Nothing Special   »   [go: up one dir, main page]
You seem to have javascript disabled. Please note that many of the page functionalities won't work as expected without javascript enabled.
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
5.3
2.6
Journals
Electronics
Special Issues
Empowering IoT with AI: AIoT for Smart and Autonomous Systems
share announcement

Empowering IoT with AI: AIoT for Smart and Autonomous Systems

A special issue of Electronics (ISSN 2079-9292). This special issue belongs to the section "Computer Science & Engineering".

Deadline for manuscript submissions: 15 March 2025 | Viewed by 3458

Special Issue Editors

Dr. Zijian Li

E-Mail Website
Guest Editor
College of Artificial Intelligence, Dalian Maritime University, Dalian 116026, China
Interests: computer networking; edge intelligence; federated learning
Dr. Zihan Chen

E-Mail Website
Guest Editor
Information Systems Technology and Design, Singapore University of Technology and Design (SUTD), 8 Somapah Road, Singapore 487372, Singapore
Interests: network intelligence; federated learning; machine learning
Dr. Jiahui Li

E-Mail Website
Guest Editor
College of Computer Science and Technology, Jilin University, Changchun 130012, China
Interests: integrated air–ground networks, UAV networks, wireless energy transfer, and optimization
Dr. Longyu Zhou

E-Mail Website
Guest Editor
Information Systems Technology and Design, Singapore University of Technology and Design (SUTD), 8 Somapah Road, Singapore 487372, Singapore
Interests: digital twin; UAV communication and networking; open-RAN; deep reinforcement learning

Special Issue Information

Dear Colleagues,

The Internet of Things (IoT) is revolutionizing daily life, but artificial intelligence (AI) truly unlocks its full potential. Three pivotal emerging technologies—AI, 5G networks, and big data—are enhancing IoT, culminating in the Artificial Intelligence of Things (AIoT). The primary objective of AIoT is to bring intelligence to the edge, enabling devices to recognize data, evaluate their environments, and determine optimal actions. Through AI, IoT devices have evolved into sophisticated machines capable of autonomous analytics and independent decision making, far surpassing their initial roles as mere data transmitters.

Nonetheless, the integration of IoT and AI in the context of smart and autonomous systems introduces critical challenges for the research community. A crucial step in realizing the full potential of AIoT is fostering collaboration among massive distributed devices. Without such collaboration, AIoT systems may encounter issues such as inefficient energy use, security vulnerabilities, and inconsistent performance. Another essential aspect is the seamless integration of AIoT with other cutting-edge technologies. Both academia and industry must prioritize bridging the gap between AIoT with 5G networks, edge computing, blockchain, digital twin, and other emerging technologies to achieve comprehensive convergence.

This Special Issue seeks high-quality contributions from experts in academia and industry in the fields of AIoT, machine learning, 5G networks, and big data. We solicit high-quality original papers on the development of AI models for IoT systems and the presentation of pioneering approaches and applications.

The topics of interest include, but are not limited to, the following:

  • Novel AIoT architectures and frameworks;
  • Integration and application of 5G networks with AIoT;
  • Security and privacy solutions for AIoT against malicious attacks;
  • Intelligent edge/fog/cloud computing;
  • Integration of large language model (LLM)/large vision model (LVM) and AIoT;
  • AIoT-Enabled Semantic Communication;
  • Distributed Collaborative Learning in AioT;
  • Robust distributed AIoT design in smart and autonomous system;
  • Interplay between digital twin and AIoT over networks;
  • Novel methods for AIoT solutions with limited communication/computation resources;
  • AI-driven applications in smart homes, smart cities, and smart industries.

We look forward to receiving your contributions.

Dr. Zijian Li
Dr. Zihan Chen
Dr. Jiahui Li
Dr. Longyu Zhou
Guest Editors

Manuscript Submission Information

Manuscripts should be submitted online at www.mdpi.com by registering and logging in to this website. Once you are registered, click here to go to the submission form. Manuscripts can be submitted until the deadline. All submissions that pass pre-check are peer-reviewed. Accepted papers will be published continuously in the journal (as soon as accepted) and will be listed together on the special issue website. Research articles, review articles as well as short communications are invited. For planned papers, a title and short abstract (about 100 words) can be sent to the Editorial Office for announcement on this website.

Submitted manuscripts should not have been published previously, nor be under consideration for publication elsewhere (except conference proceedings papers). All manuscripts are thoroughly refereed through a single-blind peer-review process. A guide for authors and other relevant information for submission of manuscripts is available on the Instructions for Authors page. Electronics is an international peer-reviewed open access semimonthly journal published by MDPI.

