Socially Acceptable Human-Like Behavior Planning for Connected Cars on Signalized Road Network

This paper proposes a socially acceptable human-like behavior planning method to improve the driving efficiency for connected cars, especially in the dilemma zone on signalized road networks. Specifically, the proposed method is based on a soft-constrained model predictive control using slack variables that are assigned to the state constraints, resulting in the slight violation of the road speed limit to improve the efficiency. By exploiting the upcoming signal-phase-and-timing information of multiple traffic lights via vehicle connectivity technologies, a connected car optimizes the speed trajectory, leading to reduced stops in the dilemma zone compared to the vehicles that strictly comply with traffic rules. As a result, it is observed that connected cars could pass through multiple traffic lights when the green or yellow signals are turned on, resulting in the minimization of the entire trip time. The proposed method is comprehensively verified through simulations under various situations, and, the behavior of the connected cars is observed similar to that of experienced human drivers, particularly in the dilemma zone. We also show the efficacy of our approach through the experiments comparing the speed trajectory generated by our approach with those of different human drivers using a lab-scale driving simulator.