Synchronisation Properties of an Agent-Based Animal Behaviour Model

We developed an agent-based model of animal behaviour inspired by the model presented in [6]. The original model was expanded to allow the agents to eat and rest. The decisions for state changes are governed by a deterministic finite-state machine. In order to evaluate the validity of this model, we compared simulation results with real data obtained from video footage of different farms. During the validation process, we used a simulated annealing algorithm to find parameter sets that allow for a best match with the data. Because the agents are situated in a mutual environment, they can influence each others’ behaviour. This effect, combined with the constant change in the agents’ state between resting and standing, motivated the interpretation of the multi-agent system as a combination of weakly coupled oscillators. Using this interpretation, we investigated the possibility of synchronisation between the agents and we developed different order parameters to measure the amount of emerging synchrony. Different biological studies like [9] and [19] show that both growth rate and welfare of beef cows increase when the herd acts in unison. This renders synchronisation as an effect that is highly intended by farmers. We hence developed an algorithm that tries to optimise the layout of feeding and bedding areas in order to find a barn design that favours synchronisation. We successfully fit our model to the data, but incompleteness in the data set did not allow for a conclusive model evaluation. During the simulations, we detected emerging synchrony and increased it using the optimisation algorithm. We concluded that an agent-based model with a small set of rules can describe certain aspects of cattle behaviour and can yield to complex behaviour like synchrony.