Co-evolving semi-competitive interactions of sheepdog herding behaviors utilizing a simple rule-based multi agent framework

Sheepdog herding behaviors demonstrate an interesting form of interactions between two classes of agents - sheep and the dog. The nature of the interactions between sheep and the dog takes a special form of competition which is different to the traditional prey-predator interactions where the success of prey depends on the failure of the predator and vice versa. In consequent, the development of an appropriate objective function to efficiently co-evolve successful sheepdog herding behaviors becomes challenging. This paper presents a framework to efficiently co-evolve sheepdog herding behaviors utilizing the simple rule based agent approach in order to derive high fidelity behavior dynamics and discusses the challenges involved in the process.