Cooperative peer-to-peer multiagent based systems

TL;DR

This paper introduces a multiagent system model based on Verhulst-Lotka-Volterra equations, demonstrating how cooperation among agents leads to self-organized clusters and enhanced capacities beyond non-interactive limits.

Contribution

It presents a novel peer-to-peer interaction scheme with rigorous analysis of stable configurations and demonstrates capacity improvements through cooperation.

Findings

Agents self-organize into clusters.

Multiple stable configurations can coexist.

Cooperation allows agents to surpass capacity limits.

Abstract

A multiagent based model for a system of cooperative agents aiming at growth is proposed. This is based on a set of generalized Verhulst-Lotka-Volterra differential equations. In this study, strong cooperation is allowed among agents having similar sizes, and weak cooperation if agent have markedly different "sizes", thus establishing a peer-to-peer modulated interaction scheme. A rigorous analysis of the stable configurations is presented first examining the fixed points of the system, next determining their stability as a function of the model parameters. It is found that the agents are self-organizing into clusters. Furthermore, it is demonstrated that, depending on parameter values, multiple stable configurations can coexist. It occurs that only one of them always emerges with probability close to one, because its associated attractor dominates over the rest. This is shown through numerical integrations and simulations,after analytic developments. In contrast to the competitive case, agents are able to increase their capacity beyond the no-interaction case limit. In other words, when some collaborative partnership among a relatively small number of partners takes place, all agents act in good faith prioritizing the common good, whence receiving a mutual benefit allowing them to surpass their capacity.