Carbon-Aware Smart Transportation Simulation Driven by Deep Learning and Policy Scenarios

This study proposes a carbon-aware smart mobility simulation framework designed to support low-emission urban transportation planning through data-driven modeling and simulation. Integrating a Long Short- Term Memory (LSTM) network with a Monte Carlo-style scenario engine, the framework captures complex spatiotemporal patterns in traffic-related CO<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf> emissions and evaluates the effectiveness of policy interventions under uncertainty. Permutation-based feature attribution and perturbation sensitivity analysis reveal that moderate congestion, adverse weather and ride-sharing demand are key drivers of emissions. Simulation results across four traffic strategies - baseline, traffic reduction, signal optimization and rideshare promotion - demonstrate that demand-side interventions consistently outperform supply-side alternatives in reducing emissions, delay and energy use. The proposed approach provides a scalable and interpretable simulation tool for smart city applications, offering actionable insights for sustainable urban mobility and low-carbon traffic system design.