Speaker: Francesco Borra

Title: Optimal collision avoidance in swarms of active Brownian particles

Abstract:
Collision avoidance is a common goal in collective behaviors of animals and robots. We consider a system of active Brownian particles in two dimensions and we frame this objective as an optimization problem with a tradeoff between avoiding collisions and minimizing control costs. We employ optimal control theory and a mean-field game approach to derive an analytic solution which displays a critical behavior with a second order discontinuity in the alignment order parameter. Remarkably, we find that a mean-field version of a standard phenomenological model for collective motion performs remarkably close to the optimal control. Our results offer a theoretical ground for the use of simple biomimetic algorithms for artificial agents.