Research On Self-adaption Adjustment Mechanism Of Multi-agent Network Based On Trust Model

In recent years,with the continuous development of multi-agent technology and distributed artificial intelligence,the requirements for Agents to have independent decision-making ability are getting higher and higher.The operating environment of Multi-Agent Systems(MAS)is becoming more and more large,open,dynamic and uncertain.At this time,various intelligent technologies need to be applied to build an agent which has self-adaptive ability.In the MAS,it is a key to study the MAS by how to complete the tasks that the single agent can't accomplish.At the same time,MAS are usually used to process multiple complex tasks.These complex tasks are composed of several different sub-tasks.In this case,agents with different resource quantities and resource types need to cooperate to complete tasks.Therefore,how to improve the efficiency of task completion and shorten the task completion time has become the focus of this paper.To this end,in this paper,a composite self-adaption mechanism in an agent network is proposed,which studies the complex task decision problem from the perspective of task assignment and task execution.The main work is as follows:(1)In view of the shortcomings of the existing distributed MAS structure,this paper firstly modeled the Multi-Agent Network and adopted a connection model with weighted relations.The strength of the relationship between agents is represented by the weight.Considering that Agent needs to cooperate with other Agents to complete tasks in the interaction process,and each Agent hopes to find the most valuable partners.This paper introduces a trust model based on social network view,and evaluates the trust degree of target Agent comprehensively from different approaches of trust acquisition(direct trust and recommendation reputation).Finally,the improved reinforcement learning method is used to make agents learn how to adjust the relationship strength between agents,so as to optimize the structure of Agent Network.Experimental results show that the proposed method can optimize the system benefits and reduce the task execution time.By comparing the weighted relations with crisp relations,it is proved that the weighted relations play an important role in improving the system performance.(2)At present,unreliability is another new problem of MAS.Considering the existence of cheating agents in MAS,the cheating behavior(false declaration,breach of contract,etc.)taken in the behavior interaction has a bad influence on the comprehensive performance of MAS.Therefore,in order to improve the comprehensive performance of the MAS,we propose an agent division mechanism,which can adjust the state according to the task performance of Agent and use the method of division to efectively divide the resources,network connections and task queues of Agent in the system and realize the aggregation effect of reliable resources.Experiments show that the proposed method improves the efficiency of task completion and reduces the time consumption of task cooperation.Wireless sensor network(WSN)is a distributed self-adaption network system integrating data acquisition,processing and communication.The energy of sensor nodes in WSN is limited,and the addition of new sensor nodes or failure of existing nodes will affect the topology of WSN.Therefore,WSN system is required to have the ability to adapt to the changing environment.By using the adaptive adjustment mechanism for task allocation proposed in this paper and adjusting the relationship between sensor nodes according to the task completion,so as to optimize the topology of WSN.The characteristics of WSN itself make it more vulnerable to various spoofing attacks.The Agent division method proposed in this paper can effectively detect spoofing attacks on nodes in WSN,thus improving the reliability of data transmission and task completion efficiency in WSN system.