Research On Policy-based Self-organizing Software Systems And Implementation&Running Technologies

Self-organization is a process that system achieves its Goals and obtains theexpected global behaviors through local interactions among the entities in the systemwithout external interference and centralized control. Self-organizing system that hasthe characteristic of self-organization is generally composed of autonomous entities,located in open environment and has local interaction at the micro-level and emergencyat the macro level. Recently, self-organization technologies have been activelyinvestigated and widely utilized in several areas such as multiple robots, P2P systemdesign, big data storage, etc. Many progresses have been made in this area, like systemmodel, running mechanism, development method. However, it is still a great challengeto propose general-purpose methods to support the construction of self-organizingsystems and deal with the changes of the situated environment and self-organizationrequirementsTo deal with the problems, the thesis borrow the ideas from sociology, regardsself-organizing system as multi-agent system, and proposes an approach that utilizes thepolicy as the abstraction of self-organizing mechanism. The contributions of the thesisare summarized as the follows:1. A Policy-based Self-organizing Mutli-Agent Systems (PSOMAS) model isproposed by inspiring the principles of sociology, and the policies are taken asthe abstraction of the self-organizing mechanisms of the systems. In the model,policy direct and constrain the interaction behaviors among agents to realizeself-organization. Moreover, policies as self-organizing mechanism can beflexibly adjusted at run-time in order to adapt to the changes of changes ofself-organizing requirements and the situated environment, such approachprovides a high-level model that is independent of specific self-organizingtechnologies and platforms to investigate the self-organizing mechanism andits update.2. A BDIP agent architecture is proposed and the behavior decision algorithmsare designed to support the construction of the software entities in PSOMAS.BDIP agent can behave according to the regulations and constraints of thepolicies in the systems in order to obtain the expected macro behaviors of thesystems.3. The architecture of PSOMAS including policy management, agent re gistration,policy update and synchronization mechanisms, etc., are proposed. In order todesign of software agents and policies in PSOMAS, a design specificationlanguage PSOSSL has been proposed. Moreover, an agenda-based mechanismof BDIP agent is designed in support of the running of BDIP agent. 4. We have developed an environment called PSOMASDE to support thedevelopment and running of PSOMAS. It consists of XML-based designspecification of agent and policy, interpretation of policy, management systemof policy, and running environment of BDIP agent. Developers can adjust thepolicies of the systems online without stopping the running of the systems.5. In orde to support the systematic development of PSOMAS, a method fordeveloping PSOMAS is proposed. A sample case is studied to show thefeasibility and effectiveness of the proposed approach and environment.