A Decision Method of Mandatory Lane-Changing for Autonomous Vehicles Based on Repeated Game

Looking at the difficulty for vehicles to make a mandatory lane change in the upstream area of the off-ramp in a congested multi-lane expressway, this paper proposes a repeated game-based, mandatory lane-changing decision-making model under heterogeneous traffic flow. First, repeated game cycles are developed based on the motion characteristics of the car behind the target lane, and beliefs about the driving style of the car behind the target lane are updated based on the results of the previous games. The mandatory lane-changing model is calibrated using NGSIM data combined with a two-layer framework model, while the model is validated using various performance metrics. The simulation results show that the repetitive game-based, mandatory lane-changing model has higher lane-changing success and drive-out rate than the traditional space-based, lane-changing model in a multi-lane highway exit area with high traffic density.