A Fuel Consumption-Oriented Optimal Variable Speed Limit Control Method on Freeway Corridor

The efficient implementation of variable speed limits offers substantial benefits to traffic efficiency, safety, and environmental sustainability. Nonetheless, the problem of the temporal-spatial variable speed limit control has not been investigated enough. In light of this, this study introduces an innovative approach termed Adaptive Rolling Horizon Speed Limit (ARSL) control. This approach integrates elements of optimal control modeling, traffic flow theory, and deep reinforcement learning to establish real-time, context-sensitive variable speed limits along a freeway corridor. The overarching goal is to enhance both overall fuel efficiency and throughput simultaneously. Initially, a comprehensive traffic flow model is devised to account for the impact of variable speed limit and capacity drop phenomenon. Subsequently, an investigation is conducted to discern the intricate relationship between traffic fuel consumption and the propagation of traffic. Building upon these foundational formulations, an optimal control model is devised to identify variable speed limit solutions that minimize fuel consumption, taking into consideration the dynamic nature of traffic flow and the diverse characteristics of heterogeneous road traffic. Finally, a policy neural network based on reinforcement learning is engineered to tackle this optimal control model within a freeway corridor characterized by varying road capacities. The outcomes of our numerical experiments underscore the efficacy of ARSL in reducing total fuel consumption by over 40% while incurring only a marginal reduction in overall travel distance (approximately 0.3%). ARSL holds the potential to markedly enhance macroscopic traffic fuel efficiency while upholding robust throughput levels.