A Comprehensive Survey of Lane Change Prediction for Autonomous Vehicles

Abstract

In the field of transportation, road vehicle automation has improved in many ways, and it has also introduced challenges for existing systems. A major problem in autonomous driving is predicting lane changes, because it affects the traffic behavior at both the individual and system levels. Considering safety, driving conditions, and human behavior are important in making these decisions. Most existing approaches work on intention and trajectory separately; thus, there is a lack of in-depth quantitative analysis. In real-world applications, the limited interpretability makes their predictions tougher to trust. The challenges in vehicle lane change prediction are evaluated by classifying and assessing existing approaches. By analyzing driver behavior modeling and prediction methods, this study categorizes the lane change models. By considering multiple datasets and methodological perspectives, this evaluation is performed. It also reflects current limitations and summarizes future research directions. Furthermore, this survey highlights the use of advanced techniques in increasing the accuracy of prediction. This research article is analyzed based on datasets, tools, metrics, and publication year, revealing that Deep Learning (DL) is the most effective approach, Python is the most commonly used tool, accuracy is the primary evaluation metric, and most of the papers were published in 2025. This provides a strong path to develop an innovative model for vehicle lane change prediction based on this assessment.