Exploiting Non-Dominance in Multi Agent Systems: An Artificial Immune Algorithm for Distributed and Complex Problem Solving Environments

Abstract

An ideal Multi Agent System is flat and has no dominant hierarchy. Multi agent computational and problem solving environments have been advocated for their ability to deliver overall solutions that are innovative and creative. There is however a significant threat to the coherence of Multi Agent Systems; chaos. This paper poses a new vision to the control and immunisation of the Multi Agent Systems against chaos. Employing a complexity measure of the problem and its lower and upper bounds, and monitoring the complexity of the problem solving agents’ interactions, we propose the holistic control of the Multi Agent Systems that leads to immunisation of the system against chaos. The control however is not central and appears in the form of the agents’ common knowledge and determines their tendency to proactively communicate.