The transport sector remains one of the main sources of air pollution and climate change on our planet. With about 59% of global oil consumption and 22% of CO2 emissions, it is essential to develop strategies to reduce fuel consumption in vehicles. This could help reduce pollution while addressing global energy shortages.

An innovative approach to a global problem

Researchers from the Hong Kong University of Science and Technology have tackled this challenge by designing a computation model based on reinforcement learning. This model, detailed in a recently published paper on the arXiv preprint server, aims to optimize fuel consumption in car-following situations, especially when semi-automated and autonomous vehicles are present on the road.

Hui Zhong, co-author of the study, told Tech Xplore: “The inspiration for this report comes from the increasing demand for sustainable and energy-efficient transportation solutions.” Indeed, traffic jams and inefficient driving behaviors aggravate fuel consumption and emissions. The researchers wanted to find ways to mitigate these issues.

Development and functioning of EcoFollower

The main objective of Zhong and his team’s work was to create a computer model to optimize fuel consumption during car following. This model must ensure a safe distance between vehicles as well as efficient traffic flow. They developed EcoFollower, a deep reinforcement learning model.

According to Zhong, “EcoFollower is designed to optimize fuel consumption while driving.” The model continuously learns from its environment, adjusting following distances and acceleration patterns to adopt fuel-efficient driving behavior. Its distinctive feature lies in its ability to balance energy efficiency while maintaining smooth and safe traffic.

An advancement over traditional models

Traditional models often focus exclusively on safety or traffic smoothness. EcoFollower, on the other hand, also aims to reduce fuel consumption, providing a more comprehensive solution to the challenges of modern driving.

Evaluation and promising results

The researchers tested their model using the Next Generation Simulation (NGSIM) dataset, which compiles traffic information collected from various environments. The preliminary test results revealed that EcoFollower could significantly reduce fuel consumption in all tested situations.

“We have demonstrated that reinforcement learning can be effectively applied to real driving scenarios to decrease fuel consumption,” said Zhong. Results indicate that EcoFollower could reduce fuel consumption by 10.42% compared to typical driving situations. This figure represents a major advancement in reducing global emissions and promoting sustainable transportation options.

Future prospects for EcoFollower

In the near future, the EcoFollower model could be integrated into advanced driver assistance systems (ADAS) and autonomous driving systems. This would increase their efficiency while reducing their environmental impact. The researchers also intend to continue optimizing this model to further enhance its performance.

Conclusion

The research surrounding EcoFollower illustrates how artificial intelligence can play a crucial role in the transition towards more sustainable transportation. With innovations like these, we have the opportunity to transform how we move while protecting our planet.

“Although its performance is already superior to that of the traditional Intelligent Driving Mode (IDM) and it reduces fuel consumption by 10.42% compared to real driving scenarios, more scenarios and datasets are needed to further test and improve its generalization and robustness,” added Zhong. “For example, in a mixed autonomy traffic environment, the behavior of human-driven vehicles differs from that of autonomous vehicles, which could impact the performance of the model.”

More information: Hui Zhong et al, EcoFollower: an environmentally friendly car-following model considering fuel consumption, arXiv (2024). DOI: 10.48550/arxiv.2408.03950