MIT Researchers Find Eco-driving Measures Can Significantly Reduce Vehicle Emissions

Researchers used machine learning to optimize eco-driving strategies, reducing vehicle emissions in urban areas. The study simulated over a million traffic scenarios in three cities and found that eco-driving measures could significantly reduce emissions. The findings could inform the implementation of eco-driving technologies, which could involve vehicle dashboards or direct acceleration control through vehicle-to-infrastructure communication systems.

Researchers from MIT have leveraged machine learning to develop eco-driving strategies that significantly reduce vehicle emissions in urban areas. Their study, published in Transportation Research Part C: Emerging Technologies, simulated over a million traffic scenarios in three major U.S. cities—Atlanta, San Francisco, and Los Angeles—and found that eco-driving measures could substantially lower emissions.

Eco-driving involves dynamically adjusting vehicle speeds to minimize stopping and excessive acceleration, thereby reducing energy consumption. The MIT researchers used a powerful AI method called deep reinforcement learning to optimize these strategies. They identified 33 factors influencing vehicle emissions, including temperature, road grade, intersection topology, and driver behavior, and simulated traffic scenarios at more than 6,000 signalized intersections.

The study revealed that full adoption of eco-driving could cut annual city-wide intersection carbon emissions by 11 to 22 percent without slowing traffic throughput or affecting safety. Even if only 10 percent of vehicles employed eco-driving, it would result in 25 to 50 percent of the total reduction in CO2 emissions. Additionally, optimizing speed limits at about 20 percent of intersections provides 70 percent of the total emission benefits.

The findings suggest that eco-driving measures can be implemented gradually while still having measurable, positive impacts on mitigating climate change and improving public health. Combining eco-driving with alternative transportation decarbonization solutions, such as hybrid and electric vehicles, could yield even greater emission reductions.

The research was funded in part by Amazon and the Utah Department of Transportation. The study highlights the potential of machine learning in optimizing traffic control measures and could inform the implementation of eco-driving technologies, which could involve vehicle dashboards or direct acceleration control through vehicle-to-infrastructure communication systems.

References:
[1] https://news.mit.edu/2025/eco-driving-measures-could-significantly-reduce-vehicle-emissions-0807