Research And Application Of Multi-agent System Optimization For Domain-oriented Dynamic Task Solving

With the rapid development of computer networks communication technologies and multimedia technologies,the solution of dynamic and complex problems has become a difficult problem.Computers are gradually moving toward intelligence.The development of Multi-Agent System(MAS)is conducive to the gradual intellectualization of large and complex problems in the real world.In this paper,a domain-oriented complex problem is solved by transforming a complex problem into doing a task.Based on the existing modular knowledge system,we construct a model for solving dynamic tasks based on the Multi-Agent system.The Multi-Agent System is optimized by solving a certain number of different problems with the common type in the common field and the efficiency of solving complex problems is further improved.The main works of this paper are as follows:Firstly,this paper expounds the key point of solving dynamic complex problem and tasks,reviews the current research development of the Agent technology and analyzes the advantages and difficulties of using Agent-Based technology to solve tasks by reading the domestic and foreign documents.Secondly,a dynamic task solving model based on Multi-Agent System is established.Due to the diversity of fields and the complexity of dynamic task,it is a very difficult process to achieve the design and construction of an unified Multi-Agent System suitable for all fields.In this paper,combining the domain module knowledge system,the module knowledge system in the domain is decomposed and reorganized by clustering algorithm.The reassembled clusters are packaged to form each Agent's knowledge base and form a Multi-Agent System.Thirdly,design the task solving process.For a complex task,a “divide and conquer” strategy is adopted.The complex task is decomposed into several sub-tasks.Every sub-task is solved by several corresponding Agents.The entire complex task is solved by several Agents,which are called the Agent team.The solution of the whole complex task is solved by the teamwork of several solving Agent teams.Fourthly,a feedback optimization model is established.In order to make the established Multi-Agent System more effective in task solving,this paper establishes a feedback optimization model and optimizes the Multi-Agent system by solving different problems of the same type in the field.This paper takes the design of conveyor reducer as an example to optimize the multi-agent system in the field of mechanical product part design by using clustering method.