检索结果-内蒙古大学图书馆

36th Chinese Control and Decision Conference (CCDC)

作者： Gao, Yingning Ma, Kemao Harbin Inst Technol Control & Simulat Ctr Harbin Peoples R China

ISBN: (纸本)9798350387780;9798350387797

In this paper, online reinforcement learning solution to finite-horizon non-cooperative differential games is investigated. The main challenges are that, the Hamilton-Jacobi-Isaacs equation is time-varying, and in terms of parameter convergence, conventional learning approaches, with persistence of excitation condition satisfied, may cause undesired oscillations in the system states and deteriorate system performance, especially in finitehorizon situations. Regarding these, we develop a finite-horizon adaptive dynamic programming algorithm for continuous-time systems, where the idea of simulation of experience in concurrent learning technique is adopted to guarantee fast online parameter convergence. Boundedness of the system states and weight estimation errors is proved. Simulation results demonstrate the effectiveness of the presented method.

关键词： finite-horizon differential games reinforcement learning adaptive dynamic programming concurrent learning

来源：评论

学校读者我要写书评

暂无评论

Autonomous learning with Automatically Created Models and a Novel Model Selection

Autonomous Learning with Automatically Created Models and a ...

引用

ieee symposium Series on Computational Intelligence (ieee SSCI)

作者： Bharatia, Harshal, V ACM Student Member Plano United States

ISBN: (纸本)9781728190488

An autonomous learning approach is presented here for expansive problem domains that may undergo frequent changes. It is hard to train and adapt learning-models to changes when the problem domain is very large. With the autonomous learning approach, a system performs knowledge refinement automatically by determining what can be improved and trains itself without explicit guidance to do so. It uses novel techniques to automatically create the learning-models and select an optimal model for each prediction. It automatically builds dynamic ensembles of models that incorporate specific improvisations, such as improvements based on past knowledge or specific trends reflected in the results, and overshadows sub-optimal portions of existing model. A reinforcement learning based model selector identifies which model is optimal for handling a request using a novel approach with automatically created hierarchical states. As a result, smart adaptive solutions that go beyond initial training become possible. This also alleviates the need for big complex monolithic models that require extensive training, automatically adapts to changes in the domain and offers better control over its performance. Experimental results show this approach learns quickly, adapts to changes very rapidly and performs quite well against prevalent learning methods.

关键词： dynamic ensembles reinforcement learning supervised learning autonomous learning model selection

来源：评论

学校读者我要写书评

暂无评论

Online Off-Policy reinforcement learning for Optimal Control of Unknown Nonlinear Systems Using Neural Networks

引用

ieee TRANSACTIONS ON SYSTEMS MAN CYBERNETICS-SYSTEMS 2024年第8期54卷 5112-5122页

作者： Zhu, Liao Wei, Qinglai Guo, Ping Beijing Normal Univ Int Acad Ctr Complex Syst Zhuhai 519087 Peoples R China Beijing Normal Univ Sch Syst Sci Beijing 100875 Peoples R China Chinese Acad Sci Inst Automat State Key Lab Multimodal Artificial Intelligence S Beijing 100190 Peoples R China Univ Chinese Acad Sci Sch Artificial Intelligence Beijing 100049 Peoples R China Macau Univ Sci & Technol Inst Syst Engn Macau Peoples R China

In this article, a real-time online off-policy reinforcement learning (RL) method is developed for the optimal control problem of unknown continuous-time nonlinear systems. First, by applying the temporal difference technique to the iterative procedure of off-policy RL, the iterative value function and the iterative policy input can be learned in real-time online. It is proven that the fitting error of neural network (NN) weights is exponentially convergent in each iteration. Second, a model-free Hamilton-Jacobi-Bellman equation (MF-HJBE) is deduced by taking the limit of the iterative procedure of off-policy RL. In this manner, it not only eliminates system dynamics in the classical HJBE, but also vanishes the iteration index. By applying temporal difference to the MF-HJBE, a real-time online tuning rule is designed to learn the optimal value function and the optimal policy input. It is proven that the fitting error of NN weights caused by the real-time online tuning rule is exponentially convergent. Note that the two online tuning rules, the iterative one and the real-time one, use only current and previous state data extracted from system trajectories. Meanwhile, it is proven using the Lyapunov's direct method that the system solution is uniformly ultimately bounded. Finally, simulation results demonstrate the validity of the proffered method.

