Car-Following Model for Mixed-Traffic Systems Considering Driving Styles

With the rapid advancement of autonomous driving, its practical applications are gradually becoming a reality. In future road traffic systems, the coexistence of autonomous vehicles (AVs) and human-driven vehicles (HVs) with diverse driving styles will remain a long-term phenomenon. Developing a mathematical car-following model for mixed traffic flow that incorporates driver styles is thus of particular importance for analyzing future traffic operating conditions. For HVs, this study constructs a CTH–Gipps car-following model by integrating the constant time headway (CTH) model and the Gipps model, while introducing a driver style coefficient to characterize the car-following behavior of drivers with different styles. In addition, for autonomous driving, a softmax deep double deterministic policy gradient algorithm is adopted from deep learning, with the state space, action space, and reward function carefully designed to develop the autonomous vehicle car-following model. Simulation experiments are conducted to investigate the operating states of mixed traffic flows comprising AVs and HVs, analyzing fundamental characteristics as well as the relationships among density, flow, and speed. The results indicate that traffic flow capacity peaks vary significantly across driving styles, with the aggressive driving style generally corresponding to the highest peak value. Moreover, higher autonomous vehicle penetration rates lead to greater traffic flow values at the same density, reflecting the substantial improvement in road capacity brought by the integration of autonomous vehicles into the traffic stream. The findings of this study hold important implications for promoting the practical deployment of AVs.