Please visit the Instructions for Authors page before submitting a manuscript. The Article Processing Charge (APC) for publication in this open access journal is 2400 CHF (Swiss Francs). Submitted papers should be well formatted and use good English. Authors may use MDPI's English editing service prior to publication or during author revisions.

Keywords

  • artificial intelligence of things
  • edge computing
  • 5G networks
  • security and privacy
  • AI-driven applications

Benefits of Publishing in a Special Issue

  • Ease of navigation: Grouping papers by topic helps scholars navigate broad scope journals more efficiently.
  • Greater discoverability: Special Issues support the reach and impact of scientific research. Articles in Special Issues are more discoverable and cited more frequently.
  • Expansion of research network: Special Issues facilitate connections among authors, fostering scientific collaborations.
  • External promotion: Articles in Special Issues are often promoted through the journal's social media, increasing their visibility.
  • e-Book format: Special Issues with more than 10 articles can be published as dedicated e-books, ensuring wide and rapid dissemination.

Further information on MDPI's Special Issue policies can be found here.

Published Papers (4 papers)

Download All Papers
Order results
Result details
Show export options expand_more Show export options expand_less
Select all
Export citation of selected articles as:

Research

30 pages, 2514 KiB  
Article
FedCon: Scalable and Efficient Federated Learning via Contribution-Based Aggregation
by Wenyu Gao, Gaochao Xu and Xianqiu Meng
Electronics 2025, 14(5), 1024; https://doi.org/10.3390/electronics14051024 - 4 Mar 2025
Viewed by 226
Abstract
With the increasing application of federated learning to medical and image data, the challenges of class distribution imbalances and Non-IID heterogeneity across clients have become critical factors affecting the generalization ability of global models. In the medical domain, the phenomenon of data silos [...] Read more.
With the increasing application of federated learning to medical and image data, the challenges of class distribution imbalances and Non-IID heterogeneity across clients have become critical factors affecting the generalization ability of global models. In the medical domain, the phenomenon of data silos is particularly pronounced, leading to significant differences in data distributions across hospitals, which in turn hinder the performance of global model training. To address these challenges, this paper proposes FedCon, a federated learning method capable of dynamically adjusting aggregation weights, while accurately evaluating client contributions. Specifically, FedCon initializes aggregation weights based on client data volume and class distribution and employs Monte Carlo sampling to effectively simplify the computation of Shapley values. Subsequently, it further optimizes the aggregation weights by comprehensively considering the historical contributions of clients and the similarity between clients and the global model. This approach significantly enhances the ability to generalize and update the stability of the global model. Experimental results demonstrate that, compared to existing methods, FedCon achieved a superior generalization performance on public datasets and significantly accelerated the convergence of the global model. Full article
(This article belongs to the Special Issue Empowering IoT with AI: AIoT for Smart and Autonomous Systems)
►▼ Show Figures

