neighbor described in claim 3, the number of common neighbors with all neighbors, and the state distance with all neighbors respectively), right The value of , forming a decision matrix:

[0022]

[0023] S220: Normalization: Since different attributes are described in different units, the values ​​vary greatly, resulting in dimensional effects, so normalization is required to convert the attribute values ​​of different units to the same scale to obtain the normalized matrix. , where the conversion rules are as follows:

[0024]

[0025] S230: Determine the positive ideal state and the negative ideal state: The positive ideal state represents the set of optimal values ​​selected from all neighbors to be evaluated under a certain attribute information. Indicated by; the negative ideal state represents the set of the worst values ​​selected from all neighbors to be evaluated under a certain attribute information, and is expressed by Indicates. The set elements are defined as follows:

[0026]

[0027]

[0028] In the determination of positive and negative ideal states, the number of neighbors of the neighbor , the number of common neighbors with all neighbors is a positive indicator, the state distance from all neighbors It is a negative indicator, so in the positive ideal state, the first two attributes take the maximum value and the third attribute takes the minimum value; in the negative ideal state, the first two attributes take the minimum value and the third attribute takes the maximum value.

[0029] S240: Calculate the Euclidean distance from each neighbor state to the positive ideal state and the negative ideal state: Indicates The distance from the state of the neighboring agent to the positive ideal state; Indicates The distance from the state of a neighboring agent to the negative ideal state; the specific definition is as follows:

[0030]

[0031]

[0032] S250: Calculate the evaluation reference value of each neighboring agent state: the evaluation reference value is recorded as , which is defined as follows:

[0033]

[0034] S260: Screening important neighbor agents: Sort the evaluation reference values ​​from large to small, select the neighbor agents corresponding to the top three evaluation reference values, and add them to the important neighbor set Complete the screening.

[0035] Furthermore, the distributed consistency protocol designed in step S300 has the following steps:

[0036] S310: Calculate candidate state values: Based on the dimensions of the agent state information, the new candidate state solution can be specifically expressed as:

[0037]

[0038] in, Indicated in Candidate state values ​​on the dimension, Indicates the first among all the important neighbors selected The maximum value of the state in the dimension, Indicates the first among all the important neighbors selected The minimum value of the state on the dimension.

[0039] S320: Consistency evolution rules: According to the candidate state values ​​calculated in step S310, the consistency protocol can be represented as an agent Current state Add the calculated candidate state value , which can be expressed as:

[0040]

[0041] in, Represents the trade-off coefficient, which affects the convergence speed of the system.

[0042] Furthermore, in step S400, the system consistency determination rule is that for any two agents and , if all conditions are met:

[0043]

[0044] in, Indicates the threshold value, generally speaking , represents the L2 norm.

[0045] If the system reaches a consistent state, it can be considered that the system has converged; otherwise, repeat step S100 to proceed to the next moment. evolution.

[0046] Compared with the prior art, the present invention has the following advantages:

[0047] (1) The present invention can make decisions based on the comprehensive topological multi-attribute information. Compared with other agent selection strategies, the present invention mines and utilizes the topological structure formed by agents, incorporates multiple attribute information into decision-making references, and makes the judgment of important agents more accurate;

[0048] (2) The present invention can accurately quantify the importance of intelligent agents. Compared with other evaluation strategies, comprehensive calculation based on the TOPSIS model is more accurate and the quantification of the reference value for the evaluation of the state of intelligent agents is more reasonable;

[0049] (3) The present invention can accelerate the convergence process of multi-agent systems. Compared with traditional consensus protocols, the present invention not only reduces the number of interacting neighboring agents, but also maximizes the amount of state change at each moment, which can help agents reduce communication costs while reaching state consistency faster, thus accelerating the overall convergence of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 This is a general flow chart of the consistency method of multi-agent system under switching topology based on TOPSIS decision

[0051] Figure 2 It is an intelligent agent Example diagram of system status before screening

[0052] Figure 3 It is an intelligent agent Example diagram of selecting important neighbor agents and calculating candidate states according to the consensus protocol DETAILED DESCRIPTION

[0053] The present invention is further described in detail below in conjunction with embodiments and drawings, but the embodiments of the present invention are not limited thereto.

[0054] Example 1

[0055] Those skilled in the art will understand that the following examples are only used to illustrate the present invention and should not be considered to limit the scope of the present invention. If no specific techniques or conditions are specified in the examples, the techniques or conditions described in the literature in the art or the product instructions are used.

[0056] like Figure 1 As shown in the flowchart, the multi-agent system consistency method under the switching topology based on TOPSIS decision can be divided into four steps, as follows:

[0057] S100: Distributed discrete-time model multi-agent system under switching topology, At this moment, all agents Sense and store all neighboring agent information within the communication radius ;

[0058] S200: Combined with the topological structure formed by the agent information, a TOPSIS model-based important agent evaluation and screening strategy is designed to screen out important agents and make the agents Interact only with important agents, rather than all agents in the neighborhood;

[0059] S300: Design a consensus protocol and calculate the candidate states of the agents at the next moment based on the selected important agents;

[0060] S400: All agents calculate their own state at the next moment according to the consistency protocol and evolve, thereby obtaining the state of the entire system agent at the next moment, and performing consistency judgment on the multi-agent system.