关键词： adaptive dynamic programming nonlinear systems online learning optimal control reinforcement learning (RL)

来源：评论

学校读者我要写书评

暂无评论

Energy- and Cost-Efficient Transmission Strategy for UAV Trajectory Tracking Control: A Deep reinforcement learning Approach

引用

ieee INTERNET OF THINGS JOURNAL 2023年第10期10卷 8958-8970页

作者： Zhang, Minkai Wu, Shaohua Jiao, Jian Zhang, Ning Zhang, Qinyu Harbin Inst Technol Shenzhen Dept Elect & Informat Engn Shenzhen 518055 Peoples R China Harbin Inst Technol Shenzhen Guangdong Prov Key Lab Aerosp Commun & Networking Shenzhen 518055 Peoples R China Peng Cheng Lab Dept Broadband Commun Shenzhen 518055 Peoples R China Univ Windsor Dept Elect & Comp Engn Windsor ON N9B 3P4 Canada

this article, we consider a networked control system (NCS) with network-induced delay, in which the control center needs to control the remote unmanned aerial vehicle (UAV) to complete the trajectory tracking task. The sensor of the controlled UAV adopts the event-triggered mechanism, while the control center uses the adaptive dynamic programming (ADP) based tracking control method to generate control actions. The application of the ADP method brings new transmission options, i.e., the control center can choose to transmit control action or neural network model. Considering the fundamental tradeoff between these two transmission options with different transmission energy consumption and tracking cost, we formulate the joint optimization problem as a Markov decision process (MDP). Due to the continuous value of state in MDP, we propose the deep Q-network (DQN)-based strategy, which uses the reinforcement learning (RL) algorithm, specifically DQN. Besides, we further propose a greedy strategy by calculating the instantaneous expected cost. Simulation results show that DQN-based strategy has better performance but depends on the training process, while greedy strategy is suboptimal but easy to compute. Besides, compared with the benchmark strategies, the proposed strategies can achieve a better compromise in the long-term average energy consumption and tracking cost by adjusting the value of the weighted factor. Furthermore, by comparing the difference of transmission decisions in the proposed strategies, we show that the proper transmission sequence in DQN-based strategy can reduce the tracking cost and transmission energy at the same time.

关键词： Costs Control systems Artificial neural networks Energy consumption Autonomous aerial vehicles Internet of Things Trajectory tracking adaptive dynamic programming (ADP) networked control system (NCS) transmission strategy unmanned aerial vehicle (UAV)

来源：评论

学校读者我要写书评

暂无评论

Optimal Control for Multi-agent Systems Using Off-Policy reinforcement learning 4

Optimal Control for Multi-agent Systems Using Off-Policy Rei...

引用

4th International Conference on Control and Robotics (ICCR)

作者： Wang, Hao Chen, Zhiru Wang, Jun Lu, Lijun Li, Mingzhe Henan XJ Metering Co Ltd Xuchang Peoples R China State Grid Shandong Elect Power Co Mkt Serv Ctr Metering Ctr Jinan Peoples R China

ISBN: (纸本)9781665486415

To achieve the consensus for discrete-time multi-agent systems, an optimal control policy is designed based on off-policy reinforcement learning. By utilizing centralized learning and decentralized execution, we first define a centralized and shared value function. Then, a value iteration adaptive dynamic programming method is proposed to approach the solution of the Bellman optimality equation with convergence analysis. Furthermore, the actor-critic structure is given for the implementation purpose, where one single-critic network is given to approach the optimal centralized value function, and multi-actor networks are decentralized based on the local observation from the neighbors to obtain the optimal policy for each agent. Finally, the proposed algorithm is verified in a leader-follower consensus case.

