A Passing-Sight-Distance-Based Overtaking Assistance System Using Computer Vision and GNSS, with a Simulated Extension to Three-Vehicle Maneuver Feasibility

In this research paper, SIASS, an intelligent decision-support system for overtaking assistance on two-way rural roads, is presented that combines YOLOv8n vehicle detection, monocular ranging, optical flow, and GNSS. The proposed framework is a blend of learned visual perception with interpretable decision rules based on the passing-sight-distance requirement from the Peru's DG-2018. Oncoming-vehicle distance, oncoming-vehicle time-to-collision constraints and three-vehicle maneuver feasibility are evaluated using the physical prototype and Monte Carlo simulation. Across 7,897 synthetic scenarios at 30–60 km/h, the model achieved 94.1% accuracy, 94.4% precision, 99.0% recall, and a 96.6% F1-score, with a reported ROC-AUC of 0.908. This false-positive rate (FP) of 34.0% indicates that reliability in warning must be improved. In hardware testing, the median processing time is found to be 75ms, costs estimated to be USD 205-245. The framework holds promise for making decisions with computational assistance, but independent ground truth validation, sensor calibration and testing in the real world will be required before it is deployed in a safety-critical situation.