Dynamic-Weighted Model Predictive Control for Connected Automated Vehicles to Enhance Car-Following Safety on Foggy Freeways

Abstract Low visibility on foggy freeways impairs drivers’ perception and judgment, increasing risks of rear-end crashes. Connected automated vehicles (CAVs), supported by advanced sensors and communication systems, are less affected by visibility reductions and thus offer considerable potential to improve car-following safety in fog. This study proposes a CAV car-following strategy that integrates fuzzy control into a dynamic model predictive control (MPC) framework. By adjusting dynamic MPC weights, the strategy adapts to changing traffic conditions and satisfies the Lyapunov stability criterion, ensuring car-following’s robustness on foggy freeways. To evaluate the strategy’s performance, simulations are conducted under light and dense fog, with varying CAV penetration rates and spatial distribution patterns. Results show that car-following risks, measured by surrogate safety indicators, decline as CAV penetration rate increases. Compared with a pure human-driven vehicle fleet, a pure CAV fleet operating under the proposed strategy reduces risks by up to 62.85% (inverse time-to-collision) and 57.38% (deceleration rate to avoid crash) in light fog and by 73.33% and 64.93% in dense fog. Moreover, spatial distribution of CAVs in mixed fleets substantially affects safety improvement. When CAVs equipped with the proposed strategy are positioned at the front, they suppress disturbance propagation and enhance safety even at low penetration rates, whereas improvements are only significant above the 80% penetration rate when CAVs are placed at the rear. Overall, the proposed strategy proves more effective in dense fog and when CAVs are deployed at the front of mixed fleets, providing practical insights for improving traffic safety management under fog conditions.