关键词： reinforcement learning multi-agent systems adaptive dynamic programming

来源：评论

学校读者我要写书评

暂无评论

Evolutionary adaptive dynamic programming Algorithm for Converter Gas Scheduling of Steel Industry 6

Evolutionary Adaptive Dynamic Programming Algorithm for Conv...

引用

6th International symposium on Advanced Control of Industrial Processes (AdCONIP)

作者： Wang, Tianyu Wang, Linqing Zhao, Jun Wang, Wei Liu, Ying Dalian Univ Technol Sch Control Sci & Engn Dalian 116024 Peoples R China

ISBN: (纸本)9781509043972

It is significant to perform an effective scheduling of byproduct gas system in steel industry for reducing cost and protecting environment. The existing studies largely focused on extracting specific knowledge from human experience or directly optimizing the scheduling performance, which failed to provide a dynamic optimization process for making the scheduling scheme updated online. In this study, an action-dependent heuristic dynamic programming (ADHDP) framework is proposed for the Linz Donawitz converter gas (LDG) scheduling, in which the scheduling amount is calculated based on the gas system states by utilizing a Tagaki-Sugeno-Kang (TSK) fuzzy model, while a utility function is introduced in the critic network considering the time delay of the gas system to evaluate the scheduling performance over time. For achieving online learning process, the concept of a modified evolutionary algorithm is combined with the ADHDP to obtain the near-optimal scheduling policy at each time instance. To demonstrate the performance of the proposed method, the practical data coming from the energy center of a steel plant are employed. The results show that the proposed method can supply the human operators with effective solution for secure and economically justified optimization of the LDG system.

关键词： steel industry energy scheduling adaptive dynamic programming reinforcement learning evolutionary computing

来源：评论

学校读者我要写书评

暂无评论

adaptive dynamic programming and Data-Driven Cooperative Optimal Output Regulation with adaptive Observers 61

Adaptive Dynamic Programming and Data-Driven Cooperative Opt...

引用

ieee 61st Conference on Decision and Control (CDC)

作者： Qasem, Omar Jebari, Khalid Gao, Weinan Florida Inst Tech nology Dept Mech & Civil Engn Coll Engn & Sci Melbourne FL 32901 USA Florida Inst Technol Dept Aerosp Engn Coll Engn & Sci Melbourne FL 32901 USA

ISBN: (纸本)9781665467612

In this paper, a novel adaptive optimal control strategy is proposed to achieve the cooperative optimal output regulation of continuous-time linear multi-agent systems based on adaptive dynamic programming (ADP). The proposed method is different from those in the existing literature of ADP and cooperative output regulation in the sense that the knowledge of the exosystem dynamics is not required in the design of the exostate observers for those agents with no direct access to the exosystem. Moreover, an optimal control policy is obtained without the prior knowledge of the modeling information of any agent while achieving the cooperative output regulation. Instead, we use the state/input information along the trajectories of the underlying dynamical systems and the estimated exostates to learn the optimal control policy. Simulation results show the efficacy of the proposed algorithm, where both estimation errors of exosystem matrix and exostates, and the tracking errors converge to zero in an optimal sense, which solves the cooperative optimal output regulation problem.

关键词： Optimal control reinforcement learning adaptive dynamic programming cooperative optimal control cooperative optimal output regulation

来源：评论

学校读者我要写书评

暂无评论

A Biologically-Inspired Computational Model for Transformation Invariant Target Recognition

A Biologically-Inspired Computational Model for Transformati...

