检索结果-内蒙古大学图书馆

Explainable and Transferable Loss Meta-Learning for Zero-Touch Anticipatory network Management

IEEE TRANSACTIONS ON network AND SERVICE MANAGEMENT 2024年第3期21卷 2802-2823页

作者： Collet, Alan Bazco-Nogueras, Antonio Banchs, Albert Fiore, Marco IMDEA Networks Inst Leganes 28918 Spain Univ Carlos III Madrid Telematic Engn Dept Leganes 28911 Spain

Zero-touch network management is one of the most ambitious yet strongly required paradigms for beyond 5G and 6G mobile communication systems. Achieving full automation requires a closed loop that combines (i) network status data collection and processing, (ii) predictive capabilities based on such data to anticipate upcoming needs, and (iii) effective decision making that best addresses such future needs through proper network control and orchestration. Recent seminal works have proposed approaches to jointly implement the last two phases above via a single deep learning model trained on past network status to directly optimize future decisions. This is achieved by designing custom loss functions that directly embed the management task objective. Experiments with real-world measurement data have demonstrated that this strategy leads to substantial performance gains across diverse network management tasks. In this paper, we go one step beyond the loss tailoring schemes above, and introduce a loss meta-learning paradigm that (i) reduces the need for human intervention at model design stage, (ii) eases explainability and transferability of trained deep learning models for network management, and (iii) outperforms custom losses across a range of controlled experiments and practical use cases.

关键词： Loss meta-learning zero-touch networking anticipatory network management Loss meta-learning zero-touch networking anticipatory network management

来源：评论

学校读者我要写书评

暂无评论

SACWOM: Synergistic Adaptive Congestion Window Optimization Mechanism for Self-Clocked Algorithm

引用

IEEE ACCESS 2025年 13卷 22086-22117页

作者： Zubaydi, Haider Dhia Jagmagji, Ahmed Samir Molnar, Sandor Budapest Univ Technol & Econ Fac Elect Engn & Informat Dept Telecommun & Artificial Intelligence H-1111 Budapest Hungary

Congestion control (CC) is essential in networked systems, especially in environments with strict delay and throughput requirements. While CC algorithms like Self-Clocked Rate Adaptation for Multimedia (SCReAM) and Bottleneck Bandwidth and Round-trip propagation time (BBR) have shown progress, each faces limitations: BBR encounters difficulty in balancing responsiveness and fairness when rapid fluctuations in network conditions arise, which can lead to inefficient performance in shared environments. On the other hand, although SCReAM is designed for multimedia traffic, it struggles to dynamically adapt the congestion window to unpredictable or highly congested networks, leading to inefficient resource utilization. This paper proposes SACWOM, a hybrid congestion window optimization mechanism that combines our novel method with SCReAM's rate control and BBR's bandwidth-delay estimation to enhance throughput stability, fairness, and adaptability by dynamically adjusting the congestion window. Extensive simulations demonstrate SACWOM's significant improvements over SCReAM in managing congestion window and bytes in flight under diverse network conditions. In Phase I, SACWOM achieved up to 11.96-13.27% increase in congestion window and bytes in flight by maintaining higher data flow. Phase II analysis shows up to 20.76-22.64% improvements with optimized configurations. Finally, Phase iii, comprising 100 experiments, reveals SACWOM's dynamic adaptability, achieving up to 50-70% improvements. These results highlight SACWOM as a robust mechanism suitable for various applications across diverse network scenarios.

