A Rulefit Based Model for Driving Intention Prediction at Intersections

Understanding the driving intentions of surrounding vehicles is crucial for safe decision-making and path planning in advanced driver assistance systems and automated vehicles. This paper presents a novel Rulefit-based model to predict driving intentions at intersections. The model comprises two parts. The first part contains three sub-tasks: judging whether the vehicle is more likely to turn left or go straight, turn right or go straight, and turn left or turn right with the Rulefit algorithm. Here, a rule set is introduced to fully use the experiences learned from the historical moments and increase the prediction accuracy. The second part uses hard voting to get the classification results. Simulations validate the model's performance using data in lankershim street from the NGSIM dataset. It shows that the proposed model is more accurate than typical model-based and neural-network-based methods on validation data when the distance between the vehicle and the intersection is less than 21 meters. Moreover, the proposed method can give the rules that play the most critical roles.