检索结果-内蒙古大学图书馆

Online learning of the Kalman Filter With Logarithmic Regret

ieee TRANSACTIONS ON AUTOMATIC control 2023年第5期68卷 2774-2789页

作者： Tsiamis, Anastasios Pappas, George J. Swiss Fed Inst Technol Automat Control Lab Zurich Switzerland Univ Penn Dept Elect & Syst Engn Philadelphia PA 19104 USA

In this article, we consider the problem of predicting observations generated online by an unknown, partially observable linear system, which is driven by Gaussian noise. In the linear Gaussian setting, the optimal predictor in the mean square error sense is the celebrated Kalman filter, which can be explicitly computed when the system model is known. When the system model is unknown, we have to learn how to predict observations online based on finite data, suffering possibly a nonzero regret with respect to the Kalman filter's prediction. We show that it is possible to achieve a regret of the order of $\text{poly}\log (N)$ with high probability, where $N$ is the number of observations collected. This is achieved using an online least-squares algorithm, which exploits the approximately linear relation between future observations and past observations. The regret analysis is based on the stability properties of the Kalman filter, recent statistical tools for finite sample analysis of system identification, and classical results for the analysis of least-squares algorithms for time series. Our regret analysis can also be applied to other predictors, e.g., multiple step-ahead prediction, or prediction under exogenous inputs including closed-loop prediction. A fundamental technical contribution is that our bounds hold even for the class of nonexplosive systems (including marginally stable systems), which was not addressed before in the case of online prediction.

关键词： Kalman filters Prediction algorithms Predictive models Stability analysis System identification Upper bound Regulators Adaptive estimation Kalman filter online learning statistical learning

来源：评论

学校读者我要写书评

暂无评论

Deep learning Approach for EEG Classification in Lower-Limb Movement Phases: Towards Enhanced Brain-Computer Interface control 20

Deep Learning Approach for EEG Classification in Lower-Limb ...

引用

20th ieee/ASME International conference on Mechatronic and Embedded systems and Applications (MESA)

作者： Blanco-Diaz, Cristian Felipe Guerrero-Mendez, Cristian David Furtado, Guilherme Rangel Bastos-Filho, Teodiano de Andrade, Rafhael Milanezi Scuola Super Sant Anna Dept Excellence & Ai BioRobot Inst Pisa PI Italy Univ Estadual Campinas Dept Elect & Biomed Engn Campinas SP Brazil Univ Fed Espirito Santo Dept Elect Engn Vitoria ES Brazil Univ Fed Espirito Santo Posgrad Program Elect Engn Vitoria ES Brazil Univ Fed Espirito Santo Grad Program Mech Engn Vitoria ES Brazil

ISBN: (纸本)9798331516246;9798331516239

Embedded Brain-Computer Interfaces (BCI) have emerged as an alternative for rehabilitation/assistance of people with neuromotor impairments. data-driven methods are usually implemented to design controllers to (de)activate these systems using electroencephalography (EEG), which may limit usability and degrees of freedom. Thus, the development of more adaptive classifiers remains a challenge. This study focuses on the implementation of a Deep learning (DL) method to classify EEG signals during different phases of gait and pedaling. To address this, public and in-house databases were used, which contain information on EEG and kinematic data during walking on a treadmill and pedaling on a minibike. Kinematic labels of the phases were extracted using k-means and classified by a Convolutional Neural Network (CNN), which demonstrated accuracy above 60%. Furthermore, a negative correlation was found between velocities and classifier performance metrics, suggesting that faster movements can hinder EEG decoding. Our proposal is a starting point to identify current states of lower-limb movement, which can be used for designing state machine-based controllers of robotic BCIs with rehabilitation or assistance purposes, such as motorized mini exercise bikes or leg exoskeletons.

关键词： Brain-Computer Interfaces Embedded Human-Machine Interfaces Gait Assistance Pedaling Rehabilitation data-driven Methods

来源：评论

学校读者我要写书评

暂无评论

Neural Network Models for Predicting Magnetization Surface Switched Reluctance Motor: Classical, Radial Basis Function, and Physics-Informed Techniques

引用

ieee ACCESS 2025年 13卷 54987-54996页

作者： Demidova, Galina Justo, Jackson J. Lukichev, Dmitry Poliakov, Nikolai Anuchin, Alecksey ITMO Univ Fac Control Syst & Robot St Petersburg 197101 Russia Moscow Power Engn Inst Elect Drives Dept Moscow 111250 Russia

