Abstract

Evolutionary algorithms have been used as a mechanism for evolving behaviours in both individual robots and swarms over many years. In this talk, I will first show how evolving for functional diversity with respect to behaviour in a swarm can lead to more robust and adaptable swarms. I will then go on to discuss evolutionary approaches to design entire robots – that is, both the body and brain of a robot, in response to a given task.

I will highlight some of the latest research in this area, touching on some of the mechanisms that enable it. I will first show that even when evolving an individual design, such robots can spontaneously demonstrate collective behaviour. I will then turn the question around and ask what can be learnt from a collective over time (in this a case an evolving population) and demonstrate how the information extracted from the collective can be used to improve the design of an individual robot.