Search

Nano-networks are envisioned to allow several nanoscale devices to transmit and receive information. One form of such networks is electromagnetic nano-networks working within the THz band. However, high overall path loss and molecular noise experienced in the THz band, as well as limited energy storage capabilities, restrict the communication range of nano-nodes and impact network efficiency. Therefore, optimizing the nano-network resources is necessary. In this paper, we present an optimization framework employing mixed-integer linear programming (MILP) to determine the most energy-efficient routing, bandwidth, and sub-band allocation for each nano-node in an electromagnetic nano-network operating within the THz band. Our model was tested for two different scenarios related to the priority of energy saving. We also compare our proposed optimal bandwidth, routing, and sub-band allocation against less complex designs where sub-bands with fixed bandwidth are employed in nano-nodes. Furthermore, we investigate the impact of nano-node’s processing and sensing units on the overall network energy consumption and the associated optimal bandwidth allocation and routing strategy. Given the considered parameters and the model’s assumptions, the results show that using the optimal multi-hops paths with higher bandwidth allocation for the considered sub-bands can be more energy efficient than sending the traffic using a single hop and lower bandwidths, especially when the transmission power dominates in the nano-network. On the other hand, when the processing and sensing unit’s energy consumption is dominant, then single hop schemes with lower bandwidth allocation result in the minimum network energy consumption. Finally, we discuss the limitations of the proposed energy-efficient strategies and point toward possible future research directions to which the model can be adapted.

Tiny machine learning (TinyML) is a promising approach to enable intelligent applications relying on Human Activity Recognition (HAR) on resource-limited and low-power Internet of Things (IoT) edge devices. However, designing efficient TinyML models for these devices remains challenging due to computational resource constraints and the need for customisation to unique use cases. To address this, we propose a novel approach that utilises transfer learning (TL) techniques on edge micro-controller units (MCUs) to accelerate TinyML development. Our strategy involves pre-training generalised ML models on large-scale and varied datasets and fine-tuning them on-device for specific applications using TL. We demonstrate the effectiveness of our approach for HAR by experimenting with a convolutional neural network, long short-term memory TL (CNN-LSTM-TL) model architecture and visualisation techniques like t-distributed stochastic neighbour embedding (t-SNE) for dimensionality reduction. To further validate our model’s proficiency and adaptability, we conducted extensive testing using two distinct datasets: MotionSense and UCI. This dual-dataset approach allowed us to assess the robustness of our model across different data domains, showcasing its versatility and effectiveness in various HAR scenarios. Our results show significant model accuracy and reduced training time while maintaining high inference rates and low MCU memory footprint. We also provide insights into best practices for implementing TL on edge MCUs and evaluate classification performance metrics such as accuracy, precision, recall, F1 score, and categorical cross-entropy loss. Our work lays a solid foundation for faster and more efficient TinyML deployments through TL framework across different application domains and types of edge IoT devices.

The rapid growth of Internet of Things (IoT) devices has generated vast amounts of data, leading to the emergence of federated learning as a novel distributed machine learning paradigm. Federated learning enables model training at the edge, leveraging the processing capacity of edge devices while preserving privacy and mitigating data transfer bottlenecks. However, the conventional centralized federated learning architecture suffers from a single point of failure and susceptibility to malicious attacks. In this study, we delve into an alternative approach called decentralized federated learning (DFL) conducted over a wireless mesh network as the communication backbone. We perform a comprehensive network performance analysis using stochastic geometry theory and physical interference models, offering fresh insights into the convergence analysis of DFL. Additionally, we conduct system simulations to assess the proposed decentralized architecture under various network parameters and different aggregator methods such as FedAvg, Krum and Median methods. Our model is trained on the widely recognized EMNIST dataset for benchmarking handwritten digit classification. To minimize the model’s size at the edge and reduce communication overhead, we employ a cutting-edge compression technique based on genetic algorithms. Our simulation results reveal that the compressed decentralized architecture achieves performance comparable to the baseline centralized architecture and traditional DFL in terms of accuracy and average loss for our classification task. Moreover, it significantly reduces the size of shared models over the wireless channel by compressing participants’ local model sizes to nearly half of their original size compared to the baselines, effectively reducing complexity and communication overhead.