Neural networks have increasingly been utilized in electric drive systems to enhance modeling, control, and optimization. These data-driven techniques enable accurate predictions of complex nonlinear behaviors, including the magnetization characteristics of electric machines. This paper investigates the use of neural networks for predicting magnetization surfaces in switched reluctance motors, a key aspect of their design and operational efficiency. Three neural networks-based methods are studied: classical neural networks, radial basis function neural networks, and physics-informed neural networks. Experimental data from a 7.5 kW switched reluctance motor are used to assess the capabilities of each approach. Moreover, the study evaluates predictive accuracy, computational requirements, and the ability to reflect physical dynamics. Results demonstrate that classical neural networks and radial basis function networks can model magnetization surface, but with inaccuracy due to failure to comply with flux behavior, with radial basis function networks excelling in computational efficiency. Physics-informed neural networks achieve the highest accuracy by integrating physical laws into the learning process. This research highlights the potential of neural networks techniques in advancing switched reluctance motors modeling, paving the way for improved electric drive systems.

关键词： Accuracy Rotors Neural networks Couplings Training Torque Magnetization Biological neural networks Switched reluctance motors Predictive models Feedforward neural networks radial basis function networks physics-informed neural networks motors dc motors switched reluctance motors surface

来源：评论

学校读者我要写书评

暂无评论

data-driven Model Predictive control-Based Proactive Scheduling for Commercial Microgrid Considering Anomaly Detection

引用

ieee systems JOURNAL 2023年第1期17卷 443-454页

作者： Thanh, Vo-Van Su, Wencong Univ Michigan Dearborn Dearborn MI 48128 USA

One of the challenging tasks for commercial microgrids with distributed generation is to ensure critical loads resilience during high-impact, low-frequency extreme events. This article proposes a data-driven model predictive control-based proactive scheduling strategy for microgrids subject to extreme weather events. The goal is to maximize the utility value of the building by realizing the difference between the comfort level of critical loads and the total operating cost of the commercial microgrid during the planning horizon. A two-layer framework is proposed to solve the considered problem. First, the aggregate demand load measured by smart meters is processed using a combined long short-term memory and fuzzy neural network algorithm to eliminate anomaly data. Then, the obtained real data are applied through the deep-Q network method to seek appropriate proactive scheduling solutions. However, the rewards obtained from the deep-Q network method are random because of the uncertainties in extreme events. Therefore, robust optimization namely, robust deep reinforcement learning, is incorporated to enhance the robustness of the solutions. The proposed framework ensures the optimality and feasibility of the commercial microgrid's proactive scheduling strategy within the uncertainties in extreme events and the different intensity levels of anomalous data. Extensive simulation results prove the effectiveness of the proposed optimization model and algorithm for enhancing commercial microgrid resilience in extreme events. Furthermore, it also shows good ability in detecting smart meters anomalous data, with a stable convergence value of the performance index, to secure commercial microgrid smart meter data.

关键词： Smart meters Cogeneration HVAC Microgrids Resilience Real-time systems Costs Commercial microgrid deep Q-network demand response long short-term memory neural network proactive scheduling resilience robust deep reinforcement learning robust optimization

来源：评论

学校读者我要写书评

暂无评论

Multi-Classification Machine learning Algorithm for Predicting Pavement Maintenance Treatments

Multi-Classification Machine Learning Algorithm for Predicti...

引用

ieee International conference on Automatic control and Intelligent systems (I2CACIS)

作者： Milad, Abdalrhman Ali, Abdualmtalab Abdualaziz Yusoff, Nur Izzi Md Univ Nizwa Dept Civil Engn Nizwa Oman Azzytuna Univ Dept Civil Engn Tarhuna Libya Univ Kebangsaan Malaysia Dept Civil Engn Ukm Bangi Selangor Malaysia

ISBN: (纸本)9798350372113;9798350372106

Managing pavement assets poses a challenge for many highways worldwide, primarily attributed to funding restrictions and the absence of organized arrangement tools for maintenance and rehabilitation (M&R) activities to preserve road pavement assets. While pavement management systems (PMS) aid in predictions, the decision of when and how to administer treatments ultimately lies with pavement engineers. A model was developed to predict pavement treatment types with four input parameters: density, severity, average daily traffic (ADT) and road functional classification. The output parameters of the model correspond to various treatment techniques. The construction of this predictive model involved utilizing multi-classification machine learning (ML) algorithms, including support vector machine (SVM), Naive Bayes, K-Nearest Neighbour, and Random Forests. Furthermore, the model highlights the data categories that exert the most significant impact on the predictions. The accuracy of the model is noteworthy, with 81% superior for predicting untreated segments and exceeding 72% for predicting treated segments.

