Hierarchical Control Strategy for Cooperative On-Ramp Merging of Connected and Automated Vehicles on Multilane Highways

In mixed traffic environments, fully utilizing the traffic capacity of the multiple lanes on the inner side of highways and coordinating connected and automated vehicles (CAVs) to complete merging is an extremely challenging task. This article proposes a hierarchical cooperative merging control strategy for CAVs on multilane highways. It comprises two layers. The premerging lane-changing layer aims to guide CAVs from the outer lanes of the mainline to the inner lanes. We establish a CAV lane allocation model and propose a rule-based CAV lane-changing strategy for both free and cooperative lane-changing scenarios. In the merging control layer, we adopt a hierarchical reward optimization multiagent deep deterministic policy gradient (HRO-MADDPG) algorithm to finely optimize the merging trajectories of CAVs. To improve the convergence performance of the merging task algorithm, global and local rewards are set by considering both the overall objectives of the merging control area and the individual objectives of CAVs. A three-lane highway is used as a case study. The results indicate that the proposed strategy effectively promotes the uniform distribution of traffic flow across the mainline lanes, significantly relieving merging pressure between the outermost lane and the ramp. HRO-MADDPG outperforms existing algorithms in terms of convergence speed, efficiency, and safety. The proposed strategy also effectively improves traffic efficiency under different CAV penetration rates and ramp merging rates, providing a reference for future merging management strategies of multilane highways under mixed traffic.