An Ensemble Model for Lane Change Intention Inference for Autonomous Driving

Advanced Driver Assistance System (ADAS) requires an accurate and timely recognition and prediction of the lane change intention of the driver for decision making. Traditional intention models are based on mathematical, logical and machine learning which struggle to adequately represent the entire lane change process. Here, a novel ensemble model is proposed for identifying the lane change intention of human-driven vehicle before the trajectory begins by analyzing the various factors in discretionary lane change of vehicles in hybrid traffic scenarios where the autonomous and human-driven vehicles coexists. The proposed ensemble model combines the Support Vector Machines (SVM) and Random Forest (RF) to precisely identify the lane change behavior. The model is evaluated using the benchmarked dataset NGSIM which includes naturalistic highway driving data consisting of different lane change scenarios. The ensemble model is compared with other models and shows an average accuracy of 96% for different lane change maneuvers.