Eco-driving for connected automated hybrid electric vehicles in learning-enabled layered transportation systems

Eco-driving strategies have the potential to enhance energy savings, safety, and transportation efficiency by optimizing vehicle interactions with dynamic traffic environments. This study addresses the challenge of balancing computational efficiency and optimization effectiveness amid the high-dimensional state and control variables driven by extensive traffic information. The novelty different from existing methods lies in developing an eco-driving strategy within a traffic information cyber-physical system. The cyber-layer maps simulated road segments for training vehicles equipped with the Proximal Policy Optimization (PPO) algorithm, enabling effective planning of economical speeds. During vehicle operation, the cyber-layer maps the real-time physical environment, providing a predictive state sequence for the vehicle's adaptive equivalent fuel consumption minimization strategy. Then, optimizing the efficiency factor in a rolling manner further improves fuel economy. A comparative analysis with existing methods across different scenarios shows that the proposed strategy significantly improves fuel economy while ensuring real-time speed planning and reliable speed-tracking performance.