Institut des Systèmes Intelligents
et de Robotique

Cooperative actions exist in a multitude of different forms and are performed by as many different species. Yet their evolution has for a long time represented a puzzle for evolutionary biology. Indeed as the theory goes, a particular behaviour will be selected if it benefits the individual who performs it. However cooperation is, by definition, beneficial to other individuals. In particular, evolutionary biologists have been interested in explainaing the evolution of altruism, where cooperating is costly for the initiator. In comparison, fewer works have focused on the evolution of mutualistic cooperation, which means that all the individuals directly benefit from engaging in a cooperative behaviour. Yet, mutualistic actions often require the evolution of coordination, a particular evolutionary trait which is often ignored by classical theoretical models. We claim the behavioural mechanisms underlying the evolution of coordination are critical in understanding the evolution of cooperation.

 

In order to study this subject, we use evolutionary robotics. This approach simulates the evolution of genotypes and embeddes them in robotic agents to study and evaluate the resulting behaviours. Therefore it allows to investigate the complex mechanisms at play in the evolution of coordination. The goals of this thesis are two-fold: (1) to model the behavioural mechanisms of coordination in the evolution of mutualistic cooperation and (2) to design evolutionary robotics algorithm for the evolution of cooperation to solve complex collective robotics tasks.