Hierarchical MPC-based safe shared control for forward collision avoidance in collaborative driving

Human–machine shared control plays a vital role in enhancing driving safety through collision avoidance assistance. Existing collaborative controllers typically integrate risk assessment and control computation within one Model Predictive Control (MPC) framework, which can lead to computational challenges when incorporating essential nonlinear vehicle dynamics for accurate trajectory evaluation. To address this issue, we propose a novel safe shared control scheme with two key components: First, a trajectory generator produces multiple smooth, collision-free candidate trajectories considering both obstacle avoidance and vehicle stability constraints. Second, a hierarchical MPC module evaluates and executes these trajectories through a dual-layer structure. The upper layer uses a linear model to compute control inputs and then evaluates them with a nonlinear vehicle dynamics model for risk assessment, while the lower layer calculates the actual control commands based on a simplified linear model by tracking the selected optimal trajectory. This separation of trajectory generation, risk evaluation, and control computation significantly enhances both computational efficiency and safety assurance. A series of collision avoidance experiments was conducted on a driving simulator to evaluate the proposed method. Results show that the proposed method significantly enhances safety in human–machine shared control driving.