A methodology to improve the cooperative performance of hedonistic multi-agents

Abstract

Multi-agent systems have several performance advantages over monolithic systems, among them are: improved speed performance, spatial distribution, reduction of single point failures, and both diversity and redundancy of capability. Such systems are often difficult to design and manage due to the complexity of coordination. The communications overhead cost to achieve coordination becomes intractable as the number of agents grows large. Yet, cooperative behavior has been shown to emerge from the interactions of simple multi-agents that use self-serving fixed-action patterns. Social insects are the inspiration for this work, which explores the value of constrained communication in simple hedonistic multi-agents for the purpose of improving their summed performance from the total system level view. The objective of this work is to develop a better understanding of the influence of communication in achieving improved cooperative system level performance among self-serving agents. This work includes development of an agent/world model and three experimental studies, the last one of which focuses on the cost/value of communications.