引用

International Joint Conference on Neural Networks

作者： Iftekharuddin, Khan M. Li, Yaqin Univ Memphis Dept Elect & Comp Engn Intelligence Syst & Image Proc Lab Memphis TN 38152 USA

ISBN: (纸本)9781424418206

Transformation invariant image recognition has been an active research area due to its widespread applications in a variety of fields such as military operations, robotics' medical practices, geographic scene analysis, and many others. One of the primary challenges is detection and recognition of objects in the presence of transformations such as resolution, rotation, translation, scale and occlusion. In this work, we investigate a biologically-inspired computational modeling approach that exploits reinforcement learning (RL) for transformation-invariant image recognition. The RL is implemented in an adaptive critic design (ACD) framework to approximate the neuro-dynamic programming. Two ACD algorithms such as Heuristic dynamic programming (HDP) and Dual Heuristic dynamic programming (DHP) are investigated and compared for transformation invariant recognition. The two learning algorithms are evaluated statistically using simulated transformations in 2-D images as well as with a large-scale UMIST 2-D face database with pose variations. Our simulations show promising results for both HDP and DHP or transformation-invariant image recognition as well as face authentication. Comparing the two algorithms, DHP outperforms HDP in learning capability, as DHP takes fewer steps to perform a successful recognition task in general. On the other hand, HDP is more robust than DHP as far as success rate across the database is concerned when applied in a stochastic and uncertain environment, and the computational complexity involved in HDP is much less.

关键词： reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

Optimal Elevator Group Control via Deep Asynchronous Actor-Critic learning

引用

ieee TRANSACTIONS ON NEURAL NETWORKS AND learning SYSTEMS 2020年第12期31卷 5245-5256页

作者： Wei, Qinglai Wang, Lingxiao Liu, Yu Polycarpou, Marios M. Chinese Acad Sci Inst Automat State Key Lab Management & Control Complex Syst Beijing 100190 Peoples R China Univ Chinese Acad Sci Sch Artificial Intelligence Beijing 100049 Peoples R China Qingdao Acad Intelligent Ind Qingdao 266109 Peoples R China Chinese Acad Sci Inst Automat Beijing 100190 Peoples R China Univ Cyprus KIOS Res & Innovat Ctr Excellence CY-1678 Nicosia Cyprus Univ Cyprus Dept Elect & Comp Engn CY-1678 Nicosia Cyprus

In this article, a new deep reinforcement learning (RL) method, called asynchronous advantage actor-critic (A3C) method, is developed to solve the optimal control problem of elevator group control systems (EGCSs). The main contribution of this article is that the optimal control law of EGCSs is designed via a new deep RL method, such that the elevator system sends passengers to the desired destination floors as soon as possible. Deep convolutional and recurrent neural networks, which can update themselves during applications, are designed to dispatch elevators. Then, the structure of the A3C method is developed, and the training phase for the learning optimal law is discussed. Finally, simulation results illustrate that the developed method effectively reduces the average waiting time in a complex building environment. Comparisons with traditional algorithms further verify the effectiveness of the developed method.

关键词： Elevators Optimal control Neural networks Backpropagation Machine learning learning (artificial intelligence) Actor– critic adaptive dynamic programming deep learning (DL) elevator group control (EGC) optimal control reinforcement learning (RL)

来源：评论

学校读者我要写书评

暂无评论

Direct adaptive control of a flexible robot using reinforcement learning

Direct adaptive control of a flexible robot using reinforcem...

引用

International Conference on Industrial Electronics, Control and Robotics

作者： Subudhi, Bidyadhar Pradhan, Santanu Kumar

ISBN: (纸本)9781424485468

This paper proposes a new adaptive control using the concept of reinforcement learning to address adaptivity for varied payload conditions for a two-link flexible manipulator (TLFM). The application of reinforcement learning has been implemented using a method called adaptive dynamic programming. Decentralized controllers for the decoupled system have been also designed using LQR technique. Then the reinforcement learning is used to tune the gains of the optimal control to adapt in terms of different payload to the manipulator end effecter. Simulation results show that proposed controller provides better end point tracking then LQR fixed gain controller. © 2010 ieee.

关键词： reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：