Decision-making and planning for autonomous vehicles on foggy highway obstacle sections under visual perception

To enhance the safety and decision reliability of automated driving vehicles on foggy highway obstacle sections, this study addresses three key challenges: visual perception degradation, increased uncertainty in obstacle localization, and trajectory instability under high-speed conditions. A decision-making and planning method specifically designed for foggy scenarios is proposed. The method constructs a visual perception degradation model grounded in atmospheric scattering theory, and introduces uncertainty-inflated occupancy sets alongside blind-zone virtual occupancy sets to impose unified constraints on both visible and unobservable regions. An adaptive triggering mechanism is developed based on fog-effective time-to-collision, candidate behaviors are screened using velocity-adaptive terminal-state sampling, and trajectory optimization with collision avoidance control is accomplished via variable-step nonlinear model predictive control (NMPC) framework. The proposed method is validated across three dimensions: varying vehicle speeds, fog intensity levels, and representative obstacle scenarios. Results demonstrate that the method achieves stable, conservative, and executable closed-loop decision-making and planning under perception-limited conditions, simultaneously improving collision avoidance success rate, trajectory smoothness, and state convergence while maintaining real-time performance.