Sampled-data control through model-free reinforcement learning with effective experience replay

Reinforcement Learning (RL) based control algorithms can learn the control strategies for nonlinear and uncertain environment during interacting with it. Guided by the rewards generated by environment, a RL agent can learn the control strategy directly in a model-free way instead of investigating th...

Full description

Saved in:
Bibliographic Details
Main Authors: Bo Xiao, H.K. Lam, Xiaojie Su, Ziwei Wang, Frank P.-W. Lo, Shihong Chen, Eric Yeatman
Format: Article
Language:English
Published: KeAi Communications Co., Ltd. 2023-02-01
Series:Journal of Automation and Intelligence
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S2949855423000011
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850278414397734912
author Bo Xiao
H.K. Lam
Xiaojie Su
Ziwei Wang
Frank P.-W. Lo
Shihong Chen
Eric Yeatman
author_facet Bo Xiao
H.K. Lam
Xiaojie Su
Ziwei Wang
Frank P.-W. Lo
Shihong Chen
Eric Yeatman
author_sort Bo Xiao
collection DOAJ
description Reinforcement Learning (RL) based control algorithms can learn the control strategies for nonlinear and uncertain environment during interacting with it. Guided by the rewards generated by environment, a RL agent can learn the control strategy directly in a model-free way instead of investigating the dynamic model of the environment. In the paper, we propose the sampled-data RL control strategy to reduce the computational demand. In the sampled-data control strategy, the whole control system is of a hybrid structure, in which the plant is of continuous structure while the controller (RL agent) adopts a discrete structure. Given that the continuous states of the plant will be the input of the agent, the state–action value function is approximated by the fully connected feed-forward neural networks (FCFFNN). Instead of learning the controller at every step during the interaction with the environment, the learning and acting stages are decoupled to learn the control strategy more effectively through experience replay. In the acting stage, the most effective experience obtained during the interaction with the environment will be stored and during the learning stage, the stored experience will be replayed to customized times, which helps enhance the experience replay process.The effectiveness of proposed approach will be verified by simulation examples.
format Article
id doaj-art-eefac0a080e0465790b1777a01329b1f
institution OA Journals
issn 2949-8554
language English
publishDate 2023-02-01
publisher KeAi Communications Co., Ltd.
record_format Article
series Journal of Automation and Intelligence
spelling doaj-art-eefac0a080e0465790b1777a01329b1f2025-08-20T01:49:31ZengKeAi Communications Co., Ltd.Journal of Automation and Intelligence2949-85542023-02-0121203010.1016/j.jai.2023.100018Sampled-data control through model-free reinforcement learning with effective experience replayBo Xiao0H.K. Lam1Xiaojie Su2Ziwei Wang3Frank P.-W. Lo4Shihong Chen5Eric Yeatman6Department of Electrical and Electronic Engineering, Imperial College London, London SW7 2AZ, UK; Corresponding authors.Department of Engineering, King’s College London, London WC2B 4BG, UK; Corresponding authors.Department of Automation, Chongqing University, Shapingba District, Chong Qing, ChinaSchool of Engineering, Lancaster University, LA1 4YW, UKHamlyn Centre, Imperial College London, London SW7 2AZ, UKDepartment of Electrical and Electronic Engineering, Imperial College London, London SW7 2AZ, UKDepartment of Electrical and Electronic Engineering, Imperial College London, London SW7 2AZ, UKReinforcement Learning (RL) based control algorithms can learn the control strategies for nonlinear and uncertain environment during interacting with it. Guided by the rewards generated by environment, a RL agent can learn the control strategy directly in a model-free way instead of investigating the dynamic model of the environment. In the paper, we propose the sampled-data RL control strategy to reduce the computational demand. In the sampled-data control strategy, the whole control system is of a hybrid structure, in which the plant is of continuous structure while the controller (RL agent) adopts a discrete structure. Given that the continuous states of the plant will be the input of the agent, the state–action value function is approximated by the fully connected feed-forward neural networks (FCFFNN). Instead of learning the controller at every step during the interaction with the environment, the learning and acting stages are decoupled to learn the control strategy more effectively through experience replay. In the acting stage, the most effective experience obtained during the interaction with the environment will be stored and during the learning stage, the stored experience will be replayed to customized times, which helps enhance the experience replay process.The effectiveness of proposed approach will be verified by simulation examples.http://www.sciencedirect.com/science/article/pii/S2949855423000011Reinforcement learningNeural networksSampled-data controlModel-freeEffective experience replay
spellingShingle Bo Xiao
H.K. Lam
Xiaojie Su
Ziwei Wang
Frank P.-W. Lo
Shihong Chen
Eric Yeatman
Sampled-data control through model-free reinforcement learning with effective experience replay
Journal of Automation and Intelligence
Reinforcement learning
Neural networks
Sampled-data control
Model-free
Effective experience replay
title Sampled-data control through model-free reinforcement learning with effective experience replay
title_full Sampled-data control through model-free reinforcement learning with effective experience replay
title_fullStr Sampled-data control through model-free reinforcement learning with effective experience replay
title_full_unstemmed Sampled-data control through model-free reinforcement learning with effective experience replay
title_short Sampled-data control through model-free reinforcement learning with effective experience replay
title_sort sampled data control through model free reinforcement learning with effective experience replay
topic Reinforcement learning
Neural networks
Sampled-data control
Model-free
Effective experience replay
url http://www.sciencedirect.com/science/article/pii/S2949855423000011
work_keys_str_mv AT boxiao sampleddatacontrolthroughmodelfreereinforcementlearningwitheffectiveexperiencereplay
AT hklam sampleddatacontrolthroughmodelfreereinforcementlearningwitheffectiveexperiencereplay
AT xiaojiesu sampleddatacontrolthroughmodelfreereinforcementlearningwitheffectiveexperiencereplay
AT ziweiwang sampleddatacontrolthroughmodelfreereinforcementlearningwitheffectiveexperiencereplay
AT frankpwlo sampleddatacontrolthroughmodelfreereinforcementlearningwitheffectiveexperiencereplay
AT shihongchen sampleddatacontrolthroughmodelfreereinforcementlearningwitheffectiveexperiencereplay
AT ericyeatman sampleddatacontrolthroughmodelfreereinforcementlearningwitheffectiveexperiencereplay