A Multi-Layered Control Architecture of Intelligent Agent

Abstract

Usually control architecture of an intelligent Agent is structured as a single level, all related control issues are determined in this level. Since the Agent may work in a complex organizational context and it may have multiple complicated interdependent tasks to process. Obviously, it is difficult to construct an integrated control framework in which all the control issues are determined in an efficient way. In this paper, a three-layered control architecture for an intelligent Agent is presented. The control Agent architecture consists of three layers: global control layer, detailed control layer and execution layer, the first two control layers are decision making layers. Each control layer in this control architecture has different function in control process: global control layer deals with the formation of high level goals and objectives for the Agent. Detailed control layer deals with feasibility and implementation of operations. The execution layer executes those actions determined by the upper control layers. Furthermore we use TAEMS (task analysis environment modeling and simulation) to represent structures of those tasks that agent has to execute. The primary advantage of this three-layered control Agent architecture is that all the control issues can be addressed concurrently and done in an efficient way.