Abstract- To reduce the construction complexity of multi-agent system, the proposal is presented that combining several small systems into a larger one, and obtain the properties of the larger system through the properties of small systems. Make reference to dynamic task allocation based on contact protocol, the macroscopical model for task allocation and the microscopical model for agent inner BDI architecture are constructed by gradually refinement, which respectively studies the collaborative relationships among agents and inner mental changes of each agent. Furthermore, according the related conclusions of property preservation by Petri-net-based refinements, the properties of combined system are obtained through analyzing the properties of smaller systems, which outstanding the autonomous and initiative characteristics of agent.
展开▼