Research On Collaboration And Task Allocation Of Self-interested Agents In Coalitional Skill Games

Multiple self-interested agent collaboration is widely used in engineering and management.Coalitional skill game is a typical multiple self-interested agent system.The studies of coalitional skill games can provide theoretical and methodological support for the management and collaboration of self-interested agents.The main research work and results are summarized as follows:(1)This dissertation presents a general model of the coalitional skill game.According to factors that affect the coalition value,the coalitional skill game models are divided into characteristic function coalitional skill games and partition function coalitional skill games.The differences and connections between the two coalitional skill games are studied.(2)For the task allocation problem of self-interested agent coalitional skill game,three algorithms,named task allocation algorithm based on simple emotional states,intermediary recruitment algorithm,and task allocation algorithm based on utility distribution adjustment,were proposed.The three algorithms can ensure higher system total revenue under the premise of guaranteeing the self-interest of the agents.At the same time,the three algoriths can provide decision supports for self-interested service agents,skill agents and task agents respectively.(3)When the service agent and the task agent in the coalitional skill game are in different geographical positions and have different scope of vision,the task selection strategy of the self-interested service agent and the utility distribution strategy of the self-interested task agent are studied.The influence of the scope of vision of self-interested agent on system revenue and individual revenue are also analyzed.(4)For the multiple self-interested agent coalitional skill game,the question of how to add new service agents to an existing coalitional skill game or add new skills to an existing service agent is examined.A dynamic programming algorithm was designed to solve the problem of "computing the maximum system total revenue" of the multiple self-interested agent coalition skill game.Three methods for solving the problem are proposed,namely,the maximum system revenue based method,the maximum average contribution based algorithm,and the least ratio of supply and demand based method.