MULTI-AGENT SWARM BASED LOCALIZATION OF HAZARDOUS EVENTS

Maja Varga, Zvonimir Piskovic, Stjepan Bogdan, and Zhijun Li

Keywords

Area exploration, swarms, correlated random walk, multiagent systems

Abstract

Swarm robotic concept has become a topic of extensive research in hazardous environments where fault tolerant, robust and energy-efficient approach is needed. In this paper we present behavioural based algorithm for outdoor environment exploration. This work is inspired by ant colony behaviour models. Our goal is to control agent’s motion using simple rules and not to rely on wireless communication. Agents perform random walk combined with simple heuristic pheromone based rules which prevent frequent visits of previously visited locations. To determine heuristic parameters which maximize area coverage, we optimize parameters by simulating swarm behaviour in Matlab. Furthermore, we take into account the agent limitations, such as battery lifetime, and we model a probability success of a mission with respect to that parameter.

Important Links:



Go Back