A Causal Inference–Based Speed Control Framework for Discretionary Lane-Changing Processes

Abstract

Vehicle lane changing is one of the most important topics in autonomous driving and is the behavior most likely to cause vehicle accidents. In order to optimize the performance of vehicle lane change by controlling the speed of the vehicle, which makes the lane change safer, more efficient, and more eco-friendly, a causal inference-based speed control framework is proposed in this study. First, variables affecting the lane change efficiency are selected and their treatment effects are obtained by conducting causal effect analysis and causal refutation analysis. Second, the conditional average treatment effect is estimated by taking the safety and eco-friendly requirements into consideration and applying the double machine learning version causal forest model. Finally, a controller is built, and grid search is used to optimize lane change decision-making. By conducting numerical experiments, the average lane change duration decreased by 5.9% and average velocity increased from 7.165 m/s to 7.791 m/s. The results show that the framework has good performance on improving the efficiency and environmental friendliness of a lane change process while ensuring safety. The proposed framework provides a new approach to speed control for discretionary lane-changing behavior. The approach can be well adapted to a vehicle-to-vehicle communication environment and is expected to be applied to vehicle-assisted driving systems in the future.