Guaranteed Collision Avoidance for Autonomous Vehicles Fusing Model Predictive Control and Data Driven Reachability Analysis

Ensuring collision avoidance is a critical challenge for autonomous vehicles, particularly when faced with uncertain moving obstacles. This work presents a robust collision avoidance framework, integrating data-driven reachability analysis with Model Predictive Control (MPC). The framework is specifically designed to address scenarios where detailed information about the moving obstacles that should be avoided is unavailable. A data-driven approach is employed, which utilizes uncertain measurements corrupted by bounded noise of the obstacle. Based on the measurements, an over-approximation of the reachable sets by moving obstacles represented as zonotopes is constructed. To guarantee security, a safety margin is added to the approximation. The resulting set is employed as a polytopic collision avoidance constraint within the robust MPC scheme, enabling effective control of the autonomous vehicle while guaranteeing avoidance of impacts. The effectiveness of the data-driven collision avoidance scheme is demonstrated through extensive simulations. The presented results outline a promising advancement in collision avoidance for autonomous vehicles operating in uncertain environments.