Research On Policy-Driven And Behavior Model Of Agent

Under the guidance of Agent technology and policy, this thesis establishes a novel behavior model of the Agent. The principle and method are discussed in three aspects: single Agent's behavior model, Multi-Agent's cooperation model based on policy-driven and the architecture of Multi-Agent based on policy-driven. The main contents include the followings:Firstly, from the angle of autonomy and controllability, this thesis proposes the PGN-AO model. This model applies the idea of three-tier policy which plays the role of guiding the behavior of Agent and reflects the feature of driven. By focusing on the these characteristics of the policy, this thesis introduces five modal operators such as PLAN, OBL, ABILITY, GOAL, NORM into the decision-making process of Agent and explain the PGN-AO logic's syntax, semantics and axiomatic system.Secondly, on the basis of MPGN-AO model, this thesis proposes a cooperation mechanism for Multi-Agent system. The cooperation mechanism can be divided into two parts: Multi-Agent's behavioral model and the process of collaborative problem solving. In the aspects of Multi-Agent's behavioral model, we establish a novel semantic model for Multi-Agent system where under the guidance of policy, the decision-making process of multi-Agent has been formed. In the aspects of collaborative problem solving, we extend the cooperation model of Wooldridge and Jennings and describe the whole cooperation process of Multi-Agent based on the model of MPGN-AO.Finally, in order to improve the controllability of Multi-Agent, this thesis proposes a novel Architecture which is driven by policy. This architecture includes four layers: the policy control layer, the self-control and behavior constraint layer, the organization management layer and the middle connection layer. In the policy control layer, we design two kinds of structures and a set of behavior rules which can be considered as the upper commands of the self-control and behavior constraint layer. In the self-control and behavior constraint layer, we implement policy driven autonomic Agent model to map regulation into Agent's behavior, and finally the middle connection layer is used for connecting the above two layers.