A double-cycle coordinated control framework for expressway ramp-arterial coordination

Expressway ramp-arterial coordination at urban expressway-to-arterial transition zones is decisive for mitigating cross-network spillover-induced gridlock and enhancing regional mobility. Prevailing traffic control studies, however, primarily focus on optimizing intersection scenarios while overlooking the dynamic impacts of freeway ramp queues on surface arterials, which significantly degrades network-wide capacity. This study proposes a reinforcement learning (RL) and backpressure theory integrated method for expressway ramp-arterial coordination, transcending conventional decentralized control paradigms. Key innovations include: a dynamic ramp threshold calculation method combining CNN predictions with traffic flow equations to pinpoint coordinated intervention timing; and a closed-loop control strategy wherein backpressure algorithms quantify traffic pressure at critical intersections, generating right-of-way allocation policies optimized via Soft Actor-Critic (SAC) reinforcement learning. The SAC reward function holistically incorporates queue length, dissipation efficiency, and spillover prevention objectives to dynamically reconcile expressway-arterial throughput demands, establishing an adaptive feedback cycle. Experimental results demonstrate 41% reduction in total system delay, with ramp queues and waiting times decreasing by 33% and 37.5% respectively versus benchmark methods. These findings substantiate the paradigm-shifting role of expressway ramp-arterial coordination in resolving trans-network bottlenecks and provide a theoretical method for urban road infrastructure.