Intersection-Specific Trajectory Prediction for Road Users: A Review

Intersections are critical points in urban traffic networks, accounting for over 50% of traffic accidents and nearly 30% of fatalities, highlighting the need for enhanced safety measures. Accurate trajectory prediction at intersections is essential for advanced driver assistance systems and autonomous vehicles to predict the future states of the traffic agents, enabling safer and more efficient navigation. This review examines methodologies for predicting road user trajectories at intersections, categorizing them into traditional models, machine learning techniques, and hybrid approaches. We conduct a comparative analysis of benchmark datasets and evaluation metrics. Key challenges such as sensor fusion, adaptive modeling of dynamic traffic scenarios, and enhancing computational efficiency are suggested for future research. By addressing these challenges and emphasizing the importance of benchmarking and real-world validation, this review aims to drive advancements in trajectory prediction models, ultimately contributing to safer and more efficient urban traffic management.