Automatic driving decision-making on left-turn behavior at unsignalized intersections

Abstract

Unsignalized intersections are a common scenario in urban areas. Therefore, it is necessary to consider decisions at unsignalized intersections. Left-turn behavior is used as a representative because there are more cases of interactions with other traffic participants under left-turn behavior. First, the paper proposes a way to improve the efficiency and comfort of the host vehicle when the intersection is obscured. Second, a dynamic game theory is proposed based on the potential conflict area of the intersection, considering safety and traffic efficiency. An estimation of the aggressiveness model based on the potential conflict area is proposed. Thirdly, a driving scheme is proposed by adjusting the passing scheme when the game is yielding. Finally, it was simulated for validation at unsignalized intersections by VTD (Virtual Test Drive). Simulation results show that the proposed method can balance comfort and passing efficiency through unsignalized intersections in the presence of occlusion. The average speed of the proposed method through the intersection is higher when interacting with traffic participants.