Federated Learning (FL) presents a mechanism to allow decentralized training for machine learning (ML) models inherently enabling privacy preservation. The classical FL is implemented as a client-server system, which is known as Centralised Federated Learning (CFL). There are challenges inherent in CFL since all participants need to interact with a central server resulting in a potential communication bottleneck and a single point of failure. In addition, it is difficult to have a central server in some scenarios due to the implementation cost and complexity. This study aims to use Decentralized Federated learning (DFL) without a central server through one-hop neighbours. Such collaboration depends on the dynamics of communication networks, e.g., the topology of the network, the MAC protocol, and both large-scale and small-scale fading on links. In this paper, we employ stochastic geometry to model these dynamics explicitly, allowing us to quantify the performance of the DFL. The core objective is to achieve better classification without sacrificing privacy while accommodating for networking dynamics. In this paper, we are interested in how such topologies impact the performance of ML when deployed in practice. The proposed system is trained on a well-known MINST dataset for benchmarking, which contains labelled data samples of 60K images each with a size $28\times 28$ pixels, and 1000 random samples of this MNIST dataset are assigned for each participant’ device. The participants’ devices implement a CNN model as a classifier model. To evaluate the performance of the model, a number of participants are randomly selected from the network. Due to randomness in the communication process, these participants interact with the random number of nodes in the neighbourhood to exchange model parameters which are subsequently used to update the participants’ individual models. These participants connected successfully with a varying number of neighbours to exchange parameters and update their global models. The results show that the classification prediction system was able to achieve higher than 95% accuracy using the three different model optimizers in the training settings (i.e., SGD, ADAM, and RMSprop optimizers). Consequently, the DFL over mesh networking shows more flexibility in IoT systems, which reduces the communication cost and increases the convergence speed which can outperform CFL.

Abstract Reconfigurable intelligent surfaces (RISs) are envisioned to transform the propagation space into a smart radio environment (SRE) to realize the diverse applications of sixth‐generation (6G) wireless communication. By smartly tuning the massive number of elements via controller, an RIS can passively phase‐shift the electromagnetic (EM) waves to enhance the system performance. The absence of radio‐frequency (RF) chains makes RIS an energy‐efficient and cost‐effective solution for future wireless networks. In this paper, the state‐of‐the‐art research on different aspects of RIS‐assisted communication is explored. Specifically, the fundamentals of RIS are first introduced, including the RIS's structure, operating principle, and deployment strategies. The emerging applications of RISs are then comprehensively discussed for 6G wireless networks. In addition, the crucial challenges for RIS‐assisted networks are elaborated, namely, RIS channel state information (CSI) acquisition and passive beamforming optimization. Furthermore, the recent research contributions leveraging the artificial intelligence (AI) based techniques for channel estimation, phase‐shift optimization, and resource allocation in RIS‐assisted networks are presented. Finally, to provide effective guidance for future research, important research directions for realizing RIS‐assisted network are highlighted.

Machine Learning (ML) on the edge is key to enabling a new breed of IoT and autonomous system applications. The departure from the traditional cloud-centric architecture means that new deployments can be more power-efficient, provide better privacy and reduce latency for inference. At the core of this paradigm is TinyML, a framework allowing the execution of ML models on low-power embedded devices. TinyML allows importing pre-trained ML models on the edge for providing ML-as-a-Service (MLaaS) to IoT devices. This article presents a TinyMLaaS (TMLaaS) architecture for future IoT deployments. The TMLaaS architecture inherently presents several design trade-offs in terms of energy consumption, security, privacy, and latency. We also present how TMLaaS architecture can be implemented, deployed, and maintained for large-scale IoT deployment. The feasibility of implementation for the TMLaaS architecture has been demonstrated with the help of a case study.