关键词： Heuristic algorithms Delays Bandwidth Throughput Mathematical models Adaptive systems Real-time systems Optimization Media Estimation BBR congestion control congestion window hybrid optimization rate adaptation SCReAM self-clocked

来源：评论

学校读者我要写书评

暂无评论

Data-centric predictive control with tuna swarm optimization-backpropagation neural networks for enhanced wind turbine performance

引用

RENEWABLE ENERGY 2024年 237卷

作者： Li, Wei Pandit, Ravi Kumar Cranfield Univ Ctr Life Cycle Engn & Management Wharley End England

Wind energy is a significant renewable resource, but its efficient harnessing requires advanced control systems. This study presents a Data-Centric Predictive control (DPC) system, enhanced by a Tuna Swarm OptimizationBackpropagation Neural network (TSO-BPNN) for predictive wind turbine control. It's like a smart tool that uses innovative fusion of deep learning, predictive control, and reinforcement learning. Unlike traditional control methods, the proposed approach uses real-time data to optimize turbine performance in response to fluctuating wind conditions. The system is validated using simulations on the FAST platform, which demonstrate its superior performance in two critical operational regions. Specifically, in Region II, where the objective is to maximize power extraction from the wind, the DPC achieves a 1.07 % reduction in overshoot and an improvement of 36.14 units in steadystate error compared to traditional methods. The response time remains comparable to existing Model Predictive control (MPC) strategies, ensuring real-time applicability without sacrificing efficiency. In Region iii, where maintaining constant power output is crucial, the DPC outperforms both the baseline and MPC methods, reducing overshoot by 0.58 % and improving accuracy by 17.27 units compared to the baseline method. These results highlight the effectiveness of the proposed DPC system in optimizing turbine performance under variable wind conditions, offering a significant improvement over traditional methods in both accuracy and control precision.

关键词： Deep neural networks TSO-BPNN Model predictive control Reinforcement learning Wind turbine control

来源：评论

学校读者我要写书评

暂无评论

Artificial Neural network Based Virtual Sensor for Distributed controlled DC Microgrids 7

Artificial Neural Network Based Virtual Sensor for Distribut...

引用

IEEE 7th International conference on Industrial Cyber-Physical systems (ICPS)

作者： Devakumar, Annavaram Mishra, Sukumar IIT Delhi Dept Elect Engn New Delhi India

ISBN: (纸本)9798350363029;9798350363012

The measured output current data of distributed energy resources is crucial in realizing cyber-physical DC microgrids (DCMG) and achieving distributed control objectives. This paper proposes an artificial neural network-based output current estimation at the secondary control level. The estimated output current value is used to implement distributed control successfully for the DCMG. The proposed artificial neural network (ANN) acts as a virtual sensor to alleviate problems associated with physical sensors, such as faults and biasing effects. The ANN's input features and training performance are detailed for the considered DCMG. The performance of the proposed virtual sensor-based distributed controlled DCMG is validated on an experimental hardware setup under multiple load changes and communication delays.

关键词： dc microgrid distributed control neural network virtual sensor

来源：评论

学校读者我要写书评

暂无评论

Emotion-oriented recommender system for personalized control of indoor environmental quality

引用

BUILDING AND ENVIRONMENT 2024年 254卷

作者： Kim, Hakpyeong Hong, Taehoon Yonsei Univ Dept Architecture & Architectural Engn Seoul South Korea Yonsei Univ 50 Yonsei Ro Seoul 03722 South Korea

Indoor Environmental Quality (IEQ) significantly influences health, cognitive abilities, and even moods. With the rise of smart homes and the Internet of Things (IoT), the need for personalized IEQ control has become paramount. Traditional systems, primarily focusing on explicit feedback (i.e., preference), often neglect the role of emotions in shaping this feedback. Therefore, this study presents the Emotion-oriented Recommender System for personalized control of IEQ (ERS-IEQ), based on the R -E -C -S ontology, focusing on (i) Recognizing user's continuous emotional states, (ii) Estimating emotional similarity between users, (iii) Collecting user's feedback on IEQ conditions, and (iv) Systemizing an emotion-oriented recommender system using a graph neural network. In order to confirm the predictive performance of the ERS-IEQ and verify its higher level of excellence compared to traditional recommender systems, a private dataset composed of IEQ conditions, users' explicit feedback, and users' emotional states was built via a human participant experiment using a climate chamber. As a result, the ERS-IEQ significantly enhances the recommender system's predictive performance, particularly in thermal preferences. The number of linguistic terms of emotional similarity has a profound effect on the system's predictions, with four terms proving most effective. In the near future, ERS-IEQ will play a role as a personal assistant in smart home automation, offering emotion-based recommendations. It addresses key challenges in traditional recommender systems, such as the cold start problem and rating sparsity, and ensures personalized adjustments to IEQ conditions in both private and public spaces.

