Research On Cooperation Network Model In Multi Agents System

Multi Agent system (MAS) has many advantages, such as robust, efficient,adaptive, extendible, wide application, reducing software complexity and so on, so itis used in the economics, biology, chemistry, physics, sociology, and many otherfields. In the research of multi agents technology in MAS, the research of cooperationtechnology among multi agents is the most hot and promising. The whole system’sperformance and efficiency of accomplishing tasks depend on the quality ofcooperation among agents. Although single agent’s ability is limited because of itssimple behavior, weak cognitive ability, limited perceptive area, and so on, a largenumber of agents interact with each other through communication, coordination,cooperation can make the system emerge intelligent behavior, and the system’sintelligent behavior can quickly and efficiently complete those very complex tasks orhardly computing problem, which are difficult or impossible to complete by usingtraditional methods. Therefore, the research of multi agents’ cooperation is not only ofgreat theoretical significance, but also can promote human society greatly.In MAS, the interaction among agents decides whether the system couldaccomplish task or solve problem. The interaction among agents has two types, one isthe direct interaction among agents, and the other interaction is between agents andenvironment, which can help agents interact indirectly with other agents. Because ofthe emergency of system’s intelligent behavior is from gathering vast nonlinearinteract among agents, so this paper mainly research on the nonlinear interactionsamong agents. For the research of direct interaction among agents, we construct a newagent model, propose a partner selection mechanism based on trust and credit,describe the mechanism of update of trust and credit, discuss how to communicatewith others effectively and propose the communication mechanism similar tobillboards and phone book. For the interaction between agents and environment, thispaper proposes an adaptive cooperation network optimize algorithm based onpheromone (PCNAO). According to the information about resources and tasks incurrent environment, agents can secrete digital pheromone (including the activatorfactor and inhibitor factor), which can help agents select better behavior and make thesystem optimize adaptively its cooperation network to improve utilization ofresources and assign tasks more reasonable. So the agents can adapt currentenvironment well and improve system’s efficiency and performance. Finally, based onthe swarm simulation platform, this paper proved the cooperation model and theadaptive cooperation network optimize algorithm is viable and effective.