The goal of Internet of Medical Things (IoMT) and digital healthcare systems is to provide people with the ease of receiving quality healthcare at the comfort of their homes. Hence, the aim of IoMT is the ubiquitous deployment of home-based healthcare systems. Making such systems intelligent and efficient for timely prediction of critical diseases can save millions of lives while simultaneously reducing the burden on the traditional healthcare systems e.g., hospitals. The advancement in IoT has enabled both patients and doctors to access real time data. This advancement has reduced the cost and energy consumption of digital healthcare systems by using efficient sensors and communication technologies. This paper provides a comprehensive review of various studies conducted for the development and improvement of IoMT. It analyses different sensors used for measurement of various parameters ranging from physiological to emotional signals. It also provides a detailed investigation of different communication technologies being used, their advantages, and limitations. Moreover, digital healthcare systems are now deploying machine learning technology for the prediction of health status of patients. These techniques and algorithms are also discussed. Data security and prediction accuracy are the main concerns in the development of this area. In conclusion, this paper reviews the various digital system designs in the context of healthcare, their methodology, limitations, and the present challenges faced by the e-health sector.

The agriculture sector holds paramount importance in Pakistan due to the intrinsic agrarian nature of the economy. Pakistan has its GDP based on agriculture, however it relies on manual monitoring of crops, which is a labour intensive and ineffective method. In contrast to this, several cutting edge technology-based solutions are being employed in the developed countries to enhance the crop yield with the optimal use of resources. To this end, we have proposed an integrated approach for monitoring crop health using IoT, machine learning and drone technology. The integration of these sensing modalities generate heterogeneous data which not only varies in nature (i.e. observed parameter) but also has different temporal fidelity. The spatial resolution of these methods is also different, hence, the optimal integration of these sensing modalities and their implementation in practice are addressed in the proposed system. In our proposed solution, the IoT sensors provide the real-time status of environmental parameters impacting the crop, and the drone platform provide the multispectral data used for generating Vegetation Indices (VIs) such as Normalized Difference vegetation Index (NDVI) for analyzing the crop health. The NDVI provides information about the crop based on the chlorophyll content, which offers limited information regarding the crop health. In order to obtain a rich and detailed knowledge about crop health, the variable length time series data of IoT sensors and multispectral images were converted to a fixed-sized representation to generate crop health maps. A number of machine and deep learning algorithms were applied on the collected data wherein deep neural network with two hidden layers was found to be the most optimal model among all the selected models, providing an accuracy of (98.4%). Further, the health maps were validated through ground surveys and by agriculture experts due to the absence of reference data. The proposed research is basically an indigenous, technology based agriculture solution capable of providing important insights into the crop health by extracting complementary features from multi-modal data set, and minimizing the crop ground survey effort, particularly useful when the agriculture land is large in size.

For the first time, the design and implementation of a fully-integrated wireless information and power transfer system, operating at 24 GHz and enabling battery-less sensor nodes, is presented in this paper. The system consists of an RF power source, a receiver antenna array, a rectifier, and a battery-less sensor node which communicates via backscatter modulation at 868 MHz. The rectifier circuits use commercially available Schottky diodes to convert the RF power to DC with a measured efficiency of up to 35%, an improvement of ten percentage points compared with previously reported results. The rectifiers and the receive antenna arrays were jointly designed and optimised, thereby reducing the overall circuit size. The battery-less sensor transmitted data to a base station realised as a GNU Radio flow running on a bladeRF Software Defined Radio module. The whole system was tested in free-space in laboratory conditions and was capable of providing sufficient energy to the sensor node in order to enable operation and wireless communication at a distance of 0.15 metres.

In this paper, we establish a cryptographic primitive for wireless communications. An asymmetric physical layer encryption (PLE) scheme based on elliptic curve cryptography is proposed. Compared with the conventional symmetric PLE, asymmetric PLE avoids the need of key distribution on a private channel, and it has more tools available for processing complex-domain signals to confuse possible eavesdroppers when compared with upper-layer public key encryption. We use quantized information entropy to measure the constellation confusion degree. The numerical results show that the proposed scheme provides greater confusion to eavesdroppers and yet does not affect the bit error rate (BER) of the intended receiver (the information entropy of the constellation increases to 17.5 for 9-bit quantization length). The scheme also has low latency and complexity [O(N2.37), where N is a fixed block size], which is particularly attractive for implementation.