关键词： Indoor environmental quality Affective computing Fuzzy logic Recommender system Graph attention network

来源：评论

学校读者我要写书评

暂无评论

performance Analysis of Trickle Timer Parameters on DIO Messages in High-Density network 15

Performance Analysis of Trickle Timer Parameters on DIO Mess...

引用

15th International conference on Information and Communication systems (ICICS)

作者： Albashayreh, Amjad Yassein, Muneer Bani Jordan Univ Sci & Technol Dept Comp Sci Irbid Jordan

ISBN: (纸本)9798331540777;9798331540760

The trickle algorithm allows network nodes to share the medium for data transmission, making it possible to exchange data in a reliable, energy-efficient, simple, and scalable manner. Consequently, trickle's operation is critical for performance metrics such as packet delivery ratio (PDR), power consumption (PC), and message overhead, particularly in high-density networks. A high-density network is designed to serve a large number of WLAN clients, and the performance analysis of network protocols can be challenging due to its large scale. This paper presents comprehensive analysis results for the RPL protocol in high-density networks with random topology to investigate the effect of PC, and control traffic overhead when the RX ratio varies with node count. Also, this paper compares the most recent RPL performance studies and conducts a detailed study of the effect of trickle timer parameters on DIO messages. The results revealed that parameters significantly affect PDR, PC, and control traffic overhead. The minimum rank with hysteresis objective function (MRHOF) reduced control traffic overhead by transmitting 325 packets with 1.05 traffic overhead, while objective function zero (OF0) showed superior power consumption at 1.5 mW.

关键词： IoT RPL OF0 MRHOF Cooja DODAGs PDR

来源：评论

学校读者我要写书评

暂无评论

A Deep Neural network Approach for Drogue Detection Using Laboratory-Chroma Key Images

引用

IEEE ACCESS 2024年 12卷 188465-188475页

作者： Miller, Dillon Mccormick, Sean Mwaffo, Violet Costello iii, Donald H. US Naval Acad Weap Robot & Control Engn Dept Annapolis MD 21402 USA Univ Maryland A James Clark Sch Engn MATRIX Lab College Pk MD 20619 USA

This study presents a framework for developing and evaluating a deep neural network model trained on a synthetic dataset of aerial refueling equipment. The data set was generated in a controlled laboratory environment with green screen backgrounds. The model's performance is rigorously compared to a counterpart trained on real-world data, revealing that the synthetic data approach not only offers a cost-effective alternative but also achieves comparable accuracy in identifying critical components for uncrewed aerial refueling missions. Despite minor classification errors, particularly with small, low-contrast objects, the results demonstrate the strong potential of synthetic data in advancing autonomous aerial refueling systems.

关键词： Synthetic data Aircraft Data models Probes Atmospheric modeling Adaptation models Accuracy Receivers Couplers Green products Autonomous aerial refueling deep neural networks synthetic datasets unmanned aerial systems

来源：评论

学校读者我要写书评

暂无评论

Learning to control Autonomous Fleets from Observation via Offline Reinforcement Learning

Learning to Control Autonomous Fleets from Observation via O...

引用

European control conference (ECC)

作者： Schmidt, Carolin Gammelli, Daniele Pereira, Francisco Camara Rodrigues, Filipe Tech Univ Denmark Dept Technol Management & Econ Lyngby Denmark Stanford Univ Dept Aeronaut & Astronaut Stanford CA 94305 USA

