Using a learning mechanism in matchmaking process of agents´ negotiation in multi-agent systems

Negotiation between agents is an important aspect of the multi-agent systems´ efficiency. The purpose of this paper is to enhance the efficiency of the multi-agent systems by improving the negotiation between agents in such a system. Here, focus is on the matchmaking mechanism of negotiation in which a matchmaker agent is responsible to find a proper server agent in response to a request of a client agent. The matchmaker agent needs to have the complete knowledge about other agents and their responsibilities in all system. This paper at first proposed that the agents can be categorized in fuzzy groups by regarding previous experiences of their performance and it happens by using a learning module in system. Also determining membership function of an agent in groups is considered. So, the matchmaker agent needs to have the knowledge about fuzzy groups of agents, and refer the requests in the base of such knowledge. When the matchmaker wants to pass a message about a request, it selects an a-cut group of that fuzzy group and sends the message to them. This mechanism leads to lesser communication and message passing in the huge and or Web based systems, the lesser communication and message passing through the system will lead to more efficient system.