Research Of Communication On Multi-Agent System Coordination

In recent years, research on agent theory and technology has become one of the main subjects of several disciplines, including artificial intelligence, software engineering, internet etc. Different from traditional AI, research on agent theory prefers system's autonomy, social interaction ability, reactibility and pro-activeness. Agent research mainly prefers to several disciplines, including multi-agent theory, communication and interaction technology, system architecture and organization form, agent-oriented design and programming, multi-agent coordination,cooperation and negotiation etc. Communication is the foundation and important way of negotiation and cooperation among agents. Without communication, the result of negotiation and cooperation will not be good. To achieve coordination among agents, there must be effective communication and interaction between agents. Communication Model among agents pay much more attention to the realization of communication transmit protocol, and ignore the foundation of agent on the knowledge level. Combining the ontology theory of agent, we discuss the importance of the ontology on agent communication model, and provide the idea of building agent knowledge base in construction level, which is on communication level of agent. Comparing to object, agent has some different characters including pro-activeness, objective, object-oriented behavior, self-learning, and self-adaption. These characters determine that communication mode among agents is different from message passing between objects. Combining with the BDI theory based individual rational behaviors, provide the belief library, object library and intention library in construction level, provide the communication rules, communication intention and the ontology in communication level, and then get BDI-based multi-agent communication model. In MAS coordination communication, we must care about the expression of coordination knowledge in addition to the knowledge of agent. Coordinative knowledge may get the general character through summarizing the commonness of coordinative problems. Based the TAXI coordination problems, we particular describe this coordination problem, provide agent state transition, communication on knowledge level and reasoning mechanism combining with BDI-based multi-agent communication model.