[0061] Preferably, in step S100, Figure 2 For example, each agent in the multi-agent system has the same and equal properties, and its communication radius is limited. Assume that the communication radius of the agent is .exist At this moment, with intelligent agents For example, its status information is expressed as , by sensing the agents within the communication radius, obtaining the state information of neighboring agents and counting them into the set , and count the attribute information of all neighbors, including: the number of neighbor agents, the number of neighbor agents and The number of common neighbors, neighbor agents and The state distance is recorded in the set , , The number of common neighbors with all neighbors is defined as follows:

[0062]

[0063] in, Representing a collection The modulus of ; the state distance definition expression with all neighbors is as follows:

[0064]

[0065] in, represents the L2 norm.

[0066] Preferably, in step S200, a TOPSIS model-based important agent evaluation and screening strategy is designed to Figure 2 The following example illustrates the steps:

[0067] S210: Constructing a decision matrix: The decision matrix is ​​used Indicates that each column is an attribute information, and each row represents the information of a neighboring agent to be evaluated. Represents neighbor agents In the The value under each attribute. Figure 2 For example, the agent exist At the moment, there are 6 neighbor agents, and their corresponding set is , the attribute information set is (corresponding to the three attributes of the number of neighbors of the neighbor in step S100, the number of common neighbors with all neighbors, and the state distance with all neighbors respectively), right The value of , forming a decision matrix:

[0068]

[0069] S220: Normalization: Since different attributes are described in different units, the values ​​vary greatly, resulting in dimensional effects, so normalization is required to convert the attribute values ​​of different units to the same scale to obtain the normalized matrix. , where the conversion rules are as follows:

[0070]

[0071] S230: Determine the positive ideal state and the negative ideal state: The positive ideal state represents the set of optimal values ​​selected from all neighbors to be evaluated under a certain attribute information. Indicated by; the negative ideal state represents the set of the worst values ​​selected from all neighbors to be evaluated under a certain attribute information, and is expressed by Indicates. The set elements are defined as follows:

[0072]

[0073]

[0074] In the determination of positive and negative ideal states, the number of neighbors of the neighbor , the number of common neighbors with all neighbors is a positive indicator, the state distance from all neighbors It is a negative indicator, so in the positive ideal state, the first two attributes take the maximum value and the third attribute takes the minimum value; in the negative ideal state, the first two attributes take the minimum value and the third attribute takes the maximum value.

[0075] S240: Calculate the Euclidean distance from each neighbor state to the positive ideal state and the negative ideal state: Indicates The distance from the neighbor state to the positive ideal state; Indicates The distance from the neighbor state to the negative ideal state; the specific definition is as follows:

[0076]

[0077]

[0078] S250: Calculate the evaluation reference value of each neighbor state: the evaluation reference value is recorded as , which is defined as follows:

[0079]

[0080] S260: Select important neighbor agents: According to the neighbor status evaluation reference value obtained in step S250 , select the neighboring agents corresponding to the top three evaluation reference values ​​in order from large to small, such as Figure 3 Agents 2, 3, and 4 in the example are added to the important neighbor set ;

[0081] Preferably: the distributed consistency protocol designed in step S300 is Figure 3 The example is used to illustrate the steps:

[0082] S310: Calculate candidate state values: Based on the dimensions of the agent state information, the new candidate state solution can be specifically expressed as:

[0083]

[0084] in, Indicated in Candidate state values ​​on the dimension, Indicates the first among all the important neighbors selected The maximum value of the state in the dimension, Indicates the first among all the important neighbors selected The minimum value of the state on the dimension.

[0085] S320: Consistency evolution rules: According to the candidate state values ​​calculated in step S310, the consistency protocol can be represented as an agent Current state Add the calculated candidate state value , which can be expressed as:

[0086]

[0087] in, Represents the trade-off coefficient, which affects the convergence speed of the system and can be configured in combination with the specific working conditions of the intelligent agent. Figure 3 As shown, at the next moment, the agent Will move to the current state and candidate status values On the connection line.

[0088] Preferably: in step S400, all agents evolve according to the candidate states calculated in step S300 to obtain the next moment The state of the agent; at the same time, the agent Calculate the L2 norm of its and all neighbor states. If for any neighbor agent , all satisfy the conditions:

[0089]

[0090] in, Indicates the threshold value, generally speaking , represents the L2 norm.

[0091] This means that the intelligent agent If it reaches an agreement with its neighbors, it needs to continue to judge the situation of other agents; otherwise, it means that the agent No consensus has been reached with neighboring agents.

[0092] If all agents in the system agree with their neighboring agents, it means that the system has reached a consistent state and has converged. Otherwise, go to step S100 to proceed to the next moment. evolution.

