Multi-agent paradigms and connectionist models of associated architectures in robotics

Abstract

The objective of the paper is to describe some of the concepts that have emerged over the last few years in the study of colonies of interacting robots. Specifically, the multi-agent paradigm is explained and its application to such systems is considered. The concept of cooperation is explored and a holistic definition is introduced. The tropism based cognitive architecture and a behavior-based architecture and associated learning methodologies are described. An attempt is made to describe the possible use of a connectionist structure utilizing a Kohonen network to model the learning and decision making in these architectures. A rule to model 'forgetting' is also suggested.