Security is always an important issue in wireless communications. Physical layer encryption (PLE) is an effective way to enhance wireless communication security and prevent eavesdropping. Rather than replacing cryptography at higher layers, PLE's benefit is to enable using lightweight cryptosystems or provide enhanced security at the signal level. The upper cryptography is faced with a noise-free channel, and the processing object is bit data. In PLE, the effects of channel and noise can be exploited to enhance security and prevent deciphering. In addition, since the processing object is complex vector signals, there are more operational functions to select and design for PLE. The mathematical models, design frameworks, and cryptographic primitives of PLE are established. Two design frameworks are proposed: stream PLE and block PLE. For stream PLE, a new 3D security constellation mapping is derived. For block PLE, two types of sub-transforms are defined: isometry transformations and stochastic transformations. Furthermore, a practical system operation mode PLE-block chaining (PBC) is proposed to enhance the practical system security. The proposed PLE framework can resist known plaintext attacks and chosen-plaintext attacks. The simulation shows that the proposed isometry transformation method has good performances in terms of bit error ratio (BER) penalty and confusion degree.

This paper reports on the design and three-dimensional (3D) integration of low-cost, low-loss and easy-to-fabricate 3D-printed integrated lens antennas (ILAs) for 5G broadband wireless communications in the 28 GHz frequency range. The ILA designs consist of an extended hemispherical lens, fed by two different types of source antennas, a substrate integrated waveguide (SIW) slot antenna array and a microstrip patch antenna (MPA) array. Results from comprehensive parametric analyses of the infill pattern and density of the 3D printed dielectric lenses are also intensively investigated and characterized for their electromagnetic properties, e.g., electric-field distribution. The ILAs are fabricated using polylactic acid (PLA) as the fused deposition modelling (FDM) polymer with an optimized infill density of 50%, which speeds up prototyping time and decreases the relative permittivity, dielectric loss, manufacturing cost, and overall mass of the lens. These features are illustrated through experimental verification and characterization of at least three samples. From the measurement results, the ILAs achieve a fractional bandwidth of 10.7%, ranging from 26.5 to 29.4 GHz with a maximum gain of 15.6 dBi at boresight and half power beam-width of approximately 58° and 75° in the E and H planes, respectively.

The existing sub-6 GHz band is insufficient to support the bandwidth requirement of emerging data-rate-hungry applications and Internet of Things devices, requiring ultrareliable low latency communication (URLLC), thus making the migration to millimeter-wave (mmWave) bands inevitable. A notable disadvantage of a mmWave band is the significant losses suffered at higher frequencies that may not be overcome by novel optimization algorithms at the transmitter and receiver and thus result in a performance degradation. To address this, Intelligent Reflecting Surface (IRS) is a new technology capable of transforming the wireless channel from a highly probabilistic to a highly deterministic channel and as a result, overcome the significant losses experienced in the mmWave band. This paper aims to survey the design and applications of an IRS, a 2-dimensional (2D) passive metasurface with the ability to control the wireless propagation channel and thus achieve better spectral efficiency (SE) and energy efficiency (EE) to aid the fifth and beyond generation to deliver the required data rate to support current and emerging technologies. It is imperative that the future wireless technology evolves toward an intelligent software paradigm, and the IRS is expected to be a key enabler in achieving this task. This work provides a detailed survey of the IRS technology, limitations in the current research, and the related research opportunities and possible solutions.

In this paper, we develop a comprehensive statistical framework to characterize and model large-scale unmanned aerial vehicle-enabled post-disaster recovery cellular networks. In the case of natural or man-made disasters, the cellular network is vulnerable to destruction resulting in coverage voids or coverage holes. Drone-based small cellular networks (DSCNs) can be rapidly deployed to fill such coverage voids. Due to capacity and back-hauling limitations on drone small cells (DSCs), each coverage hole requires a multitude of DSCs to meet the shortfall coverage at a desired quality-of-service. Moreover, ground users also tend to cluster in hot-spots in a post-disaster scenario. Motivated by this fact, we consider the clustered deployment of DSCs around the site of a destroyed BS. Joint consideration partially operating BSs and deployed DSCs yields a unique topology for such public safety networks. Borrowing tools from stochastic geometry, we develop a statistical framework to quantify the down-link performance of a DSCN. Our proposed clustering mechanism extends the traditional Matern and Thomas cluster processes to a more general case, where cluster size is dependent upon the size of the coverage hole. We then employ the newly developed framework to find closed-form expressions (later verified by Monte-Carlo simulations) to quantify the coverage probability, area spectral efficiency, and the energy efficiency for the down-link mobile user. Finally, we explore several design parameters (for both of the adopted cluster processes) that address optimal deployment of the network (i.e., number of drones per cluster, drone altitudes, and transmit power ratio between the traditional surviving base stations and the drone base stations).

