Research on single-layer MPC control for autonomous vehicles based on a safety-first strategy in signalized mixed traffic environments

Abstract

To address frequent start-stop fluctuations and safety conflicts in mixed traffic at signalized intersections, this paper proposes a single-layer Model Predictive Control (MPC) framework based on a safety-priority strategy to optimize the longitudinal trajectories of autonomous vehicles (AVs) in non-steady-state flow. By integrating a "risk boundary wall" and hard safety constraints within a 10s prediction horizon, the model achieves a multi-objective trade-off between safety, efficiency, and comfort; simulation results demonstrate that under 100% AV penetration, the minimum Time-to-Collision (TTC) increases from 0.56s to 1.78s, establishing a robust physical safety boundary while reaching a Pareto equilibrium between safety-induced delays and traffic flow smoothness.