[0093] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A multi-agent system consistency method under switching topology based on TOPSIS comprehensive decision, characterized in that: Combine the information of the topological structure formed by the intelligent agent to accelerate the convergence of the system. The steps are as follows: Step 1: Switch the topology of the distributed discrete-time model multi-agent system. At this moment, all agents Sense and store all neighboring agent information within the communication radius ; Step 2: Combine the topological structure formed by the agent information and design an important agent evaluation and screening strategy based on the TOPSIS model to screen out important agents and make the agents Interact only with important agents, rather than all agents in the neighborhood; Step 3: Design a consensus protocol and calculate the candidate states of the agents at the next moment based on the selected important agents; Step 4: All agents calculate their own state at the next moment according to the consistency protocol and evolve, and then obtain the state of the entire system agent at the next moment, and perform consistency judgment on the multi-agent system.

2. According to claim 1, a multi-agent system consistency method under switching topology based on TOPSIS comprehensive decision-making is characterized in that: The multi-agent topology is time-varying, and different moments It is formed according to the state and communication radius of the intelligent agent, using the adjacency matrix Represents the topology.

3. According to claim 1, a multi-agent system consistency method under switching topology based on TOPSIS comprehensive decision-making is characterized in that: In step 1, the communication radius is a fixed value. , read and store the agent Neighbor agent information , mainly includes two aspects: (1) Neighbor agent status information: ; (2) Topological structure attribute information: the number of neighbors of a neighbor, denoted as , the number of common neighbors with all neighbors, denoted as , and the state distance with all neighbors, denoted as ; The expression for the number of common neighbors with all neighbors is as follows: in, Representing a collection The modulus of ; the state distance definition expression with all neighbors is as follows: in, represents the L2 norm.

4. According to claim 1, a multi-agent system consistency method under switching topology based on TOPSIS comprehensive decision-making is characterized in that: In the step 2, a TOPSIS model-based important agent evaluation and screening strategy is designed, which includes: (1) Construction of decision matrix: Assume that the agent exist At this moment neighbor agents, whose corresponding set is , the attribute information set is (corresponding to the three attributes of the number of neighbors of neighbors, the number of common neighbors with all neighbors, and the state distance with all neighbors), right The value of , forming a decision matrix: (2) Normalization: Since different attributes are described in different units, the values ​​vary greatly, resulting in dimensional effects. Therefore, normalization is required to convert the attribute values ​​in different units to the same scale and obtain the normalized matrix. , where the conversion rules are as follows: (3) Determine the positive ideal state and the negative ideal state: The positive ideal state represents the set of optimal values ​​selected from all neighbors to be evaluated under a certain attribute information, and is expressed as Indicated by; the negative ideal state represents the set of the worst values ​​selected from all neighbors to be evaluated under a certain attribute information, and is expressed by Represents, where the set elements are defined as follows: (4) Calculate the Euclidean distance from each neighbor state to the positive ideal state and the negative ideal state: Indicates The distance from the state of the neighboring agent to the positive ideal state; Indicates The distance from the state of a neighboring agent to the negative ideal state; the specific definition is as follows: (5) Calculate the evaluation reference value of each neighboring agent state: the evaluation reference value is recorded as , which is defined as follows: (6) Screening important neighbor agents: Sort the evaluation reference values ​​from large to small, select the neighbor agents corresponding to the top three evaluation reference values, and add them to the important neighbor set. Complete the screening.

5. According to claim 4, a multi-agent system consistency method under switching topology based on TOPSIS comprehensive decision-making is characterized in that: In the determination of positive and negative ideal states, the number of neighbors of the neighbor , the number of common neighbors with all neighbors is a positive indicator, the state distance from all neighbors It is a negative indicator, so in the positive ideal state, the first two attributes take the maximum value and the third attribute takes the minimum value; in the negative ideal state, the first two attributes take the minimum value and the third attribute takes the maximum value.

6. According to claim 1, a multi-agent system consistency method under switching topology based on TOPSIS comprehensive decision-making is characterized in that: In step 3, a consistency protocol is designed, and the new candidate state solution can be specifically expressed as: in, Indicated in Candidate state values ​​on the dimension, Indicates the first among all the important neighbors selected The maximum value of the state in the dimension, Indicates the first among all the important neighbors selected The minimum value of the state on the dimension; the consensus protocol can be expressed as an agent Current state Add the calculated candidate state value , which can be expressed as: in, Represents the trade-off coefficient, which affects the convergence speed of the system.

7. According to claim 1, a multi-agent system consistency method under switching topology based on TOPSIS comprehensive decision-making is characterized in that: In step 4, the condition for determining whether the system has reached consistency is that for any two agents and , if all conditions are met: in, Indicates the threshold value, generally speaking , represents the L2 norm; Then the system can be considered to have reached a consistent state and converged; otherwise, repeat step 1 and proceed to the next moment. evolution.

8. An intelligent electronic device, comprising a sensor, a memory, a processor, and a program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps in the multi-agent system consistency method under the switching topology based on TOPSIS comprehensive decision are implemented as described in any one of claims 1-7.

Images