关键词： PMS Machine learning Decision Making Treatment Classification Prediction Pavement Maintenance

来源：评论

学校读者我要写书评

暂无评论

Hierarchical Coordination of Networked-Microgrids Toward Decentralized Operation: A Safe Deep Reinforcement learning Method

引用

ieee TRANSACTIONS ON SUSTAINABLE ENERGY 2024年第3期15卷 1981-1993页

作者： Xia, Yang Xu, Yan Feng, Xue Nanyang Technol Univ Ctr Power Engn Sch Elect & Elect Engn Singapore 639798 Singapore Singapore Inst Technol Engn Cluster Singapore 138683 Singapore

Multiple individual microgrids can be integrated as a networked microgrid system for enhanced technical and economic performance. In this paper, a two-stage data-driven method is proposed to hierarchically coordinate individual microgrids towards decentralized operation in a networked microgrid (NMG) system. The first stage schedules active power outputs of micro-turbines and energy storage systems (ESSs) on an hourly basis for energy balancing and cost minimization, where ESSs are controlled by a local P/SoC droop scheme. In the second stage, the reactive power outputs of PV inverters are dispatched every three minutes based on a Q/V droop controller, aiming to reduce network power losses and regulate the voltage under real-time uncertainties. At offline training stage, a multi-agent deep reinforcement learning model is trained to learn an optimal coordination policy, enhanced by a safety model framework. For online application, the trained agent can work locally in a decentralized manner without information exchanges, and the safety model can also be applied to monitor and guide online actions for safety compliance. Numerical test results validate the effectiveness and advantages of the proposed method.

关键词： Microgrids Safety Voltage control Energy management Reactive power Inverters Costs Networked-microgrid hierarchical coordination decentralized operation deep reinforcement learning safety model

来源：评论

学校读者我要写书评

暂无评论

learning Adaptive Subspace Differential Evolution by Evolution Strategies 6

Learning Adaptive Subspace Differential Evolution by Evoluti...

引用

6th International conference on data-driven Optimization of Complex systems (DOCS)

作者： Zhang, Haotian Shi, Jialong Sun, Jianyong Xu, Zongben Xi An Jiao Tong Univ Frontier Inst Sci & Technol Xian Peoples R China Xi An Jiao Tong Univ Sch Math & Stat Xian Peoples R China

ISBN: (纸本)9798350377859;9798350377842

The subspace method is one of the most used optimization methods, which means optimizing not all elements of the decision variable but only part of it. Subspace methods can obtain a faster convergence rate on objective functions with special structures, such as sparse objective functions. The core idea of the subspace method is selecting the proper subspace (i.e. elements) to be optimized in each iteration, which is always selected by hand. Recently the idea has been borrowed in some studies of evolutionary algorithms, however, the subspace-selecting mechanism is still handcrafted. In our opinion, the selecting mechanism can be learned from optimizing related optimization problems. In this paper, we propose an adaptive subspace differential evolution algorithm, whose subspace is selected by a graph neural network (GNN). The GNN is trained by Natural Evolution Strategies. The experiments on the CEC 2018 test suite show that the proposed method can perform significantly better than classic differential evolution and the proposed method can generalize on problems with different dimensions.

关键词： Differential evolution Parameter control learning to optimize

来源：评论

学校读者我要写书评

暂无评论

Dynamic Fault Reconfiguration of Distribution Networks in Ship Power systems Based on Deep Reinforcement learning Approach

引用

ieee TRANSACTIONS ON TRANSPORTATION ELECTRIFICATION 2024年第3期10卷 7076-7089页

作者： Shang, Chengya Fu, Lijun Bao, Xianqiang Xiao, Haipeng Xu, Xinghua Liu, Lufeng Naval Univ Engn Natl Key Lab Electromagnet Energy Wuhan 430033 Peoples R China

When a fault occurs, driven by the reconfiguration strategy, the topology of the ship power system (SPS) will be changed to isolate the fault area and restore the lost load. However, traditional optimization methods have limitations, such as getting stuck in suboptimal solutions or not offering real-time solutions. This article proposes a novel deep reinforcement learning (DRL) method to address the dynamic fault reconfiguration problem in real time for SPS. While considering the load priorities, the fault reconfiguration model is formulated with the goal of the maximum weighted load power and the minimum switching action. Then, a deep Q-network (DQN) combined with an action mask mechanism, what is called the DQN-mask algorithm, is applied to optimize the switch action. The proposed method enables end-to-end control from the fault node data to the switch sequences and has migration capability in different scenarios. Two case studies are analyzed based on historical fault datasets of medium voltage dc (MVDC) SPS. The numerical results verify the effectiveness, real-time performance, migration, and scalability of the proposed DQN-mask algorithm.

