2821

2768

كارشناسي ارشد

مهندسي صنايع

اصفهان: دانشگاه اصفهان، دانشكده صنايع وسيستمها

1383

سيزده، 249، [II]ص.: مصور، جدول، شكل، نمودار

جمشيد پرويزيان

مهدي بيجاري

فارسي به انگليسي

21/4/1384

قاسم مصلحي، علي شاهنده

1396/10/26

كتابنامه

صنايع و سيستمها

مهندسي صنايع و سيستم ها

ID2768

به فارسي و انگليسي: قابل رويت در نسخه ديجيتالي

Abstract The focus of this thesis is on the applications of competitive and cooperative multi agent systems MAS in distributed problem solving MAS concept is the latest approachto the artificial intelligence with the distinguished distributed computation aspect thatmakes it powerful to deal with a broad range of complex systemsand problems There are two main contributions to the filed presentedin this thesis The first one is inthe mechanism design techniques which form the interactions of agents in a competitivesituation efficiently The second is a new framework based on the swarm intelligenceconcept to form interactions among agentsin cooperative systems Rational emotional agent which is for first time proposed in this thesis is a novel ideain MAS problem solving methods Although it can be addressed to the engineeringsciences literature the concept as proposed here is in many aspects new and suitable forboth of human involved systems and distributed problem solving tasks A rational emotional agent may choose to act optimum only according to its utility function and itspersonality but not necessarily in the designer s or any other homogeneous with differentpersonality agents point of view In other words in a multi emotional agent system twohomogeneousagents with different personalities may act differently in the same situationto reach a similar goal In the MAS design point of view emotions make MAS approachflexible to be able to deal with incomplete definitions of the agents behaviors and environment by the designer In addition it acts as an environment abstraction tool andthus leads to more efficient and easier MAS design Using the emotional agents ensure the ability of MAS to obtain the global optimum solution In addition a mechanism is designed by defining a governor level between therational emotional agents in MAS By defining a government the common single levelagents society converts to a bi level one This makes it possible to facilitate the learningprocedure of entire system also to increase the convergence speed of the system to the global or near global optimum The idea of rational emotional agents is then extended to include cooperative systems Emotional Cooperative Multi Searcher Agents ECM is a model based search algorithm which based on the essential concepts of swarm intelligence The main advantage of this algorithm is in its ability to self determine parameters of the search Numerical results show that the proposed method dominates the state of the art algorithms to solve some of the difficult combinatorial optimization problems A novel cooperative algorithm which is called Cooperative Multi Expert Agent Systems COMES is also proposed to solve multi objective combinatorial optimization problems It uses some aspects of group problem solving methods used by humankind societies especially value and concurrent engineering to provide a good estimation from Pareto front COMES is compared by the state of the art evolutionary methods to solve some multi objective combinatorial optimization problems and its efficiency is welljustified Two new hyper heuristic algorithms a novel heuristic algorithm for architecture designing and training of the feed forward neural networks as well as using self organizing maps as one of the most powerful data mining tools for analyzing the numerical results and visualizing the searchspace are other contributions of this thesis cJ j if Lf VI II IJ

جمشيد پرويزيان

مهدي بيجاري

قاسم مصلحي، علي شاهنده