The agent architecture InteRRaP : concept and application

Abstract

One of the basic questions of research in Distributed Artificial Intelligence (DAI) is how agents have to be structured and organized, and what functionalities they need in order to be able to act and to interact in a dynamic environment. To cope with this question is the purpose of models and architectures for autonomous and intelligent agents. In the first part of this report, InteRRaP, an agent architecture for multi-agent systems is presented. The basic idea is to combine the use of patterns of behaviour with planning facilities in order to be able to exploit the advantages both of the reactive, behaviour-based and of the deliberate, plan-based paradigm. Patterns of behaviour allow an agent to react flexibly to changes in its environment. What is considered necessary for the performance of more sophisticated tasks is the ability of devising plans deliberately. A further important feature of the model is that it explicitly represents knowledge and strategies for cooperation. This makes it suitable for describing high-level interaction among autonomous agents. In the second part of the report, the loading-dock domain is presented, which has been the first application the InteRRaP agent model has been tested with. An automated loading-dock is described where the agent society consists of forklifts which have to load and unload trucks in a shared, dynamic environment.