The escalating impact of vehicular Carbon Dioxide (CO2) emissions on air pollution, global warming, and climate change necessitates innovative solutions. This paper proposes a comprehensive Internet-of-Vehicles (IoV) network for real-time CO2 emissions estimation and reduction. We implemented and tested an on-board device that estimates the vehicle’s emissions and transmits the data to the network. The estimated CO2 emissions values are close to the standard emissions values of petrol and diesel vehicles, accounting for expected discrepancies due to vehicles’ age and loading. The network uses the aggregate emissions readings to inform the Reinforcement Learning (RL) algorithm, enabling the prediction of optimal speed limits to minimize vehicular emissions. The results demonstrate that employing the RL algorithm can achieve an average CO2 emissions reduction of 11 kg/h to 150 kg/h.
