Comparison and Evaluation of Algorithms for LiDAR-Based Contour Estimation in Integrated Vehicle Safety

Many nations and organizations are committing to achieving the goal of 'Vision Zero' which aims to bring the number of road deaths close to zero by the year 2050. The core of the strategy is a safe transportation system with optimized vehicles and transportation routes. The industry continues to develop integrated safety systems to make vehicles safer, smarter, and more capable in safety-critical scenarios. Passive safety systems are now focusing on pre-crash deployment of restraint systems to better protect vehicle passengers. Current commonly used bounding box methods for the shape estimation of potential crash partners lack the fidelity required for edge case collision detection and advanced crash modeling of future pre-crash technologies. This has led to the development of novel algorithms for vehicle contour estimation in literature. With this work, we present a framework for assessing and comparing different contour estimation algorithms, including a simple bounding box, oriented bounding box, polynomial fit estimation, complemented convex hull, and three-arc fit. Tests on simulated virtual and experimental LiDAR measurements of a simplified vehicle contour have been conducted to determine performance at varying relative angles and distances. It has been concluded that the convex hull and the three-arc methods are the best performing of the studied algorithms, with each having different strengths. The three-arc algorithm offers higher accuracy estimations at low relative angles and near-mid distances, whereas the convex hull method requires low computation time and can provide accurate estimations even at extreme relative angles and distances.