Evolution of collective behaviors for a real swarm of aquatic surface robots

Swarm robotics is a promising approach for the coordination of large numbers of robots.While previous studies have shown that evolutionary robotics techniques can be applied toobtain robust and efficient self-organized behaviors for robot swarms, most studies havebeen conducted in simulation, and the few that have been conducted on real robots havebeen confined to laboratory environments. In this paper, we demonstrate for the first time aswarm robotics system with evolved control successfully operating in a real and uncon-trolled environment. We evolve neural network-based controllers in simulation for canonicalswarm robotics tasks, namely homing, dispersion, clustering, and monitoring. We thenassess the performance of the controllers on a real swarm of up to ten aquatic surfacerobots. Our results show that the evolved controllers transfer successfully to real robots andachieve a performance similar to the performance obtained in simulation. We validate thatthe evolved controllers display key properties of swarm intelligence-based control, namelyscalability, flexibility, and robustness on the real swarm. We conclude with a proof-of-con-cept experiment in which the swarm performs a complete environmental monitoring task bycombining multiple evolved controllers.