Robust Motion Coordination with Covariance Steering Model Predictive Control in Bandwidth Limited Scenarios

The study of motion coordination has highlighted its prevailing advantages in efficiency and deadlock-free. However, ensuring efficient, but tight and safe multi-robot coordination in uncertain and bandwidth limited environments is extremely challenging. In this paper, we propose a robust motion coordination method which encompasses a multi-agent covariance steering model predictive control (MCSMPC) and its underlying initialization strategy. The MCSMPC is presented to directly control the evolution of the trajectory distribution during the planning while providing probabilistic safety guarantees. In particular, probabilistic collision avoidance constraints are tightly formulated to mitigate underlying collisions. To further enhance multi-robot coordination safety, we propose a context-aware initialization strategy that adaptively prioritizes the state initializing urgency in bandwidth-limited scenarios. The strategy considers critical state conditions and surrounding risks to reduce potential collision risks for the most urgent robots. Our coordination method is tested in different environments and achieves significantly tighter coordination, higher efficiency and safety compared to state-of-the-art benchmarks. We also reveal the influence of varying bandwidth resources and their usage on coordination safety.