Distributed Model Predictive Control of Connected Multi-Vehicle Systems at Unsignalized Intersections

Qiang Ge,Guillaume Sartoretti,Jingliang Duan,Shengbo Eben Li,Yuming Yin,Sifa Zheng

DOI: https://doi.org/10.1109/icus55513.2022.9986954

2022-01-01

Abstract:This paper proposes a four-component framework for connected vehicles in general scenarios, which consists of the following: 1) an information flow topology; 2) cooperative controllers; 3) node dynamics; and 4) formation geometry, extending original concepts from vehicular platoon control. In this framework, we present a real-time planning and distributed cooperative control method, which is scenario-agnostic and scalable to larger groups of vehicles. Specifically, we first develop a real-time receding-horizon planning approach that improves plan consistency by reasoning about and reusing the previous trajectory. Our distributed model predictive controllers track the resulting reference trajectories and avoid collisions by allowing neighboring vehicles to exchange their intentions. In contrast to many existing methods, our approach to decentralized cooperation does not need to assume fixed paths for vehicles or assign any priority among them. We present simulation results that apply to up to eight heterogeneous vehicles, showing how our framework allows agents to establish efficient interactions to simultaneously traverse an unsignalized intersection smoothly and safely. Video results are also available at https://youtu.be/Q3KjKuIquAo and https://youtu.be/x1zS4ynaW-s.