An integrated control strategy for freeway weaving area considering CAV dedicated lanemerging demand

Abstract:

To improve freeway operational efficiency and optimize spatiotemporal resource allocation in weaving area under the setting of dedicated lane for connected and autonomous vehicles (CAVs), a deep reinforcement learning-based integrated control strategy is proposed. This strategy aims to ensure the efficient and safe merging of CAVs into the dedicated lane. The research focuses on a three-lane freeway configuration, with the innermost lane designated for CAVs. A multi-objective reward function is developed to address the dedicated lane merging demand of CAVs while simultaneously considering the efficiency of mainline traffic and the length of ramp queue. The deep deterministic policy gradient (DDPG) algorithm is employed to implement the integrated control strategy, which encompasses entrance ramp signal control, variable speed limit at the lane level, and adjustment to the gaps between CAV platoon. A simulation environment for the freeway weaving area is designed using SUMO and Python to assess the effectiveness of the proposed control strategy. The results demonstrate that, with a CAV penetration rate of 30%, the integrated control strategy advances the longitudinal positioning of CAVs entering the dedicated lane, merging success rate increases of 19.34%, 22.86%, and 25.55% under low, medium, and high traffic demand scenarios, respectively. Additionally, average vehicle travel time is reduced by 5.42%, 17.41%, and 20.65% under the same conditions. The proposed integrated control strategy for weaving area demonstrates significant effectiveness by not only achieving effective guidance for CAV merging dedicated lane but also enhancing the traffic efficiency and operational safety of the mainline, providing a theoretical basis and technical reference for optimizing the traffic operation in weaving areas of freeway under CAV dedicated lane conditions.