Distributed Sampling-Based Model Predictive Control via Belief Propagation for Multi-Robot Formation Navigation

Abstract

Sampling-based stochastic optimal control has become an appealing robotic control framework due to its ability to handle complex and general forms of dynamics models and task specifications. Although sampling-based methods have been shown successful in a variety of single-robot control tasks, studies on their extension to multi-robot problems are limited. In this letter, we propose a distributed framework for sampling-based optimal control. The framework formulates multi-robot optimal control as probabilistic inference over graphical models and leverages belief propagation to achieve inference via distributed computation. We developed a distributed sampling-based model predictive control (MPC) algorithm based on the proposed framework, which obtains optimal controls via variational inference. The algorithm was validated in a multi-robot formation navigation problem. The simulation results show the efficacy of our proposed method with improved control performance over a gradient-based distributed MPC algorithm.