Multi-agent coordination through mutualistic interactions

Abstract

In this paper we present an ecologically-inspired approach to agent coordination. Mutualistic networks of interacting species in nature possess characteristics that provide the systems they represent with features of stability, minimised competition, and increased biodiversity. We take inspiration from some of the ecological mechanisms that operate at the interaction level in mutualistic interactions, and which are believed to be responsible for the emergence of these system level patterns, in order to promote this structural organisation in networks of interacting agents, enhancing in this way their cooperative abilities. We demonstrate that given plausible starting conditions, we can expect mutualistic features to appear in self-organising agent systems, and we compare them with natural ones to show how the characteristics displayed by ecologically inspired networks of agents are similar to those found in natural communities. We argue that the presence of these patterns in agent interaction networks confer these systems with properties similar to those found in mutualistic communities found in the real world.