ISBN: (纸本)9798331540920;9783907144107

Autonomous Mobility-on-Demand (AMoD) systems are an evolving mode of transportation in which a centrally coordinated fleet of self-driving vehicles dynamically serves travel requests. The control of these systems is typically formulated as a large network optimization problem, and reinforcement learning (RL) has recently emerged as a promising approach to solve the open challenges in this space. Recent centralized RL approaches focus on learning from online data, ignoring the per-sample-cost of interactions within real-world transportation systems. To address these limitations, we propose to formalize the control of AMoD systems through the lens of offline reinforcement learning and learn effective control strategies using solely offline data, which is readily available to current mobility operators. We further investigate design decisions and provide empirical evidence based on data fromreal-worldmobility systems showing how offline learning allows to recover AMoD control policies that (i) exhibit performance on par with online methods, (ii) allowfor sample-efficient online fine-tuning and (iii) eliminate the need for complex simulation environments. Crucially, this paper demonstrates that offlineRLis a promising paradigm for the application of RL-based solutions within economically-critical systems, such as mobility systems.

关键词： Reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

Tuning the feedback controller gains is a simple way to improve autonomous driving performance 14

Tuning the feedback controller gains is a simple way to impr...

引用

14th International conference on control (control)

作者： Liang, Wenyu Baldivieso, Pablo R. Drummond, Ross Shin, Donghwan Univ Sheffield Dept Automat Control & Syst Engn Sheffield England Univ Sheffield Dept Comp Sci Sheffield England

ISBN: (纸本)9798350374278;9798350374261

Typical autonomous driving systems are a combination of machine learning algorithms (often involving neural networks) and classical feedback controllers. Whilst significant progress has been made in recent years on the neural network side of these systems, only limited progress has been made on the feedback controller side. Often, the feedback control gains are simply passed from paper to paper with little re-tuning taking place, even though the changes to the neural networks can alter the vehicle's closed loop dynamics. The aim of this paper is to highlight the limitations of this approach;it is shown that retuning the feedback controller can be a simple way to improve autonomous driving performance. To demonstrate this, the PID gains of the longitudinal controller in the TCP autonomous vehicle algorithm are tuned. This causes the driving score in CARLA to increase from 73.21 to 77.38, with the results averaged over 16 driving scenarios. Moreover, it was observed that the performance benefits were most apparent during challenging driving scenarios, such as during rain or night time, as the tuned controller led to a more assertive driving style. These results demonstrate the value of developing both the neural network and feedback control policies of autonomous driving systems simultaneously, as this can be a simple and methodical way to improve autonomous driving system performance and robustness.

关键词： Autonomous driving systems feedback control neural networks

来源：评论

学校读者我要写书评

暂无评论

Co-Design of a Wireless networked control System for Reliability and Resource-Efficiency 19

Co-Design of a Wireless Networked Control System for Reliabi...

引用

19th Baltic Electronics conference

作者： Ashraf, Kanwal Le Moullec, Yannick Pardy, Tamas Rang, Toomas Tallinn Univ Technol Thomas Johann Seebeck Dept Elect Tallinn Estonia

ISBN: (纸本)9798350365122;9798350365115

We present a co-design method for wireless networked control systems (WNCS) that optimizes both the network and control parameters for optimal performance. The objective is to make the system resource-efficient by introducing the possibility of switching the control policy from reliable control to energy-efficient control, or by optimizing the power consumption using variable inter-packet gap (IPG) in the wireless communication layer. An RL-based approach is used to ensure efficient resource allocation and create effective control performance under moderate to high packet loss. This is achieved by formulating a multi-objective problem that considers both resource efficiency and reliability. Subsequently, the proposed algorithm is used to solve this multi-objective problem. Simulation results show that despite situations where the network experiences packet losses, the proposed co-design reinforcement learning (RL)-based control technique effectively maintains, or in some cases even improves control performance. In contrast to conventional control, where a 15% packet loss in the network causes control performance to completely fail (unstable behavior), the proposed RL-based approach is immune to network packet loss (acceptable performance, i.e., control stability until 30% of packet loss) and also demonstrates the capability to achieve an average reduction in transmission power of up-to 10%.

关键词： Reinforcement Learning (RL) Wireless networked control systems co-design latency inter-packet gap packet loss optimal 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：