Congestion is a major constraint on traffic efficiency, energy use, and carbon emissions. Intelligent flow regulation and resource allocation have therefore become central issues in transportation research, with dynamic emergency lane management recognized as a key strategy to reduce congestion, increase throughput, and support green traffic management. To address the challenges of limited predictive foresight, lack of decision-making mechanisms, and inefficient sensor deployment, this study proposes a closed-loop framework integrating monitoring, prediction, decision-making, and optimization. A dual-stream iterative prediction model is developed to suppress long-horizon error accumulation by incorporating segment length, lane status, and temporal factors as controllable inputs, enabling accurate downstream forecasting. Decision rationale is established through the congestion potential field, which combines macroscopic flow and microscopic disturbances, while a multi-dimensional evaluation system based on the analytic hierarchy process (AHP) provides forward-looking activation criteria. Sensor deployment is optimized using a multi-objective model that balances decision efficacy and cost, solved with particle swarm optimization enhanced by a Gaussian process surrogate model. Experiments conducted on a PreScan/CarSim/MATLAB/PyCharm co-simulation platform demonstrate notable improvements in prediction accuracy, decision quality, and overall system performance, offering a theoretical basis for sustainable and intelligent traffic management. • Closed-loop emergency lane control • Congestion potential field modeling • Dual-stream prediction network • Multi-objective sensor optimization • High-fidelity co-simulation validation
Tang et al. (Sun,) studied this question.