Researchers have proposed an integrated eco-driving framework for fuel cell hybrid electric vehicles in multi-lane highway scenarios, using deep reinforcement learning to optimize motion trajectory planning and energy management at the same time. The framework is designed to improve energy economy in real traffic environments by jointly controlling acceleration, steering angle, and engine power.
Eco-driving strategies are important for hybrid electric vehicles because they can reduce energy use, emissions, and operating costs. Many previous studies have focused on longitudinal car-following or lane-changing maneuvers, but these approaches may not fully account for continuous lateral dynamics. In multi-lane highway settings, vehicles must make decisions that involve both forward motion and lateral behavior, while also managing the power system efficiently.
The new study addresses this limitation by proposing a unified framework in which trajectory planning and energy management are optimized together. According to the article, the framework uses continuous control variables including acceleration, steering angle, and engine power. This allows the vehicle to consider how its motion decisions affect both driving behavior and energy consumption, rather than treating route or trajectory planning separately from power management.
A key element of the approach is the fusion of traffic spatial information and vehicular power operating conditions into the decision-making input. The authors extract key features from surrounding traffic and power-system states and formulate them as the state matrix for the deep reinforcement learning model. This matters because eco-driving in real traffic is not only a question of minimizing fuel or hydrogen use; it also depends on the positions and movements of nearby vehicles and the operating state of the power system.
The framework is optimized using the Soft Actor-Critic algorithm, a deep reinforcement learning method suited to complex strategy spaces and multi-objective optimization tasks. In practical terms, this allows the model to explore different combinations of driving and energy-management actions while learning policies that balance comfort, speed, energy economy, and power-system health. The article also notes that the project is available online at https://github.com/sicilyala/EcoAD.
The study?s analysis of the co-optimization process helps reveal the energy conservation mechanism behind the proposed eco-driving strategy. Instead of simply slowing down or applying a fixed rule, the framework coordinates transverse and longitudinal motion with energy management. This collaborative optimization is especially relevant in multi-lane environments, where steering and acceleration decisions can influence both comfort and energy demand.
Experimental results reported in the paper indicate substantial improvements in several performance measures. The proposed eco-driving strategy achieved better transverse-longitudinal comfort and energy economy while sacrificing 14.07% of average speed. In return, it produced an 87.65% improvement in the state-of-health performance of the power system, while reducing hydrogen consumption and driving cost by 86.17% and 89.58%, respectively.
These results suggest that eco-driving can be more effective when vehicle motion and power-system operation are treated as a coupled optimization problem. For fuel cell hybrid electric vehicles, this could help extend system life, lower operating costs, and reduce hydrogen use, while still maintaining acceptable driving behavior in traffic. The trade-off with average speed also highlights an important practical issue: energy-saving policies must be evaluated alongside mobility and user-acceptance requirements.
Further validation will still be needed across broader road networks, traffic densities, vehicle platforms, and real-world driving conditions. Even so, the study offers a strong indication that deep reinforcement learning can support integrated eco-driving strategies for intelligent hybrid vehicles. As transportation systems become more automated and energy-aware, frameworks that combine trajectory planning with energy management may play an important role in improving both efficiency and sustainability.
Reference
Author:
Weiqi Chen a , Jiankun Peng a , Yuhan Ma a , Hongwen He b , Tinghui Ren a , Chunhai Wang c
Title of original paper:
Eco-driving framework for hybrid electric vehicles in multi-lane scenarios by using deep reinforcement learning methods
Article link:
https://www.sciencedirect.com/science/article/pii/S2773153725000593
Journal:
Green Energy and Intelligent Transportation
DOI:
10.1016/j.geits.2025.100309
Affiliations:
a School of Transportation, Southeast University, Nanjing 211189, China
b School of Mechanical Engineering, Beijing Institute of Technology, Beijing 100081, China
c Sky-well New Energy Automobile Group Co. Ltd., Nanjing 211100, China
Green Energy and Intelligent Transportation
Experimental study
Not applicable
Eco-driving framework for hybrid electric vehicles in multi-lane scenarios by using deep reinforcement learning methods
31-Dec-2025