Optimal temporal-spatial variable speed limit profiles for a heterogeneous freeway corridor: A real-time control and physical model-aided reinforcement learning solution method

Variable speed limit control has been confirmed for improving traffic efficiency, safety, and environmental performance in literature. However, most existing approaches focus on isolated bottlenecks and therefore do not fully exploit coordinated speed control across an entire freeway corridor. This study develops an Adaptive Rolling Horizon Speed Limit (ARSL) control framework that dynamically coordinates speed limits across space and time along a heterogeneous freeway corridor. The framework integrates a traffic flow model into an optimal control formulation to jointly optimize traffic flow energy consumption and throughput. Two key modeling components are introduced: a Speed Limit Adaptive Cell Transmission Model (SLA-CTM) that captures traffic propagation in response to variable speed limits, and a traffic flow energy consumption model expressed in terms of traffic density and speed. To solve the resulting large-scale optimal control problem efficiently, we develop an Actor-Critic reinforcement learning framework combined with Monte Carlo Tree Search (MCTS) to guide policy exploration. Numerical experiments conducted on a 7‑mile freeway corridor in Seattle, WA, demonstrate that the proposed method achieves up to a 40% reduction in energy consumption, with minor throughput decrease by less than 0.3% compared with conventional uniform speed‑limit control.