Collective behavior generation of evolutionary robotic swarm in the pursuit problem

Abstract

The pursuit problem in the continuous torus domain is discussed in this paper. It is assumed that there are three preys and twenty predators on the torus surface. The task for predators is to capture all the preys that are getting away from them by surrounding each not to move at the position for a sufficient time length. Predators are assumed to be a homogenous robotic swarm with evolving artificial neural networks (EANNs). A series of computer simulations is conducted to compare three types of EANNs to find the efficient artificial evolution of the robot controllers.