Personalized Lane-changing Behavior Decision Model Considering Driving Habits

The development of automated driving technologies for humans should adopt a human-centered approach to ensure safety and satisfactory user experience. Making automated lane change decisions in high-speed traffic flows is extremely challenging, especially when considering the habits and preferences of different drivers. To address this problem, this paper proposes a personalized lane change behavior decision model considering driving habits. Instead of directly mimicking human behavior, the method employs inverse reinforcement learning to learn the driver-driving behavior generation mechanism from personal historical driving trajectories and select personalization indicators to describe the driver’s preferences in lane-changing decisions. The risk field is invoked to design a reward function to enhance the interpretability of personalized decision-making by portraying the driver’s perception of risk. The model is trained and tested in a multi-lane highway scenario. Results show that the proposed method can make lane-changing decisions, which are consistent with the driver’s driving habits. Additionally, our model analyzes the characterization of relevant indicators that influence lane-changing decisions.