关键词： Switches Power system dynamics Circuit faults Real-time systems Power supplies Optimization Distribution networks Action mask mechanism directed graph dynamic fault reconfiguration reinforcement learning (RL) ship power system (SPS)

来源：评论

学校读者我要写书评

暂无评论

Adaptive Event-Triggered control for Uncertain Nonlinear Full-State Constrained CPSs Under Deception Attacks

引用

ieee TRANSACTIONS ON CYBERNETICS 2025年第4期55卷 1893-1902页

作者： Zhang, Jiaming Niu, Ben Wang, Yueying Zhao, Xudong Beihang Univ Sch Automat Sci & Elect Engn Beijing 100190 Peoples R China Shandong Normal Univ Sch Informat Sci & Engn Jinan 250014 Peoples R China Sichuan Univ Sch Elect Engn Chengdu 610065 Peoples R China Shanghai Univ Sch Mechatron Engn & Automat Shanghai 200444 Peoples R China Dalian Univ Technol Fac Elect Informat & Elect Engn Dalian 116023 Peoples R China

This article presents several adaptive event-triggered control (ETC) strategies for a class of uncertain nonlinear cyber-physical systems (CPSs) under constant or time-varying full-state constraints, as well as deception attacks via the senior network. Initially, the system under investigation is reformulated into a new system that encompasses both the original system state and the compromised state, enabling the use of compromised states for feedback control. Subsequently, two innovative asymptotic integral barrier Lyapunov functions (IBLFs) are developed by directly imposing constraints on the compromised state, thereby eliminating the necessity to convert state constraints into error constraints as required by traditional BLFs-based methods. Furthermore, controllers designed using relative/switched threshold event-triggered strategies ensure that all signals within the entire closed-loop system remain bounded, that the constant or time-varying full-state constraints are not breached, and that asymptotic stability is attained without Zeno behavior. Ultimately, simulation results validate the efficacy of the proposed strategies through a practical example.

关键词： Asymptotic stability Nonlinear systems Event detection Time-varying systems Stability criteria Power system stability Electronic mail Robots Lyapunov methods Feedback control Adaptive control cyber-physical systems (CPSs) full-state constraints integral barrier Lyapunov functions (IBLFs) sensor attacks

来源：评论

学校读者我要写书评

暂无评论

Online Koopman Operator-Based Feedforward Compensation Strategy for Fast Tool Servos With Robust High-Bandwidth control

引用

ieee TRANSACTIONS ON INDUSTRIAL ELECTRONICS 2025年第3期72卷 2958-2967页

作者： Meng, Yixuan Li, Linlin Wang, Xiangyuan Huang, Wei-Wei Wu, Edmond Q. Zhu, Limin Shanghai Jiao Tong Univ Sch Mech Engn State Key Lab Mech Syst & Vibrat Shanghai 200240 Peoples R China Weill Cornell Med Dept Anesthesiol New York NY 10065 USA Shanghai Jiao Tong Univ Dept Comp Sci & Engn Shanghai 200240 Peoples R China

To improve the servo accuracy for fast tool servo (FTS) applications, this article presents an online Koopman operator-based feedforward compensation strategy and combines it with the robust high-bandwidth controller. The robust high-bandwidth control is achieved by employing an optimized dual-loop feedback controller integrated with a disturbance observer. To realize the almost zero dynamic lag tracking, a Koopman operator-based data-driven learning methodology is proposed for real-time tracking error prediction and compensation. Thanks to the high computational efficiency, this methodology could fulfill the error prediction in a very short time, and thus is applicable to online applications extensively encountered in FTS operations. The error elimination mechanism is deduced and analyzed theoretically. The experimental results indicate that the root-mean-squared error is only 4.2 nm for tracking a random nonuniform rational B-splines curve with a travel range of 10 mu m. The proposed control strategy enables robust high-bandwidth and almost zero dynamic lag tracking for general and real-time generated trajectories, which is significant for improving the manufacturing efficiency and accuracy of the FTS.

关键词： Accuracy Tracking Real-time systems Feedforward systems Servomotors Trajectory Bandwidth Fast tool servo (FTS) Koopman operator robust high-bandwidth control

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：