Hierarchical adaptive cross-coupled control of traffic signals and vehicle routes in large-scale road network

Traffic signal timing and vehicle routing have been empirically demonstrated as the two most promising paradigms for network-level urban road traffic management. However, mainstream studies based on Wardrop's theory continues to treat these two modules separately without achieving effective coupling. Optimization-based methods face the challenge of increasing computational complexity as urban scales continue to expand, constrained to small-scale road networks. To address the above challenges, this paper proposes HAC3, a hierarchical adaptive cross-coupled control method for network-wide traffic management. HAC3 utilizes a rolling horizon architecture, comprising a fast update stage and a slow update stage. The core of the slow update stage is a spatiotemporal superposition vehicle route planning (SSP) module, which assigns the optimal route to each connected vehicle (CV) based on the road network state and the traffic signal timing of each intersection, and clarifies priority in right-of-way allocation to avoid falling into local optimal. The fast update stage is used for multi-intersection adaptive traffic signal control (TSC), taking the intersection state and vehicle routes as inputs to optimize the signal timing scheme. Through the asynchronous cross-coupling optimization of the two stages, the road network efficiency can be improved while ensuring equilibrium. Experimental results show that HAC3 achieves superior convergence performance on both synthetic and real-world road network data sets, outperforming baseline methods and proving its scalability to large-scale road networks. Plug-and-play experiments indicate the proposed HAC3 framework can integrate with other mainstream signal control models.