Explore the vast range of titles available.

Embedded systems technology is undergoing a phase of transformation owing to the novel advancements in computer architecture and the breakthroughs in machine learning applications. The areas of applications of embedded machine learning (EML) include accurate computer vision schemes, reliable speech recognition, innovative healthcare, robotics, and more. However, there exists a critical drawback in the efficient implementation of ML algorithms targeting embedded applications. Machine learning algorithms are generally computationally and memory intensive, making them unsuitable for resource-constrained environments such as embedded and mobile devices. In order to efficiently implement these compute and memory-intensive algorithms within the embedded and mobile computing space, innovative optimization techniques are required at the algorithm and hardware levels. To this end, this survey aims at exploring current research trends within this circumference. First, we present a brief overview of compute intensive machine learning algorithms such as hidden Markov models (HMM), k-nearest neighbors (k-NNs), support vector machines (SVMs), Gaussian mixture models (GMMs), and deep neural networks (DNNs). Furthermore, we consider different optimization techniques currently adopted to squeeze these computational and memory-intensive algorithms within resource-limited embedded and mobile environments. Additionally, we discuss the implementation of these algorithms in microcontroller units, mobile devices, and hardware accelerators. Conclusively, we give a comprehensive overview of key application areas of EML technology, point out key research directions and highlight key take-away lessons for future research exploration in the embedded machine learning domain.

The growth in artificial intelligence and its applications has led to increased data processing and inference requirements. Traditional cloud-based inference solutions are often used but may prove inadequate for applications requiring near-instantaneous response times. This review examines Tiny Machine Learning, also known as TinyML, as an alternative to cloud-based inference. The review focuses on applications where transmission delays make traditional Internet of Things (IoT) approaches impractical, thus necessitating a solution that uses TinyML and on-device inference. This study, which follows the PRISMA guidelines, covers TinyML’s use cases for real-world applications by analyzing experimental studies and synthesizing current research on the characteristics of TinyML experiments, such as machine learning techniques and the hardware used for experiments. This review identifies existing gaps in research as well as the means to address these gaps. The review findings suggest that TinyML has a strong record of real-world usability and offers advantages over cloud-based inference, particularly in environments with bandwidth constraints and use cases that require rapid response times. This review discusses the implications of TinyML’s experimental performance for future research on TinyML applications.

Standard-sized autonomous vehicles have rapidly improved thanks to the breakthroughs of deep learning. However, scaling autonomous driving to mini-vehicles poses several challenges due to their limited on-board storage and computing capabilities. Moreover, autonomous systems lack robustness when deployed in dynamic environments where the underlying distribution is different from the distribution learned during training. To address these challenges, we propose a closed-loop learning flow for autonomous driving mini-vehicles that includes the target deployment environment in-the-loop. We leverage a family of compact and high-throughput tinyCNNs to control the mini-vehicle that learn by imitating a computer vision algorithm, i.e., the expert, in the target environment. Thus, the tinyCNNs, having only access to an on-board fast-rate linear camera, gain robustness to lighting conditions and improve over time. Moreover, we introduce an online predictor that can choose between different tinyCNN models at runtime—trading accuracy and latency—which minimises the inference’s energy consumption by up to 3.2×. Finally, we leverage GAP8, a parallel ultra-low-power RISC-V-based micro-controller unit (MCU), to meet the real-time inference requirements. When running the family of tinyCNNs, our solution running on GAP8 outperforms any other implementation on the STM32L4 and NXP k64f (traditional single-core MCUs), reducing the latency by over 13× and the energy consumption by 92%.

