| Researches on Intelligent agents and multi-agent systems have been an active field in artificial intelligence society even in mainstream computer science in recent years. Some famous agent models and systems have been proposed. Due to the variety of complex application problems, the models do not always provide satisfying solutions. Such is the case for e-business applications, cruise missile control, air-battle management, network management, etc. To provide better supports to the applications, in this thesis, an intelligent agent model, IDEAL (I&barbelow;ntegrated D&barbelow;eliberative and rE&barbelow;active A&barbelow;rchitectures with L&barbelow;earning abilities) is proposed. An IDEAL agent is horizontal hierarchical hybrid architecture. It is partly based on BDI paradigm, thus making use of the three agent's primary mental attitudes (beliefs, desires, and intentions) as data structures, but also uses reactive patterns to deal with unpredictable events or events requiring fast response times. Two effective learning mechanisms are integrated into the architecture. Not only do they bridge the gaps between reactive and deliberative paradigms thus enhancing reaction speed for processing complex problems, but they also improve the performance of the system through optimization of the agent's behaviors. A specially designed dynamic planner with pseudo-parallel processing abilities is used to improve real time planning abilities. Qualitative comparisons of an IDEAL agent and some other agent models are presented.; On the basis of the IDEAL agent model, a multi-agents frame with a hierarchical broadcast frame is proposed. We show that it can be used to simplify and improve the design and development of intelligent agent applications. A detailed application example is given. |
