Towards optimization-based multi-agent collision-avoidance under continuous stochastic dynamics

Abstract

In our ongoing work, we aim to control a team of agents so as to achieve a prescribed goal state while being confident that collisions with other agents are avoided. Each agent is associated with a feedback controlled plant, whose continuous state trajectories follow some stochastic differential dynamics. To this end we describe a collision-detection module based on a distribution-independent probabilistic bound and employ a fixed priority method to resolve collisions. Due to their practical importance, multi-agent collision avoidance and control have been extensively studied across different communities including AI, robotics and control. However, these works typically assume linear and discrete dynamic models; by contrast, our work intends to overcome these limitations and to present solutions for continuous state space. While our current experiments were conducted with linear stochastic differential equation (SDE) models with state-independent noise (yielding Gaussian processes) we believe that our approach could also be applicable to nonGaussian cases with state-dependent uncertainties.