Navegação segregada em enxames de robôs

Abstract

A robotic swarm is a particular type of multi-robot system that employs a large number of simple agents in order to cooperatively perform different types of tasks. In this context, a topic that has received much attention in recent years is the concept of segregation, which exists in several biological systems, and can be useful in many different tasks and scenarios. This concept is important, for example, in tasks that require maintaining robots with similar features or objectives arranged in cohesive groups, while robots with different characteristics remain separate on their own groups. In this paper we propose a decentralized methodology to navigate heterogeneous groups of robots whilst maintaining segregation among different groups. Our approach consists of extending the ORCA (Optimal Reciprocal Collision Avoidance) algorithm with a modified version of the classical flocking behaviors. Numerous trials that were carried out in simulated environments provide statistical insight on the performance of the proposed technique. The results show that the groups remained cohesive during navigation in all evaluated scenarios. Furthermore, the methodology allowed for a faster convergence of the group to the goal when compared to the evaluated algorithms.