The Robot Swarm Re-localization Problem

Nicolas Bredeche (1, 2, 3) , Yann Chevaleyre (4)

1 TANC - Algorithmic number theory for cryptology
2 LRI - Laboratoire de Recherche en Informatique
3 TAO - Machine Learning and Optimisation
4 LAMSADE - Laboratoire d'analyse et modélisation de systèmes pour l'aide à la décision

This paper tackles the problem of re-localization in a swarm of moving virtual robots where some robots might be lost and where the distance between each neighboring robot is known within a limited communication radius (e.g. using the strength of a radio/wifi signal), whereas the precise localization is not known. The basic process in relocalization relies in a robot swarm that re-aggregates based on local information only by recruiting lost robots so as to build a swarm configuration where localization is possible and non-ambiguous. The main issues in relocalization is to provide highly decentralised behaviors for ensuring efficient and fast re-aggregation of the swarm that is scale independent. In this paper, the problem of relocalization is defined as well as criteria for evaluating swarm relocalization efficiency. Moreover, a set of decentralized behaviors based on local reactive behaviors is presented and experimentally studied.

Domaines: Intelligence artificielle [cs.AI]

Nicolas Bredeche, Yann Chevaleyre. The Robot Swarm Re-localization Problem. 5th IEEE International Conference on Robotics and Biomimetics (ROBIO 2008), 2008, Bangkok, Thailand. ⟨inria-00336690⟩