Reinforcement Learning–Based Energy Management Strategy for a Hybrid Electric Tracked Vehicle

This paper presents a reinforcement learning (RL)–based energy management strategy for a hybrid electric tracked vehicle. A control-oriented model of the powertrain and vehicle dynamics is first established. According to the sample information of the experimental driving schedule, statistical charac...

Full description

Bibliographic Details
Main Authors:	Teng Liu, Yuan Zou, Dexing Liu, Fengchun Sun
Format:	Article
Language:	English
Published:	MDPI AG 2015-07-01
Series:	Energies
Subjects:	reinforcement learning (RL) hybrid electric tracked vehicle (HETV) Q-learning algorithm Dyna algorithm dynamic programming (DP) stochastic dynamic programming (SDP)
Online Access:	http://www.mdpi.com/1996-1073/8/7/7243

id	doaj-fb147cff328a45c6b73e1cd7df0b8342
record_format	Article
spelling	doaj-fb147cff328a45c6b73e1cd7df0b83422020-11-24T22:32:10ZengMDPI AGEnergies1996-10732015-07-01877243726010.3390/en8077243en8077243Reinforcement Learning–Based Energy Management Strategy for a Hybrid Electric Tracked VehicleTeng Liu0Yuan Zou1Dexing Liu2Fengchun Sun3Collaborative Innovation Center of Electric Vehicles in Beijing, School of Mechanical Engineering, Beijing Institute of Technology, Beijing 100081, ChinaCollaborative Innovation Center of Electric Vehicles in Beijing, School of Mechanical Engineering, Beijing Institute of Technology, Beijing 100081, ChinaCollaborative Innovation Center of Electric Vehicles in Beijing, School of Mechanical Engineering, Beijing Institute of Technology, Beijing 100081, ChinaCollaborative Innovation Center of Electric Vehicles in Beijing, School of Mechanical Engineering, Beijing Institute of Technology, Beijing 100081, ChinaThis paper presents a reinforcement learning (RL)–based energy management strategy for a hybrid electric tracked vehicle. A control-oriented model of the powertrain and vehicle dynamics is first established. According to the sample information of the experimental driving schedule, statistical characteristics at various velocities are determined by extracting the transition probability matrix of the power request. Two RL-based algorithms, namely Q-learning and Dyna algorithms, are applied to generate optimal control solutions. The two algorithms are simulated on the same driving schedule, and the simulation results are compared to clarify the merits and demerits of these algorithms. Although the Q-learning algorithm is faster (3 h) than the Dyna algorithm (7 h), its fuel consumption is 1.7% higher than that of the Dyna algorithm. Furthermore, the Dyna algorithm registers approximately the same fuel consumption as the dynamic programming–based global optimal solution. The computational cost of the Dyna algorithm is substantially lower than that of the stochastic dynamic programming.http://www.mdpi.com/1996-1073/8/7/7243reinforcement learning (RL)hybrid electric tracked vehicle (HETV)Q-learning algorithmDyna algorithmdynamic programming (DP)stochastic dynamic programming (SDP)
collection	DOAJ
language	English
format	Article
sources	DOAJ
author	Teng Liu Yuan Zou Dexing Liu Fengchun Sun
spellingShingle	Teng Liu Yuan Zou Dexing Liu Fengchun Sun Reinforcement Learning–Based Energy Management Strategy for a Hybrid Electric Tracked Vehicle Energies reinforcement learning (RL) hybrid electric tracked vehicle (HETV) Q-learning algorithm Dyna algorithm dynamic programming (DP) stochastic dynamic programming (SDP)
author_facet	Teng Liu Yuan Zou Dexing Liu Fengchun Sun
author_sort	Teng Liu
title	Reinforcement Learning–Based Energy Management Strategy for a Hybrid Electric Tracked Vehicle
title_short	Reinforcement Learning–Based Energy Management Strategy for a Hybrid Electric Tracked Vehicle
title_full	Reinforcement Learning–Based Energy Management Strategy for a Hybrid Electric Tracked Vehicle
title_fullStr	Reinforcement Learning–Based Energy Management Strategy for a Hybrid Electric Tracked Vehicle
title_full_unstemmed	Reinforcement Learning–Based Energy Management Strategy for a Hybrid Electric Tracked Vehicle
title_sort	reinforcement learning–based energy management strategy for a hybrid electric tracked vehicle
publisher	MDPI AG
series	Energies
issn	1996-1073
publishDate	2015-07-01
description	This paper presents a reinforcement learning (RL)–based energy management strategy for a hybrid electric tracked vehicle. A control-oriented model of the powertrain and vehicle dynamics is first established. According to the sample information of the experimental driving schedule, statistical characteristics at various velocities are determined by extracting the transition probability matrix of the power request. Two RL-based algorithms, namely Q-learning and Dyna algorithms, are applied to generate optimal control solutions. The two algorithms are simulated on the same driving schedule, and the simulation results are compared to clarify the merits and demerits of these algorithms. Although the Q-learning algorithm is faster (3 h) than the Dyna algorithm (7 h), its fuel consumption is 1.7% higher than that of the Dyna algorithm. Furthermore, the Dyna algorithm registers approximately the same fuel consumption as the dynamic programming–based global optimal solution. The computational cost of the Dyna algorithm is substantially lower than that of the stochastic dynamic programming.
topic	reinforcement learning (RL) hybrid electric tracked vehicle (HETV) Q-learning algorithm Dyna algorithm dynamic programming (DP) stochastic dynamic programming (SDP)
url	http://www.mdpi.com/1996-1073/8/7/7243
work_keys_str_mv	AT tengliu reinforcementlearningbasedenergymanagementstrategyforahybridelectrictrackedvehicle AT yuanzou reinforcementlearningbasedenergymanagementstrategyforahybridelectrictrackedvehicle AT dexingliu reinforcementlearningbasedenergymanagementstrategyforahybridelectrictrackedvehicle AT fengchunsun reinforcementlearningbasedenergymanagementstrategyforahybridelectrictrackedvehicle
_version_	1725734742425337856

Reinforcement Learning–Based Energy Management Strategy for a Hybrid Electric Tracked Vehicle

Similar Items