Research On Autonomic Collaboration And Policy Oriented Trusted Model Of Agents

With the gradually increasing demand for resource-shared computing and collaborative services in distributed environment, which requires software development, deployment, operation and maintenance in network environment which is the open, dynamic and difficult to control. Agent, with reactivity, autonomy and sociality characteristics, is widely considered to be a key technology for large-scale, open and distributed information systems to achieve dynamic services integration and collaborating work. But the traditional agent model, which is difficult to predict and control the agent’s behavior, lack of assessment of credibility in individual and their behavior and lack of conflict resolution mechanisms, is difficult to ensure the credibility of the system and agent’s behavior.For these problems, this thesis proposes a new solution, which uses open distributed system environment as its background, combines with behavior control and inspiration, and finally uses the extended defeasible logical framework to establish the credibility of model theory in agent individual and group collaboration. The main work is as follows:1. A policy-oriented individual model which is based on credibility. In the traditional rational model based on the BDI, combining with the policy-oriented role to constrain and regulate the behavior of Agent from the macro and micro levels and enhance outside effect for decision-making, so as to make the Agent autonomic and be more controllable. In decision-making process, using credibility mechanism for making Agent take full account of direct trust and reputation factors in order to improve the correct of decision and ensure the model’s reliability.2. A policy-oriented collaborative credibility framework. Unified the two aspects of agent trust behavioral control and behavior inspiration, and established appropriate mechanisms to ensure the system is credible. The framework contains four sub-processes:coalition statuses updating, generating coalition tasks, generating individual tasks and implementing of individual behavior. Four sub-processes are constrained by policies and affected from the credibility center so as to make Agents collaborative process more credible.3. A formal framework based on the extended defeasible logical, which make the established model can dynamically accept the changing of rules, deal with real-time conflicts of rules flexibly and can be non-monotonic reasoning in highly efficiency. Successful introduction of reliable regular text so that extended defeasible logical can be more reasonable description of the rules of credibility and provides a theoretical basis for the introduction of credibility mechanism in Agent models.