Game Theory-Based Harmonious Decision-Making for Autonomous Bus Lane Change

Blended traffic, comprising autonomous buses (ABs) and human-driven vehicles (HDVs), is becoming increasingly common, yet the lane change decision-making for ABs remains challenging due to complex interactions with heterogeneous HDVs. To address the challenge above, this paper proposes a game theory-based harmonious decision-making (GTHD) algorithm considering nuance of driving styles of HDVs, achieving human-like performance in interactions with HDVs. Technically, a game theoretic model of the GTHD uses predictions of the opposing vehicle’s motion and the information from preplanned trajectories. Besides, a prior estimation for driving styles is obtained utilizing clustering of historical data, and refined in real time through Bayesian estimation. Then, the driving style estimation is utilized to modify the game theoretic model. The modified model provides a closer depiction of the opponent’s preferences, meanwhile adjusts self-preferences to adapt to the opponent. The efficacy of GTHD is validated using a hardware and human in loop simulator and datasets in MLC scenarios. It is shown that the GTHD achieves human-like performance with 91.50%-98.50% accuracy compared with human bus driver under different conditions, better than several lane change models based on data driven methods. The code is open source and available at https://github.com/guofan999/GTHD.