Integrating Agent Interaction into a Planner-Reactor Architecture*

摘要

This paper deals with the development of methodologies and architectures for autonomous systems that act and interact in a shared, dynamic environment. We present the agent architecture INTERRAP 1. INTERRAP extends previous approaches to building layered architectures by a cooperation layer, and thus allows the designer of a system not only to build agents with both reactive and deliberative facilities, but also to provide important mechanisms for communication, coordination, and collaboration with other agents. The components of the architecture and their interplay are described. It is evaluated by the simulation of an automated loading dock, an interacting robots application, where autonomous forklifts have to perform loading and unloading tasks. We show how different functionalities of the forklift agents can be modeled in an adequate manner at different layers of the INTERRAP architecture. Thus, the main contribution of this work is the reconciliation of (D)AI concepts such as goals, plans, conflict resolution, and negotiation with mobile robot research which is absolutely necessary in order to model the lower layers of the agent, such as sensor data processing, obstacle avoidance, navigation, and physical control.