This work proposes a robust control strategy for an autonomous vehicle to overtake safely and comfortably a human-driven vehicle. The proposed scheme designs a collisionavoidance constraint setup that comprehensively coordinates dimension-based and velocity-dependent constraints to fulfil the safety criteria. A three-phase control framework is proposed for the overtaking task, subject to separate collision-avoidance constraints in lane changing, passing, and merging phases. Moreover, the proposed method utilises a Stackelberg game model to interactively involve the human-driven overtaken vehicle behaviours in the online optimisation loop. To further cope with uncertainties caused by the human driver, the optimisation is solved by a robust model predictive controller to guarantee the avoidance of collisions. Numerical case studies verify that the proposed framework is capable of overtaking not only a cooperative human-driven vehicle but also an uncooperative human-driven vehicle with safe and comfortable trajectories.
Game-Theoretic Model Predictive Control for Safety-Assured Autonomous Vehicle Overtaking in Mixed-Autonomy Environment
2024-07-01
In: Proceedings of the 22nd European Control Conference. Institute of Electrical and Electronics Engineers (IEEE) (2024) (In press).
Paper
Elektronische Ressource
Englisch
DDC: | 629 |
Model Predictive Control for Full Autonomous Vehicle Overtaking
Transportation Research Record | 2023
|