Multi-agent coordination mechanism in distributed environment

This paper presents how to make different agents cooperate to accomplish complex tasks and coordinate their objects, planning, knowledge and action, etc in dynamic and open multi-agent system. Using a hierarchical architecture, we implement a multi-agent prototype system-NHMAS in distributed network system environment. After briefly describing the framework of different agents and their functions, the paper discusses coordination mechanism of NHMAS in detail. We explore a heuristic strategy in task allocation. An advantage of this task allocation strategy is that it does not put extra loads on the system at critical times. For the need of coordinating different agent actions, we add partial order constraints between actions that have probability of conflicts and introduce our planning algorithm, conflict resolution and communication language. At last, we give a briefly performance analysis on Aglet workbench for NHMAS.