On Situated Reasoning in Multi-Agent Systems

M. Prokopenko

In this paper we aim to analyse relationships between different approaches to formalising interactivity in dynamic systems. Approaches developed in the framework of Reasoning about Action are mostly logic-based, rely on a centralised world model, and try to (explicitly) capture various aspects of rationality. Another methodology evolved in the field of Multi-agent Systems. It usually considers autonomous agents reacting to changes in external environment and (ideally) exhibiting emergent behaviour. We first attempt to formally define various types of situated agent architectures and encapsulate them in a hierarchical framework. We then analyse and identify domain classes and action theories corresponding to given agent architecture types. This approach can specifically assist in mapping logic theories of actions to reactive agent architectures, where ramifications are embedded in situated behaviours. The described hierarchical framework has been used in the RoboCup Simulation League domain, resulting in implementation of the Cyberoos'98 - a heterogeneous soccer team of autonomous software agents (3rd place winner of the Pacific Rim series at PRICAI-98).

This page is copyrighted by AAAI. All rights reserved. Your use of this site constitutes acceptance of all of AAAI's terms and conditions and privacy policy.