Guiding Herds Through Transhumance Corridors: Harnessing Reinforcement Learning With Vicsek Model

This paper introduces a new approach based on reinforcement learning, which consists of confining livestock to transhumance corridors during seasonal migrations. A model is proposed that combines the Bellman equation with the Vicsek model and the conditions for keeping cattle in the corridor. The reinforcement learning algorithm is used to define a model that meets the requirements stipulated in this paper. The model is implemented and tested by varying several parameters. The model is trained on the shepherd. It is followed by the animals through interactions between them. This training evolved progressively through two essential stages, namely the shepherd’s exploration of the corridor and the shepherd’s autonomy in exploring the environment. The results obtained lay the foundations for moving towards implementation in real-life conditions. By confining animals and shepherds to virtual corridors during transhumance, conflicts between farmers and herders due to declining resources and population growth could be avoided.