The Researches On Software Architecture And Implementation Technology Of Policy-based Self-Organizing Systems

With the rapid development of information technologies, the distributed softwaresystems (e.g. the wireless sensor network systems and the P2P systems) become moreand more large and complicated with characteristics like autonomy of individuals,decentralized control, local interactions, etc. Under this background, many researchersfrom academy and industry pay attention to the researches of the self-organizingsystems, and some research efforts have been made such as the self-organizingmechanisms, system models and development methodologies. However, because of thecomplexness of self-organiazing system, the development of such systems still facessome challenges:1) How to model and abstract the self-organizing system?2) How todesign the appropriate interactions among individuals to satisfy the global macroemergence?3) What technologies should be provided to support the construction of theself-organizing systems?To tackle these problems, this thesis takes the entities of the self-organizing systemas agents and proposes a policy-based self-organizing method for developingmulti-agent system based on the concept of policy in the human society. In our approach,we use policies to regulate agent’s behaviors so as to obtain the expected emergentresults in an effective way. The key issue of the constructing such system is how toimplement the software agent of the systems, which not only can perceive andunderstand system’s policy, but also can automatically decide and adjust its behaviorsbased on the policy. The main contributions of this paper are described as follows:1) We propose a BDIP(Belief-Desire-Intention-Policy) software agent architecturefor the PSOMAS (Policy-based Self-Organizing Multi-Agent System), and thecorresponding behavior decision algorithms of the BDIP agent;2) We design and implement the runtime mechanism and the runtime model ofBDIP agent based on the extended BDI agent runtime mechanism.We alsoprovide an development model of BDIP agent and policy, and implement thescheme of the development model with XML data binding techonology, whichcould help the developers develop the agents and the policies of the PSOMAS;3) We implement a development platform called PSOMASDE (PSOMASDevelopment Environment). This platform can support the running of the self- organizing system as well as the development of BDIP agent and policy. It alsocan support the dynamic update of the policies at the run-time.