Reinforcement Learning-Based Cooperative Variable Speed Limit Control Strategy for Merging Area

Addressing the mutual influence between mainline traffic flow and on-ramp traffic flow in freeway on-ramp merging area, a reinforcement learning-based cooperative variable speed limit control model for mainline and on-ramp is developed in this study. Existing freeway variable speed limit control primarily focuses on single-dimensional control of mainline traffic flow; however, grounded in the mixed traffic flow environment, this research integrates on-ramp control with mainline variable speed limit to realize comprehensive control of the entire merging area. Specifically, the variable speed limit control method is applied to the on-ramp, where the action spaces of both the on-ramp and mainline are discretized and integrated. Meanwhile, the reward function incorporates both the operational efficiency of the merging area and the queue length on the on-ramp. Finally, simulations are conducted to verify the model’s performance under the condition of 30% connected vehicle penetration rate. Compared with the scenario without variable speed limit control, the proposed model reduces the average travel time of vehicles by 4.87% and increases the average speed of vehicles by 3.16%; within the merging area, the total travel time of vehicles is decreased by 5.63%. Consequently, the proposed variable speed limit control strategy for merging area can effectively alleviate traffic congestion and improve the operational status of merging area.