Figure 1

Figure 1
<p>The FedCon framework: (<b>A</b>) the calculation of client data quality in the first round using discrepancies between global and local data distributions to determine initialization weights; (<b>B</b>) the dynamic adjustment of aggregation weights based on precise client contribution computations (using Shapley values and similarity metrics) in each round, improving the model convergence stability.</p> Full article ">Figure 2
<p>The above chart is the heatmap of data partitioning for CIFAR10-NIID-1.</p> Full article ">Figure 3
<p>The above chart is the heatmap of data partitioning for CIFAR10-NIID-2.</p> Full article ">Figure 4
<p>Convergence analysis of different methods on CIFAR10 and CIFAR100 datasets under independent and identically distributed (Homo) settings. (<b>a</b>) shows the convergence analysis of CIFAR10 under Homo settings, and (<b>b</b>) shows the convergence analysis of CIFAR100 under Homo settings.</p> Full article ">Figure 5
<p>Convergence analysis of different methods on the CIFAR10 dataset under two Non-IID data partitioning strategies, NIID-1 and NIID-2. The figure above demonstrates that the FedCon method achieved a faster convergence and outperformed the other methods in terms of performance. (<b>a</b>) shows the convergence analysis of CIFAR10 under NIID-1 settings, and (<b>b</b>) shows the convergence analysis of CIFAR10 under NIID-2 settings.</p> Full article ">Figure 6
<p>Convergence analysis of various methods on the CIFAR100 dataset under two Non-IID data partitioning strategies, NIID-1 and NIID-2. (<b>a</b>) shows the convergence analysis of CIFAR100 under NIID-1 settings, and (<b>b</b>) shows the convergence analysis of CIFAR100 under NIID-2 settings.</p> Full article ">Figure 7
<p>Convergence analysis of different methods on the HAR and HAM10000 datasets under the NIID-1 partitioning strategy (since these two datasets have different numbers of classes, only the NIID-1 partitioning strategy could be applied). (<b>a</b>) shows the convergence analysis of HAR under NIID-1 settings, and (<b>b</b>) shows the convergence analysis of HAM10000 under NIID-1 settings.</p> Full article ">Figure 8
<p>Convergence analysis of different methods on the OrganAMNIST dataset under the NIID-1 and NIID-2 Non-IID data partitioning strategies. (<b>a</b>) shows the convergence analysis of OrganAMNIST under NIID-1 settings, and (<b>b</b>) shows the convergence analysis of OrganAMNIST under NIID-2 settings.</p> Full article ">Figure 9
<p>Convergence analysis of different methods on the OrganCMNIST dataset under the NIID-1 and NIID-2 Non-IID data partitioning strategies. (<b>a</b>) shows the convergence analysis of OrganCMNIST under NIID-1 settings, and (<b>b</b>) shows the convergence analysis of OrganCMNIST under NIID-2 settings.</p> Full article ">Figure 10
<p>Convergence analysis of different methods on the OrganSMNIST dataset under the NIID-1 and NIID-2 Non-IID data partitioning strategies. (<b>a</b>) shows the convergence analysis of OrganSMNIST under NIID-1 settings, and (<b>b</b>) shows the convergence analysis of OrganSMNIST under NIID-2 settings.</p> Full article ">Figure 11
<p>The figure above presents the RMSE of the eight baseline methods and the FedCon method across the different datasets and partitioning strategies. Notably, the performance of FedCon was particularly remarkable on the HAM10000 dataset. On other datasets, FedCon also demonstrated superior performance, highlighting its advanced capabilities.</p> Full article ">Figure 12
<p>Comparison of communication time, round time, and accuracy for NIID-1 setting. (<b>a</b>) shows communication time and accuracy on NIID-1 setting. (<b>b</b>) shows round time and accuracy on NIID-1 setting.</p> Full article ">Figure 13
<p>Comparison of communication time, round time, and accuracy for NIID-2 setting. (<b>a</b>) shows communication time and accuracy on NIID-2 setting. (<b>b</b>) shows round time and accuracy on NIID-2 setting.</p> Full article ">Figure 14
<p>The above figure shows the hyperparameter analysis under CIFAR10 with the NIID-1 data partitioning.</p> Full article ">Figure 15
<p>The above figure shows the hyperparameter analysis under CIFAR10 with the NIID-2 data partitioning.</p> Full article ">Figure 16
<p>Violin plot comparing FedCon without A, FedCon without B, FedCon, and other baseline methods.</p> Full article ">
18 pages, 1071 KiB  
Article
Optimal Reconfigurable Intelligent Surface Deployment for Secure Communication in Cell-Free Massive Multiple-Input Multiple-Output Systems with Coverage Area
by Jie Zhao, Qi Zhang, Tianyu Ai, Xianhu Wei and Fengqiang Peng
Electronics 2025, 14(2), 241; https://doi.org/10.3390/electronics14020241 - 8 Jan 2025
Viewed by 522
Abstract
This paper investigates the secure communication in the reconfigurable intelligent surface (RIS)-aided cell-free massive multiple-input multiple-output (CF-mMIMO) system in the presence of an eavesdropper (Eve). Since the RIS can only reflect the incident signal from its front, we define the RIS coverage and [...] Read more.
This paper investigates the secure communication in the reconfigurable intelligent surface (RIS)-aided cell-free massive multiple-input multiple-output (CF-mMIMO) system in the presence of an eavesdropper (Eve). Since the RIS can only reflect the incident signal from its front, we define the RIS coverage and non-coverage area based on whether the incident signals can be reflected. The RIS coverage area is affected by the deployment position and rotation angle, and thus, we take both of these two factors into account and a closed-form approximation for the ergodic secrecy rate of the legitimate user is derived. Based on it, the optimal RIS deployment position and phase shift are obtained through an alternating iteration method, and the optimal RIS angle is achieved through an exhaustive enumeration of angles with a certain interval. Simulations confirm that our optimal RIS deployment can achieve a superior secrecy rate. We find that to guarantee the best secrecy rate, the RIS should be placed near the target user, and its rotation angle should be adjusted to make as many access points (APs) as possible within the RIS coverage area. Full article
(This article belongs to the Special Issue Empowering IoT with AI: AIoT for Smart and Autonomous Systems)
►▼ Show Figures

