Evolutionary robotics: a tool to understand and design collective behaviours in collective adaptive systems - School of Computing Seminar Series

Nicolas Bredeche, ISIR, UPMC, Sorbonnes Université (France)

Collective adaptive systems are ubiquitous: from the many examples that can be observed in nature to artificial systems such as collective robotics or self-adaptive distributed algorithms. In this talk, I will describe two of our recent works on adaptive mechanisms in collective systems. Firstly, I will provide some insights related to the evolution of mutualistic cooperation in natural systems, using evolutionary robotics as a methodological tool to extend game theoretical models of cooperative hunting (work published in Plos Computational Biology). Secondly, I will address the problem of on-line distributed learning of behavioural specialisation in a robot swarm, for which evolutionary robotics stands as a promising, but currently limited, design method (work published in Frontiers in AI and Robotics). I will also briefly describe some other works related to the possible application of collective adaptive artificial systems, including the design of a micro-scale robot swarm.