Model-Based Reinforcement Learning for Eco-Driving Control of Electric Vehicles

Lee, Heeyun; Kim, Nam Wook; Cha, Sungdeok

doi:10.1109/access.2020.3036719

Cited by 52 publications

(25 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…MBRL incorporates indirect learning using a model, which is constructed by observing the environment [32,33]. In this paper, a novel MBRL algorithm using the ECMS is developed based on our previous study [29,30,34], in which MBRL was successfully applied to the optimal control problem of the HEV and electric vehicle. Here, domain knowledge of the vehicle system's dynamics is used for constructing a model to reduce the uncertainty of the entire environment, while ensuring that the vehicle system remains model-free by updating the value based on observation.…”

Section: Rl Based On Ecmsmentioning

confidence: 99%

Reinforcement Learning Based on Equivalent Consumption Minimization Strategy for Optimal Control of Hybrid Electric Vehicles

Lee

Cha

2021

IEEE Access

Self Cite

View full text Add to dashboard Cite

Section: Rl Based On Ecmsmentioning

confidence: 99%

Reinforcement Learning Based on Equivalent Consumption Minimization Strategy for Optimal Control of Hybrid Electric Vehicles

Lee

Cha

2021

IEEE Access

Self Cite

View full text Add to dashboard Cite

“…In addition, with the development of machine learning, the field of eco-driving strategy has also begun to use intelligent algorithms to solve optimal control. Lee et al [ 53 ] developed a model-based reinforcement learning algorithm (MBRL) that considers the road gradient for the eco-driving of electric vehicles. The simulation results indicated that the speed profile optimized using model-based reinforcement learning had similar behavior to the global solution obtained via DP and exhibited an energy-saving performance of 1.2–3.0%, which is similar to DP.…”

Section: Eco-driving Theorymentioning

confidence: 99%

An Overview of Eco-Driving Theory, Capability Evaluation, and Training Applications

et al. 2021

Sensors

View full text Add to dashboard Cite

Constrained by traditional fuel-saving technologies that have almost reached the limit of fuel-saving potential, the difficulty in changing urban congestion, and the low market penetration rate of new energy vehicles, in the short term, eco-driving seems to be an effective way to achieve energy-saving and emissions reduction in the transportation industry. This paper reviews the energy-saving theory and technology of eco-driving, eco-driving capability evaluation, and the practical application of eco-driving, and points out some limitations of previous studies. Specifically, the research on eco-driving theory mostly focuses on a single vehicle in a single scene, and there is a lack of eco-driving research for fleets or regions. In addition, the parameters used to evaluate eco-driving capabilities mainly focus on speed, acceleration, and fuel consumption, but external factors that are not related to the driver will affect these parameters, making the evaluation results unreasonable. Fortunately, vehicle big data and the Internet of Vehicles (V2I) provides an information basis for solving regional eco-driving, and it also provides a data basis for the study of data-driven methods for the fair evaluation of eco-driving. In general, the development of new technologies provides new ideas for solving some problems in the field of eco-driving.

show abstract

“…Sakhdari and Azad proposed an adaptive tube-based nonlinear MPC method for economic autonomous cruise control of plug-in hybrid electric vehicles, which can improve the economy and tracking of ACC systems while maintaining robustness to disturbances and modeling errors [9]. To reduce the energy consumption, Lee et al presented a novel model-based reinforcement learning algorithm for eco-driving control of EVs, where the domain knowledge of vehicle dynamics and the powertrain system is utilized for the reinforcement learning process while modelfree characteristics are maintained by updating the approximation model using experience replay [20]. In addition, Li and Görges put forward an model-free Eco-ACC strategy based on reinforcement learning for fuel vehicles, which can improve fuel economy and enhance vehicle safety [21].…”

Section: Introductionmentioning

confidence: 99%

Economic Adaptive Cruise Control for Electric Vehicles Based on ADHDP in a Car-Following Scenario

et al. 2021

View full text Add to dashboard Cite

In order to ensure vehicle safety, enhance riding comfort, extend the battery life of electric vehicles (EVs), and improve the energy economy, an ADHDP-based economic adaptive cruise control (Eco-ACC) strategy for EVs in car-following scenarios is proposed in this paper. First, the longitudinal dynamics of EVs is modeled, and the control objectives are presented; then, the actor-critic structure of ADHDP is introduced, and the policy iteration formulas of the critic and actor networks in the ADHDP framework are given; finally, after the state variables, control variables, unity function and value function are determined, the ADHDP-based Eco-ACC strategy for EVs is designed. Extensive simulation results under different driving cycles show that the proposed Eco-ACC strategy can not only ensure vehicle safety, improve riding comfort and reduce energy consumption, but also significantly reduce the battery capacity loss and extend the battery life compared with the benchmark algorithm. In addition, the proposed Eco-ACC strategy is model-free and real-time, and can be robust in different car-following scenarios.

show abstract

Model-Based Reinforcement Learning for Eco-Driving Control of Electric Vehicles

Cited by 52 publications

References 21 publications

Reinforcement Learning Based on Equivalent Consumption Minimization Strategy for Optimal Control of Hybrid Electric Vehicles

Reinforcement Learning Based on Equivalent Consumption Minimization Strategy for Optimal Control of Hybrid Electric Vehicles

An Overview of Eco-Driving Theory, Capability Evaluation, and Training Applications

Economic Adaptive Cruise Control for Electric Vehicles Based on ADHDP in a Car-Following Scenario

Contact Info

Product

Resources

About