Figure 1

Figure 1
<p>The reconfigurable intelligent surface (RIS)-aided cell-free massive multiple-input multiple-output (CF-mMIMO) system with an eavesdropper (Eve).</p> Full article ">Figure 2
<p>Secrecy rate vs. access point(AP) number, where <math display="inline"><semantics> <mrow> <mi>L</mi> <mo>=</mo> <mn>64</mn> </mrow> </semantics></math> and <math display="inline"><semantics> <mrow> <msub> <mi>p</mi> <mi>d</mi> </msub> <mo>=</mo> <mn>30</mn> </mrow> </semantics></math> dBm.</p> Full article ">Figure 3
<p>Secrecy rate vs. AP transmit power, where <math display="inline"><semantics> <mrow> <mi>M</mi> <mo>=</mo> <mn>30</mn> </mrow> </semantics></math>, <math display="inline"><semantics> <mrow> <mi>K</mi> <mo>=</mo> <mn>20</mn> </mrow> </semantics></math>, and <math display="inline"><semantics> <mrow> <mi>L</mi> <mo>=</mo> <mn>64</mn> </mrow> </semantics></math>.</p> Full article ">Figure 4
<p>Secrecy rate vs. AP number, where <math display="inline"><semantics> <mrow> <mi>L</mi> <mo>=</mo> <mn>64</mn> </mrow> </semantics></math> and <math display="inline"><semantics> <mrow> <msub> <mi>p</mi> <mi>d</mi> </msub> <mo>=</mo> <mn>30</mn> </mrow> </semantics></math> dBm.</p> Full article ">Figure 5
<p>Optimal RIS positioning for different target user and Eve—Scenario 1.</p> Full article ">Figure 6
<p>Optimal RIS positioning for different target users and Eve—Scenario 2.</p> Full article ">Figure 7
<p>Secrecy rate vs. RIS component number, where <math display="inline"><semantics> <mrow> <mi>M</mi> <mo>=</mo> <mn>30</mn> </mrow> </semantics></math>, <math display="inline"><semantics> <mrow> <mi>K</mi> <mo>=</mo> <mn>20</mn> </mrow> </semantics></math>, and <math display="inline"><semantics> <mrow> <msub> <mi>p</mi> <mi>d</mi> </msub> <mo>=</mo> <mn>30</mn> </mrow> </semantics></math> dBm.</p> Full article ">Figure 8
<p>Secrecy rate vs. iteration number.</p> Full article ">
18 pages, 406 KiB  
Article
A Blockchain Multi-Chain Federated Learning Framework for Enhancing Security and Efficiency in Intelligent Unmanned Ports
by Zeqiang Xie and Zijian Li
Electronics 2024, 13(24), 4926; https://doi.org/10.3390/electronics13244926 - 13 Dec 2024
Viewed by 690
Abstract
The integration of blockchain and federated learning (FL) has emerged as a promising solution to address data privacy and security challenges in Intelligent Unmanned Ports (IUPs). However, existing blockchain federated learning (BFL) frameworks encounter significant limitations, including high latency, inefficient data processing, and [...] Read more.
The integration of blockchain and federated learning (FL) has emerged as a promising solution to address data privacy and security challenges in Intelligent Unmanned Ports (IUPs). However, existing blockchain federated learning (BFL) frameworks encounter significant limitations, including high latency, inefficient data processing, and limited scalability, particularly in scenarios with sparse and distributed data. This paper introduces a novel multi-chain federated learning (MFL) framework to overcome these challenges. The proposed MFL architecture interconnects multiple BFL chains to facilitate the secure and efficient aggregation of data across distributed devices. The framework enhances privacy and efficiency by transmitting aggregated model updates rather than raw data. A low-frequency consensus mechanism is employed to improve performance, leveraging game theory for representative selection to optimize model aggregation while reducing inter-chain communication overhead. The experimental results demonstrate that the MFL framework significantly outperforms traditional BFL in terms of accuracy, latency, and system efficiency, particularly under the conditions of high data sparsity and network latency. These findings highlight the potential of MFL to provide a scalable and secure solution for decentralized learning in IUP environments, with broader applicability to other distributed systems such as the Industrial Internet of Things (IIoT). Full article
(This article belongs to the Special Issue Empowering IoT with AI: AIoT for Smart and Autonomous Systems)
►▼ Show Figures