Industrial assets often feature multiple sensing devices to keep track of their status by monitoring certain physical parameters. These readings can be analyzed with machine learning (ML) tools to identify potential failures through anomaly detection, allowing operators to take appropriate corrective actions. Typically, these analyses are conducted on servers located in data centers or the cloud. However, this approach increases system complexity and is susceptible to failure in cases where connectivity is unavailable. Furthermore, this communication restriction limits the approach’s applicability in extreme industrial environments where operating conditions affect communication and access to the system. This paper proposes and evaluates an end-to-end adaptable and configurable anomaly detection system that uses the Internet of Things (IoT), edge computing, and Tiny-MLOps methodologies in an extreme industrial environment such as submersible pumps. The system runs on an IoT sensing Kit, based on an ESP32 microcontroller and MicroPython firmware, located near the data source. The processing pipeline on the sensing device collects data, trains an anomaly detection model, and alerts an external gateway in the event of an anomaly. The anomaly detection model uses the isolation forest algorithm, which can be trained on the microcontroller in just 1.2 to 6.4 s and detect an anomaly in less than 16 milliseconds with an ensemble of 50 trees and 80 KB of RAM. Additionally, the system employs blockchain technology to provide a transparent and irrefutable repository of anomalies.

For the next coming years, metaverse, digital twin and autonomous vehicle applications are the leading technologies for many complex applications hitherto inaccessible such as health and life sciences, smart home, smart agriculture, smart city, smart car and logistics, Industry 4.0, entertainment (video game) and social media applications, due to recent tremendous developments in process modeling, supercomputing, cloud data analytics (deep learning, etc.), communication network and AIoT/IIoT/IoT technologies. AIoT/IIoT/IoT is a crucial research field because it provides the essential data to fuel metaverse, digital twin, real-time Industry 4.0 and autonomous vehicle applications. However, the science of AIoT is inherently multidisciplinary, and therefore, it is difficult for readers to understand its evolution and impacts. Our main contribution in this article is to analyze and highlight the trends and challenges of the AIoT technology ecosystem including core hardware (MCU, MEMS/NEMS sensors and wireless access medium), core software (operating system and protocol communication stack) and middleware (deep learning on a microcontroller: TinyML). Two low-powered AI technologies emerge: TinyML and neuromorphic computing, but only one AIoT/IIoT/IoT device implementation using TinyML dedicated to strawberry disease detection as a case study. So far, despite the very rapid progress of AIoT/IIoT/IoT technologies, several challenges remain to be overcome such as safety, security, latency, interoperability and reliability of sensor data, which are essential characteristics to meet the requirements of metaverse, digital twin, autonomous vehicle and Industry 4.0. applications.

The rapid emergence of low-power embedded devices and modern machine learning (ML) algorithms has created a new Internet of Things (IoT) era where lightweight ML frameworks such as TinyML have created new opportunities for ML algorithms running within edge devices. In particular, the TinyML framework in such devices aims to deliver reduced latency, efficient bandwidth consumption, improved data security, increased privacy, lower costs and overall network cost reduction in cloud environments. Its ability to enable IoT devices to work effectively without constant connectivity to cloud services, while nevertheless providing accurate ML services, offers a viable alternative for IoT applications seeking cost-effective solutions. TinyML intends to deliver on-premises analytics that bring significant value to IoT services, particularly in environments with limited connection. This review article defines TinyML, presents an overview of its benefits and uses and provides background information based on up-to-date literature. Then, we demonstrate the TensorFlow Lite framework which supports TinyML along with analytical steps for an ML model creation. In addition, we explore the integration of TinyML with network technologies such as 5G and LPWAN. Ultimately, we anticipate that this analysis will serve as an informational pillar for the IoT/Cloud research community and pave the way for future studies.

Vehicles are the major source of air pollution in modern cities, emitting excessive levels of CO2 and other noxious gases. Exploiting the OBD-II interface available on most vehicles, the continuous emission of such pollutants can be indirectly measured over time, although accuracy has been an important design issue when performing this task due the nature of the retrieved data. In this scenario, soft-sensor approaches can be adopted to process engine combustion data such as fuel injection and mass air flow, processing them to estimate pollution and transmitting the results for further analyses. Therefore, this article proposes a soft-sensor solution based on an embedded system designed to retrieve data from vehicles through their OBD-II interface, processing different inputs to provide estimated values of CO2 emissions over time. According to the type of data provided by the vehicle, two different algorithms are defined, and each follows a comprehensive mathematical formulation. Moreover, an unsupervised TinyML approach is also derived to remove outliers data when processing the computed data stream, improving the accuracy of the soft sensor as a whole while not requiring any interaction with cloud-based servers to operate. Initial results for an embedded implementation on the Freematics ONE+ board have shown the proposal’s feasibility with an acquisition frequency equal to 1Hz and emission granularity measure of gCO2/km.

