A Hierarchical Control Framework for Coordinating CAV-Dedicated Lane Allocation and Signal Timing at Isolated Intersections in Mixed Traffic Environments

Abstract

With the rapid development of connected and automated vehicles (CAVs), numerous studies have demonstrated that CAV-dedicated lanes (CAV-DLs) can significantly enhance traffic efficiency. However, most existing studies primarily focus on optimizing either CAV trajectory planning or traffic signal control, and the integration of CAV-DLs and signal control for improved spatiotemporal resource utilization remains underexplored. To address this challenge, this study proposes a hierarchical control framework that integrates CAV-DLs allocation with signal control. The framework employs two collaborative agents based on the dueling double deep Q-network (D3QN) algorithm. The upper-level agent recommends optimal CAV-DLs configurations based on long-term traffic flow patterns, while the lower-level agent focuses on real-time signal control by adjusting signal parameters and green time allocations in response to current traffic demand. Simulation results demonstrate that the proposed model effectively adapts to dynamic traffic conditions, significantly improving intersection capacity and reducing delays. Compared with benchmark approaches, the model achieves an average improvement of 31.8% in traffic efficiency. Additionally, the study identifies CAV penetration rate (CAV PR) thresholds of 30% and 60% as appropriate for allocating one and two CAV-DLs, respectively, at intersections with high traffic volumes. These findings provide valuable theoretical insights and practical guidance for the effective configuration of CAV-DLs in future traffic systems.