Continuous Approximation Of A Discrete Situated And Reactive Multi-Agent System: Contribution To Agent Parameterization

We propose a formal model for situated and reactive multi-agent systems based on correlated discrete random walks. In order to study this model, we construct a continuous approximation ending up on the Fokker-Planck equation. This result allows us to determine an optimal parameterization for the agents, with respect to the system's objective. Numerical simulations confirm the approach from two points of view, the validity of the continuous model and the optimality of the agents' parameterization.