A Multi-agent Negotiation Protocol Based on Extended Case Based Reasoning

This paper describes a new Multi-Agent architecture and a task allocation protocol proposed inter-agent cooperation in a multi-agent system. The negotiation protocol is contract net protocol (CNP). The broadcast of the Task Announcement on CNP tends to consume much communication load and its performance degrades drastically when the number of communicating agents and the number of tasks announced increase. In order to overcome this problem, the CNP based on extended case based reasoning (ECBR) is presented, it can reduce the scope of competitors and communication load on task negotiation, make a distinct improvement to the efficiency. To ensure a better cooperation between agents and adapt to higher flexibility of system, Several parameters such as forget coefficienttime coefficient and a special case are defined in extended CBR. Experimental results show the performance of this negotiation protocol based on extended CBR improves significantly.