The penetration of wearable devices in our daily lives is unstoppable. Although they are very popular, so far, these elements provide a limited range of services that are mostly focused on monitoring tasks such as fitness, activity, or health tracking. Besides, given their hardware and power constraints, wearable units are dependent on a master device, e.g., a smartphone, to make decisions or send the collected data to the cloud. However, a new wave of both communication and artificial intelligence (AI)-based technologies fuels the evolution of wearables to an upper level. Concretely, they are the low-power wide-area network (LPWAN) and tiny machine-learning (TinyML) technologies. This paper reviews and discusses these solutions, and explores the major implications and challenges of this technological transformation. Finally, the results of an experimental study are presented, analyzing (i) the long-range connectivity gained by a wearable device in a university campus scenario, thanks to the integration of LPWAN communications, and (ii) how complex the intelligence embedded in this wearable unit can be. This study shows the interesting characteristics brought by these state-of-the-art paradigms, concluding that a wide variety of novel services and applications will be supported by the next generation of wearables.

Driven by the rapid development of the Internet of Things (IoT), deploying deep learning models on resource-constrained hardware has become an increasingly critical challenge, which has propelled the emergence of TinyML as a viable solution. This study aims to deploy lightweight deep learning models for human activity recognition (HAR) using TinyML on edge devices. We designed and evaluated three models: a 2D Convolutional Neural Network (2D CNN), a 1D Convolutional Neural Network (1D CNN), and a DeepConv LSTM. Among these, the DeepConv LSTM outperformed existing lightweight models by effectively capturing both spatial and temporal features, achieving an accuracy of 98.24% and an F1 score of 98.23%. After performing full integer quantization on the best model, its size was reduced from 513.23 KB to 136.51 KB and was successfully deployed on the Arduino Nano 33 BLE Sense Rev2 using the Edge Impulse platform. The device’s memory usage was 29.1 KB, flash usage was 189.6 KB, and the model’s average inference time was 21 milliseconds, requiring approximately 0.01395 GOP, with a computational performance of around 0.664 GOPS. Even after quantization, the model maintained an accuracy of 97% and an F1 score of 97%, ensuring efficient utilization of computational resources. This deployment highlights the potential of TinyML in achieving low-latency and efficient HAR systems, making it suitable for real-time human activity recognition applications.

This paper proposes small and efficient machine learning models (TinyML) for resource-constrained edge devices, specifically for on-device indoor localisation. Typical approaches for indoor localisation rely on centralised remote processing of data transmitted from lower powered devices such as wearables. However, there are several benefits for moving this to the edge device itself, including increased battery life, enhanced privacy, reduced latency and lowered operational costs, all of which are key for common applications such as health monitoring. The work focuses on model compression techniques, including quantization and knowledge distillation, to significantly reduce the model size while maintaining high predictive performance. We base our work on a large state-of-the-art transformer-based model and seek to deploy it within low-power MCUs. We also propose a state-space-based architecture using Mamba as a more compact alternative to the transformer. Our results show that the quantized transformer model performs well within a 64 KB RAM constraint, achieving an effective balance between model size and localisation precision. Additionally, the compact Mamba model has strong performance under even tighter constraints, such as a 32 KB of RAM, without the need for model compression, making it a viable option for more resource-limited environments. We demonstrate that, through our framework, it is feasible to deploy advanced indoor localisation models onto low-power MCUs with restricted memory limitations. The application of these TinyML models in healthcare has the potential to revolutionize patient monitoring by providing accurate, real-time location data while minimising power consumption, increasing data privacy, improving latency and reducing infrastructure costs.