The deployment of heterogeneous networks (HetNets) inevitably demands the design of interference management techniques to elevate the overall network performance. This paper presents a novel interference mitigation technique known as reverse frequency allocation (RFA), which provides an efficient resource allocation compared with the other state-of-the-art techniques. RFA reverses the transmission direction of interferers, thereby minimizing the cross-tier interference. Eventually, better coverage as well as increased data rates are achieved by providing complementary spectrum to the macro and pico users. In this paper, we present a tractable approach for modeling HetNets under the proposed RFA scheme. Specifically, we employ well known tools from stochastic geometry to derive closed-form expressions for the coverage probability and rate coverage in two-tier cellular network employing RFA and its variants. The modeling is performed using two approaches; first, where the base stations and users are modeled as independent Poisson point processes (PPPs) and second, the interference is approximated using the fluid model. It is shown that the results obtained from the PPP model are accurate for higher values of path loss exponents, while the results from fluid model are useful for smaller values of path loss exponents. The plausibility of model is validated through the Monte-Carlo simulations and the network performance is evaluated in terms of coverage probability, coverage rate, and outage capacity. The results demonstrate that 2-RFA yields outage capacity gains of 13% as compared with the soft fractional frequency reuse scheme, whereas, the performance gains can be further improved by 14% by employing the proposed variants of RFA.

Weft knitted conductive fabrics can act as excellent textile strain sensors for human motion capture. The loop architecture dictates the overall electrical properties of weft knit strain sensors. Therefore, research into loop architecture is relevant for comprehensively investigating the design space of e-textile sensors. There are three main types of knit stitches, Knitted loop stitch, Miss stitch, and Tuck stitch. Nevertheless, most of the research into weft knit strain sensors has largely focused on fabrics with only knitted loop stitches. Miss and tuck stitches will affect the contact points in the sensor and, consequently, its piezoresistivity. Therefore, this paper investigates the impact of incorporating miss and tuck stitches on the piezoresistivity of a weft knit sensor. Particularly, the electromechanical models of a miss stitch and a tuck stitch in a weft knit sensor are proposed. These models were used in order to develop loop configurations of sensors that consist of various percentages of miss or tuck stitches. Subsequently, the developed loop configurations were simulated while using LTspice and MATLAB software; and, verified experimentally through a tensile test. The experimental results closely agree with the simulated results. Furthermore, the results reveal that increases in the percentage of tuck or miss stitches in weft knit sensor decrease the initial and average resistance of the sensor. In addition, it was observed that, although the piezoresistivity of a sensor with tuck or miss stitches is best characterised as a quadratic polynomial, increases in the percentage of tuck stitches in the sensor increase the linearity of the sensor’s piezoresistivity.

Internet of Things (IoT)-based automation of agricultural events can change the agriculture sector from being static and manual to dynamic and smart, leading to enhanced production with reduced human efforts. Precision Agriculture (PA) along with Wireless Sensor Network (WSN) are the main drivers of automation in the agriculture domain. PA uses specific sensors and software to ensure that the crops receive exactly what they need to optimize productivity and sustainability. PA includes retrieving real data about the conditions of soil, crops and weather from the sensors deployed in the fields. High-resolution images of crops are obtained from satellite or air-borne platforms (manned or unmanned), which are further processed to extract information used to provide future decisions. In this paper, a review of near and remote sensor networks in the agriculture domain is presented along with several considerations and challenges. This survey includes wireless communication technologies, sensors, and wireless nodes used to assess the environmental behaviour, the platforms used to obtain spectral images of crops, the common vegetation indices used to analyse spectral images and applications of WSN in agriculture. As a proof of concept, we present a case study showing how WSN-based PA system can be implemented. We propose an IoT-based smart solution for crop health monitoring, which is comprised of two modules. The first module is a wireless sensor network-based system to monitor real-time crop health status. The second module uses a low altitude remote sensing platform to obtain multi-spectral imagery, which is further processed to classify healthy and unhealthy crops. We also highlight the results obtained using a case study and list the challenges and future directions based on our work.