Research On Issues Of Agents Coalition In MAS

Research on cooperative mechanisms among Agents has been recently a hotspot in DAI(Distributed Artificial Intelligence) domain. Agents coalition is an important method of cooperation of Agents and is a very important problem in Distributed Problem Solving(DSP), Multi-Agents System(MAS) and Game Theories.We analyzed firstly the relations among DSP, MAS and Game Theories by the review of development process of DAI and then the similarity and the difference on Agents coalition problems in the three different domains were pointed out. This is extremely good for the more all-sided understanting of the concept of Agents coalition and deep research on the coalition problem.Searching the optimal coalition structure which optimizes the total of values of coalitions is all a global problem in DPS, MAS and Game Theories. But, generally,the number of the coalition structure is too large to realize the searching of all the coalition structures. In this paper, a genetic algorithm was presented to solve the problem of the searching of the optimal coalition structure. We designd one-dimension chromosome coding structure and the corresponding improved uniformity crossover operator and the heuristic mutation operator. Compared to OBGA ( Order-based Genetic Algorithms),our algorithm can obtain the result of the higher quality and has the higher ability of globally searching and the shorter time of searching, so having the better feasibility.The task-oriented coalition formation mechanism is an important problem in DPS domain. But, researchers often consider or presume the conditions that one Agent can only join one coalition and one coalition can only execute one task, this presumption will result in the big waste of resources and abilities of Agents. So we presented a new concept of complex coalition, which allows that one Agent can join several coalitions and one coalition can also execute several tasks, and gave the formation mechanism of the complex coalition, which decreased the waste of resources and abilities and increased the system utilities.