Energy-Sustainable IoT Connectivity: Vision, Technological Enablers, Challenges, and Future Directions

IEEE Open Journal of the Communications Society, Год журнала: 2023, Номер 4, С. 2609 - 2666

Опубликована: Янв. 1, 2023

Technology solutions must effectively balance economic growth, social equity, and environmental integrity to achieve a sustainable society. Notably, although the Internet of Things (IoT) paradigm constitutes key sustainability enabler, critical issues such as increasing maintenance operations, energy consumption, manufacturing/disposal IoT devices have long-term negative economic, societal, impacts be efficiently addressed. This calls for self-sustainable ecosystems requiring minimal external resources intervention, utilizing renewable sources, recycling materials whenever possible, thus encompassing sustainability. In this work, we focus on energy-sustainable during operation phase, our discussions sometimes extend other aspects lifecycle phases. Specifically, provide fresh look at identify provision, transfer, efficiency three main energy-related processes whose harmonious coexistence pushes toward realizing systems. Their related technologies, recent advances, challenges, research directions are also discussed. Moreover, overview relevant performance metrics assess energy-sustainability potential certain technique, technology, device, or network, together with target values next generation wireless systems, discuss protocol, integration, implementation issues. Overall, paper offers insights that valuable advancing goals present future generations.

Язык: Английский

---

Large Language Models Empowered Autonomous Edge AI for Connected Intelligence

IEEE Communications Magazine, Год журнала: 2024, Номер 62(10), С. 140 - 146

Опубликована: Янв. 8, 2024

The evolution of wireless networks gravitates towards connected intelligence, a concept that envisions seamless interconnectivity among humans, objects, and intelligence in hyper-connected cyber-physical world. Edge artificial (Edge AI) is promising solution to achieve by delivering high-quality, low-latency, privacy-preserving AI services at the network edge. This article presents vision autonomous edge systems automatically organize, adapt, optimize themselves meet users' diverse requirements, leveraging power large language models (LLMs), i.e., Generative Pretrained Transformer (GPT). By exploiting powerful abilities GPT understanding, planning, code generation, as well incorporating classic wisdom such task-oriented communication federated learning, we present versatile framework efficiently coordinates cater personal demands while generating train new manner. Experimental results demonstrate system's remarkable ability accurately comprehend user demands, execute with minimal cost, effectively create highperformance servers.

Язык: Английский

---

Noninvasive Diabetes Detection through Human Breath Using TinyML-Powered E-Nose

Sensors, Год журнала: 2024, Номер 24(4), С. 1294 - 1294

Опубликована: Фев. 17, 2024

Volatile organic compounds (VOCs) in exhaled human breath serve as pivotal biomarkers for disease identification and medical diagnostics. In the context of diabetes mellitus, noninvasive detection acetone, a primary biomarker using electronic noses (e-noses), has gained significant attention. However, employing e-noses requires pre-trained algorithms precise detection, often requiring computer with programming environment to classify newly acquired data. This study focuses on development an embedded system integrating Tiny Machine Learning (TinyML) e-nose equipped Metal Oxide Semiconductor (MOS) sensors real-time detection. The encompassed 44 individuals, comprising 22 healthy individuals diagnosed various types mellitus. Test results highlight XGBoost algorithm’s achievement 95% accuracy. Additionally, integration deep learning algorithms, particularly neural networks (DNNs) one-dimensional convolutional network (1D-CNN), yielded efficacy 94.44%. These outcomes underscore potency combining TinyML systems, offering approach mellitus

Язык: Английский

---

Fast Resource Estimation of FPGA-Based MLP Accelerators for TinyML Applications

Electronics, Год журнала: 2025, Номер 14(2), С. 247 - 247

Опубликована: Янв. 9, 2025

Tiny machine learning (TinyML) demands the development of edge solutions that are both low-latency and power-efficient. To achieve these on System-on-Chip (SoC) FPGAs, co-design methodologies, such as hls4ml, have emerged aiming to speed up design process. In this context, fast estimation FPGA’s utilized resources is needed rapidly assess feasibility a design. paper, we propose resource estimator for fully customized (bespoke) multilayer perceptrons (MLPs) designed through hls4ml workflow. Through analysis bespoke MLPs synthesized using Xilinx High-Level Synthesis (HLS) tools, developed models dense layers’ arithmetic modules registers. These consider unique characteristics inherent nature MLPs. Our was evaluated six different architectures synthetic real benchmarks, which were Vitis HLS 2022.1 targeting ZYNQ-7000 FPGAs. experimental demonstrates our can accurately predict required in terms Look-Up Tables (LUTs), Flip-Flops (FFs), Digital Signal Processing (DSP) units less than 147 ms single-threaded execution.

