Lane Change Prediction in Autonomous Driving: A Statistical Survey on Intention and Trajectory Forecasting

Lane-change prediction has emerged as one of the core challenges in autonomous driving research over the past decade. Previous surveys have focused on either lane-change intention or trajectory prediction, without offering a unified view of both tasks. They provide limited statistical analysis, offering mostly qualitative insights without capturing research trends. Additionally, they overlook interpretability and explainability, which crucial for trustworthy prediction systems. In this survey, we conduct a comprehensive review of over 150 research articles to systematically analyze how the field has evolved across different key dimensions: prediction tasks (intention, trajectory, or both), types of sensors used, considered inputs, datasets employed, modeling approaches, and evaluation practices. Beyond summarizing existing efforts, this survey also presents a statistical analysis that exposes critical gaps in the literature. The analysis reveals a lack of focus on interpretability, transparency, and explainability of the models, despite their importance for building reliable and trustworthy prediction systems. Additionally, only a small subset of studies incorporate decision-making modules that act upon the predicted behaviors, and even fewer offer experimental or real-world validation. Based on our findings, we outline recommendations for future research to advance toward more explainable, decision-aware, and deployable lane-change prediction systems.