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).

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.

Emerging technologies are continually redefining the paradigms of smart farming and opening up avenues for more precise and informed farming practices. A tiny machine learning (TinyML)‐based framework is proposed for unmanned aerial vehicle (UAV)‐assisted smart farming applications. The practical deployment of such a framework on the UAV and bespoke internet of things (IoT) sensors which measure soil moisture and ambient environmental conditions is demonstrated. The key objective of this framework is to harness TinyML for implementing transfer learning (TL) using deep neural networks (DNNs) and long short‐term memory (LSTM) ML models. As a case study, this framework is employed to predict soil moisture content for smart agriculture applications, guiding optimal water utilisation for crops through time‐series forecasting models. To the best of authors’ knowledge, a framework which leverages UAV‐assisted TL for the edge internet of things using TinyML has not been investigated previously. The TL‐based framework employs a pre‐trained data model on different but similar applications and data domains. Not only do the authors demonstrate the practical deployment of the proposed framework but they also quantify its performance through real‐world deployment. This is accomplished by designing a custom sensor board for soil and environmental sensing which uses an ESP32 microcontroller unit. The inference metrics (i.e. inference time and accuracy) are measured for different ML model architectures on edge devices as well as other performance metrics (i.e. mean square error and coefficient of determination [ R ² ]), while emphasising the need for balancing accuracy and processing complexity. In summary, the results show the practical feasibility of using drones to deliver TL for DNN and LSTM models to ultra‐low performance edge IoT devices for soil humidity prediction. But in general, this work also lays the foundation for further research into other applications of TinyML usage in many different aspects of smart farming.

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.

The advancement of sustainable energy sources necessitates the development of robust forecasting tools for efficient energy management. A prominent player in this domain, solar power, heavily relies on accurate energy yield predictions to optimize production, minimize costs, and maintain grid stability. This paper explores an innovative application of tiny machine learning to provide real-time, low-cost forecasting of solar energy yield on resource-constrained edge internet of things devices, such as micro-controllers, for improved residential and industrial energy management. To further contribute to the domain, we conduct a comprehensive evaluation of four prominent machine learning models, namely unidirectional long short-term memory, bidirectional gated recurrent unit, bidirectional long short-term memory, and simple bidirectional recurrent neural network, for predicting solar farm energy yield. Our analysis delves into the impacts of tuning the machine learning model hyperparameters on the performance of these models, offering insights to improve prediction accuracy and stability. Additionally, we elaborate on the challenges and opportunities presented by the implementation of machine learning on low-cost energy management control systems, highlighting the benefits of reduced operational expenses and enhanced grid stability. The results derived from this study offer significant implications for energy management strategies at both household and industrial scales, contributing to a more sustainable future powered by accurate and efficient solar energy forecasting.