Figure 1

Figure 1
<p>The figure illustrates the workflow of the proposed MFL framework. Data are collected locally from Port Equipment and Cargo Ship Equipment, where self-training is performed to generate gradients. These gradients are securely stored and processed in the Blockchain Storage Units, supported by Blockchain Computing Nodes. The Cross-Chain Communication Nodes enable efficient gradient exchange between blockchains, facilitating global model updates while ensuring data privacy and system scalability.</p> Full article ">Figure 2
<p>Workflow of the low-frequency consensus.</p> Full article ">Figure 3
<p>Impact of data sparsity on model performance.</p> Full article ">Figure 4
<p>Comparison of model performance–performance by the times.</p> Full article ">Figure 5
<p>Comparison of model performance–loss by the times.</p> Full article ">Figure 6
<p>Comparison of convergence, performance by the times, and loss by the times.</p> Full article ">Figure 7
<p>Comparisons of the system delay between different nodes of BFL and MFL.</p> Full article ">
15 pages, 1402 KiB  
Article
Enhancing Anomaly Detection in Maritime Operational IoT Time Series Data with Synthetic Outliers
by Hyunjoo Kim and Inwhee Joe
Electronics 2024, 13(19), 3912; https://doi.org/10.3390/electronics13193912 - 3 Oct 2024
Viewed by 1213
Abstract
Detecting anomalies in engine and machinery data during ship operations is crucial for maintaining the safety and efficiency of the vessel. We conducted experiments using device data from the maritime industry, consisting of time series records from IoT (Internet of Things) datasets such [...] Read more.
Detecting anomalies in engine and machinery data during ship operations is crucial for maintaining the safety and efficiency of the vessel. We conducted experiments using device data from the maritime industry, consisting of time series records from IoT (Internet of Things) datasets such as cylinder and exhaust gas temperatures, coolant temperatures, and cylinder pressures collected from various sensors on the ship’s equipment. We propose data enrichment and validation techniques by generating synthetic outliers through data degradation and data augmentation with a Transformer backbone, utilizing the maritime operational data. We extract a portion of the input data and replace it with synthetic outliers. The created anomaly data are then used to train the model via a self-supervised learning approach. Synthetic outliers are generated using methods such as the arithmetic mean, geometric mean, median, local scale, global scale, and magnitude warping. With our methodology, we achieved a 17.23% improvement in F1 performance compared to existing state-of-the-art methods across five publicly available datasets and actual maritime operational data collected from the industry. Full article
(This article belongs to the Special Issue Empowering IoT with AI: AIoT for Smart and Autonomous Systems)
►▼ Show Figures

Figure 1

Figure 1
<p>Various types of anomalies in time series data: global anomaly (red), contextual anomaly (green), seasonal anomaly (blue), trend anomaly (purple), shapelet anomaly (orange), and original data (black).</p> Full article ">Figure 2
<p>This diagram depicts an anomaly detection framework for maritime operational data. Outliers are generated through the synthetic outlier generation module and used to train the Transformer backbone model, which produces an F1 score. The model with the best score is determined through voting to identify the best model, which is then used for anomaly detection.</p> Full article ">Figure 3
<p>(<b>a</b>) Describes important data from ship equipment related to the engine; (<b>b</b>) details the model selection method used to choose the best model for anomaly detection.</p> Full article ">Figure 4
<p>Synthetic outlier type for anomaly detection using arithmetic mean and geometric mean.</p> Full article ">Figure 5
<p>Synthetic outlier type for anomaly detection using median and local scaling.</p> Full article ">Figure 6
<p>Synthetic outlier type for anomaly detection using global scaling and magnitude warping.</p> Full article ">Figure 7
<p>F1 score compared with SOTA. The blue bar represents AnomalyBERT, the orange bar represents CARLA, and the green bar represents the F1 score.</p> Full article ">Figure 8
<p>F1 score for external interval replacement. The blue bar represents the Flip method, the orange bar denotes the arithmetic mean method, the green bar represents the geometric mean method, and the red bar represents the median method. The purple bar denotes the global scale method. The brown bar represents the local scale method. The Pink bar represents the magnitude warping method.</p> Full article ">Figure 9
<p>Performances of different global scales for different datasets. WADI (blue), SWaT (orange), MSL (green), SMAP (red), SMD (purple), and IMOD (brown).</p> Full article ">Figure 10
<p>Loss with global scales for the IMOD dataset.</p> Full article ">Figure 11
<p>Anomaly detection results for IMOD. The red point denotes the test dataset. The top part shows the original data for Cylinder7 Pmax, while the bottom part displays the anomaly score results for the corresponding indices.</p> Full article ">Figure 12
<p>Anomaly detection results for IMOD. The red point is the label for the test dataset. The top part shows the original data for the bearing temperature, cylinder exhaust gas outlet temperature, cylinder Pmax, engine load, and power, while the bottom part displays the anomaly score results for the corresponding indices.</p> Full article ">
Show export options expand_more Show export options expand_less
Select all
Export citation of selected articles as:
 
Displaying articles 1-4
Back to TopTop