This study proposes a risk-predictive path planning and control framework for autonomous driving (AD) and advanced driver-assistance systems (ADAS) operating on narrow urban roads with limited visibility. When driving through unsignalized intersections, buildings or parked vehicles often create blind spots where pedestrians or other road users may suddenly dart out, becoming a significant safety risk. To address this challenge, this research defines a method to simultaneously determine the safe speed and safe lateral distance based on the geometric relationship between the ego vehicle and the occluded area. The safe speed is defined as the maximum velocity at which a collision can be avoided through emergency braking, while the lateral gap ensures avoidance of side-faced collisions. Using these risk-predictive parameters, a non-linear polynomial path generation method is formulated, in which cubic and quartic functions describe the vehicle’s speed profile and trajectory, respectively. The coefficients of these polynomials are optimized to minimize deviations from desired paths while preventing the vehicle from entering the predicted high-risk zones by the proposed method. Comprehensive MATLAB simulations were performed using a 3-DoF vehicle model to validate the proposed method. Results show that the framework successfully generated smooth and safe trajectories, allowing the vehicle to avoid potential collisions under various road widths and darting-out conditions. Even in constrained road scenarios, the system maintained stability and avoided frontal collisions. The proposed method effectively balances safety and mobility, demonstrating strong adaptability for future integration into real-world AD/ADAS systems.
Fujinami et al. (Thu,) studied this question.