Yuecheng Li

Beijing Information Science & Technology University

Author Statistics

Papers

Citation

H-Index

i-10 index

Research Trends

Author Order

Document Type

Co-Authors

Hongwen He

Beijing Institute of Technology

Jiankun Peng

Southeast University

Hongwen He

Beijing Institute of Technology

Yong Chen

Beijing Information Science & Technology University

Jingda Wu

Nanyang Technological University

Wenyan Jia

University of Pittsburgh

Zhi‐Hong Mao

University of Pittsburgh

Hao Wang

McMaster University

Amir Khajepour

University of Waterloo

Mingui Sun

University of Pittsburgh

Cooperative Institutions

Chinese Academy of Sciences

Shanghai Jiao Tong University

University of Chinese Academy of Sciences

Beijing Institute of Technology

Peking University

Chinese Academy of Medical Sciences & Peking Union Medical College

Huazhong University of Science and Technology

Xi'an Jiaotong University

Sichuan University

Tsinghua University

Author Statistics

Papers

Citation

H-Index

i-10 index

Research Field

Energy management for a power-split hybrid electric bus via deep reinforcement learning with terrain information

Applied Energy (2019)

Yuecheng Li Hongwen He Amir Khajepour Hong Wang Jiankun Peng

Powertrain

10.1016/j.apenergy.2019.113762

Cite

Citations (150)

Energy Management Strategy for a Series Hybrid Electric Vehicle Using Improved Deep Q-network Learning Algorithm with Prioritized Replay

DEStech Transactions on Environment Energy and Earth Science (2019)

Yuecheng Li Hongwen He Jiankun Peng Jingda Wu

On the basis of previous deep Q-network (DQN) based energy management strategy (EMS), two aspects of efforts are explored to improve it for more efficient and stable training and performances in this paper. First, the architecture of original DQN is changed in order to learn separately from the value of current driving/vehicle states and advantages of EMSs under current states; a duplicated network is adopted for Q-value calculation. Second, prioritized replay is introduced for more efficient data utilization during training of DQN based EMS. Simulation results show that the convergence of improved DQN based EMS is faster and higher reward can be achieved compared with original DQN based EMS. Simulation on China typical urban driving cycle for the series hybrid electric vehicle indicates that the fuel economy performance of improved DQN (6.07L/100km) is 8.4% higher than DP based EMS, exceeding original DQN based EMS (6.24L/100km) by about 3%.

Driving cycle

10.12783/dteees/iceee2018/27794

Cite

Citations (30)

Vehicle State Estimation Based on Sage–Husa Adaptive Unscented Kalman Filtering

World Electric Vehicle Journal (2023)

Yong Chen Hao Yan Yuecheng Li

To combat the impacts of uncertain noise on the estimation of vehicle state parameters and the high cost of sensors, a state-observer design with an adaptive unscented Kalman filter (AUKF) is developed. The design equation of the state observer is derived by establishing the vehicle’s three degrees-of-freedom (DOF) model. On this basis, the Sage–Husa algorithm and unscented Kalman filter (UKF) are combined to form the AUKF algorithm to adaptively update the statistical feature estimation of measurement noise. Finally, a co-simulation using Carsim and Matlab/Simulink confirms the algorithm is effective and reasonable. The simulation results demonstrate that the proposed algorithm, compared with the UKF algorithm, increases estimation accuracy by 19.13%, 32.8%, and 39.46% in yaw rate, side-slip angle, and longitudinal velocity, respectively. This is because the proposed algorithm adaptively adjusts the measurement noise covariance matrix, which can estimate the state parameters of the vehicle more accurately.

CarSim

Observer (physics)

Unscented transform

Yaw

10.3390/wevj14070167

Cite

Citations (7)

Research on Obstacle Avoidance Tracking Control Method for Unmanned Vehicles Based on Model Predictive Control in Variable Time Domain

2022 4th International Conference on Intelligent Control, Measurement and Signal Processing (ICMSP) (2023)

Zhong Qin-hong Yong Chen Yuecheng Li

Aiming at the obstacle avoidance tracking control problem of unmanned vehicles, a model predictive controller is designed in this paper. The controller adopts a layered structure. The upper layer performs local path planning during obstacle avoidance based on the nonlinear vehicle point-mass model, and the lower layer implements path tracking based on the monorail vehicle dynamics model. Through the joint simulation of Carsim and Matlab, the genetic algorithm is used to optimize the controller parameters, and the performance of the controller is tested at different speeds. The simulation results show that: compared with the controller before improvement, the controller is at 30km/h, 60km/h and 80km/h, the lateral deviation after obstacle avoidance decreased by 34.8%, 47.2%, and 13% respectively, and the yaw angle deviation decreased by 18.7%, 49.9%, and 29% respectively, Finally, smooth obstacle avoidance at various speeds is achieved through layered control.