Язык: Английский

---

Energy-aware tinyML model selection on zero energy devices

Internet of Things, Год журнала: 2025, Номер 30, С. 101488 - 101488

Опубликована: Янв. 31, 2025

Язык: Английский

---

Early Termination for Hyperdimensional Computing Using Inferential Statistics

Опубликована: Фев. 3, 2025

Язык: Английский

---

TinyML-Based Approach for Dynamic Transmission Power in LoRaWAN Network

Communications in computer and information science, Год журнала: 2025, Номер unknown, С. 217 - 227

Опубликована: Янв. 1, 2025

Язык: Английский

---

Design and Implementation of ESP32-Based Edge Computing for Object Detection

Sensors, Год журнала: 2025, Номер 25(6), С. 1656 - 1656

Опубликована: Март 7, 2025

This paper explores the application of ESP32 microcontroller in edge computing, focusing on design and implementation an server system to evaluate performance improvements achieved by integrating cloud computing. Responding growing need reduce burdens latency, this research develops server, detailing hardware architecture, software environment, communication protocols, framework. A complementary framework is also designed support processing. deep learning model for object recognition selected, trained, deployed server. Performance evaluation metrics, classification time, MQTT (Message Queuing Telemetry Transport) transmission data from various brokers are used assess performance, with particular attention impact image size adjustments. Experimental results demonstrate that significantly reduces bandwidth usage effectively alleviating load study discusses system’s strengths limitations, interprets experimental findings, suggests potential future applications. By AI IoT, demonstrates benefits localized processing enhancing efficiency reducing dependency.

Язык: Английский

---

High Accuracy of Epileptic Seizure Detection Using Tiny Machine Learning Technology for Implantable Closed-Loop Neurostimulation Systems

BioMedInformatics, Год журнала: 2025, Номер 5(1), С. 14 - 14

Опубликована: Март 10, 2025

Background: Epilepsy is one of the most common and devastating neurological disorders, manifesting with seizures affecting approximately 1–2% world’s population. The criticality seizure occurrence associated risks, combined overwhelming need for more precise innovative treatment methods, has led to development invasive neurostimulation devices programmed detect apply electrical stimulation therapy suppress reduce burden. Tiny Machine Learning (TinyML) a rapidly growing branch machine learning. One its key characteristics ability run learning algorithms without high computational complexity powerful hardware resources. featured work utilizes TinyML technology implement an algorithm that can be integrated into microprocessor implantable closed-loop brain system accurately in real-time by analyzing intracranial EEG (iEEG) signals. Methods: A dataset containing iEEG signal values from both non-epileptic epileptic individuals was utilized implementation proposed algorithm. Appropriate data preprocessing performed, two training datasets 1000 records signals were created. test independent 500 also web-based platform Edge Impulse used model generation visualization, different architectures explored tested. Finally, metrics accuracy, confusion matrices, ROC curves evaluate performance model. Results: Our demonstrated performance, achieving 98% 99% accuracy on validation datasets, respectively. results support use epilepsy, as it contributes significantly speed detection. Conclusions: reliable detection distinguishing activity normal activity. These findings highlight potential systems enhancing

Язык: Английский

---

Deep Learning for Non-Invasive Blood Pressure Monitoring: Model Performance and Quantization Trade-Offs

Electronics, Год журнала: 2025, Номер 14(7), С. 1300 - 1300

Опубликована: Март 26, 2025

The development of non-invasive blood pressure monitoring systems remains a critical challenge, particularly in resource-constrained settings. This study proposes an efficient deep learning framework integrating Edge Artificial Intelligence for continuous estimation using photoplethysmography (PPG) signals. We evaluate three architectures: residual-enhanced convolutional neural network, transformer-based model, and attentive BPNet. Using the MIMIC-IV waveform database, we implement signal processing pipeline with adaptive filtering, statistical normalization, peak-to-peak alignment. Experiments assess varying temporal windows (10 s, 20 30 s) to optimize predictive accuracy computational efficiency. Attentive BPNet achieves best performance, systolic (SBP) yielding mean absolute error (MAE) 6.36 mmHg, diastolic (DBP) MAE 4.09 arterial (MBP) 4.56 mmHg. Post-training quantization reduces model size by 90.71% (to 0.13 MB), enabling deployment on devices. These findings demonstrate feasibility deploying learning-based edge proposed provides scalable computationally solution, offering real-time, accessible that could enhance hypertension management healthcare resource utilization.

Язык: Английский

---