CarSim

Obstacle avoidance

Model Predictive Control

Yaw

10.1109/icmsp58539.2023.10171048

Cite

Citations (1)

Energy management based on reinforcement learning with double deep Q-learning for a hybrid electric tracked vehicle

Applied Energy (2019)

Xuefeng Han Hongwen He Jingda Wu Jiankun Peng Yuecheng Li

Q-learning

Benchmark (surveying)

10.1016/j.apenergy.2019.113708

Cite

Citations (210)

Learning of EMSs in Continuous State Space-Discrete Action Space

Synthesis lectures on advances in automotive technology (2022)

Yuecheng Li Hongwen He

When applying Q-learning or DP for energy management, their solving efficiency may be limited by the premise of state space discretization. For data-driven, end-to-end learning-based EMSs, we desire not only to reduce their reliance on empirical parameter tuning, but also a higher requirement for its data mining capability, i.e., the energy-saving control schemes should be learned quickly from multidimensional environmental information. The DQN method, as an early breakthrough in DRL, combines deep learning with Q-learning to construct action-value functions and achieve continuity of state space while solving MDP problems, which dramatically expands the application of reinforcement learning in complex environments. On the other hand, DQN directly traverses and searches the entire action space to find the maximum action value when making decisions, such a strategy enables a fairly concise and efficient learning algorithm in discrete action spaces. Therefore, to address energy management problems with continuous state—discrete action spaces, this chapter describes an energy management method based on deep Q-learning, and further conduct research on its learning stability, optimization, and adaptability on diverse driving cycles.

Q-learning

Adaptability

10.1007/978-3-031-79206-9_3

Cite

Citations (0)

Deep Reinforcement Learning-Based Energy Management for Hybrid Electric Vehicles

Synthesis lectures on advances in automotive technology (2022)

Yuecheng Li Hongwen He

The urgent need for vehicle electrification and improvement in fuel efficiency has gained increasing attention worldwide. Regarding this concern, the solution of hybrid vehicle systems has proven its

10.1007/978-3-031-79206-9

Cite

Citations (0)

Continuous reinforcement learning of energy management with deep Q network for a power split hybrid electric bus

Applied Energy (2018)

Jingda Wu Hongwen He Jiankun Peng Yuecheng Li Zhanjiang Li

Q-learning

10.1016/j.apenergy.2018.03.104

Cite

Citations (346)

Learning of EMSs in Continuous State-Continuous Action Space

Synthesis lectures on advances in automotive technology (2022)

Yuecheng Li Hongwen He

Depending on the modeling approach or configuration types, some energy management problems modeled as the MDP have single or multiple continuous control actions. For such problems, traditional optimization methods usually adopt discretization solutions, but their application scenarios and computational amount are vulnerable to dimensional issues. The study of continuous energy management methods that can directly search for the optimal policy in the continuous state—continuous action space will be of great benefit to the expansion of application range to a certain extent. In addition, based on the continuous energy management method, this chapter also introduces a PHEV energy management solution integrating history cumulative trip information (HCTI) to improve the EMS learning effect across a wider feasible domain of SoC.

Continuous modelling

Continuous monitoring

Continuous variable

10.1007/978-3-031-79206-9_4

Cite

Citations (0)

An Improved Energy Management Strategy for Hybrid Electric Vehicles Integrating Multistates of Vehicle-Traffic Information

IEEE Transactions on Transportation Electrification (2021)

Hongwen He Yunlong Wang Jianwei Li Jingwei Dou Renzong Lian

This study aims to answer the key question for hybrid electric vehicles (HEVs) on how to manage the power flow in HEVs with recent intelligent and electrified upgrades in automotive industries. The new energy management strategy (EMS) needs to fuse both the physic and cyber systems, reflecting the dynamic vehicle system in the physical layer, as well as taking full advantage of the outside information in the cyber layer. Given that, this article proposes the cyber-physical system (CPS)-based EMS using deep reinforcement learning (DRL) in two different types of vehicles [hybrid electric bus (HEB) and Prius]. Under the proposed framework, exploratory training is carried out for the EMS which is mediated by DRL algorithms, expert prior knowledge and multistate of traffic information. Then, the prior valid knowledge trained by HEB is applied to Prius based on deep transfer learning, accelerating the new EMS convergence and ensuring the same initial parameters of the two vehicles' deep neural networks. Moreover, the cyber information is decoupled from the vehicle itself, for the first time being visualized for comparative analysis. The results show a significant improvement by considering traffic states (TS) and using dynamic programming (DP) as a benchmark with 6.94% fuel economy improvement for deep deterministic policy gradients (DDPG) test results and 8.12% for deep Q-learning (DQL), respectively. The decoupling analysis distinguished the effect of various TS for the HEB and Prius due to their different characteristics in vehicle service, driving style, and vehicle structures, which further verifies the effectiveness of the proposed EMS.

Benchmark (surveying)

Energy management system

Hybrid vehicle

10.1109/tte.2021.3054896

Cite

Citations (82)