Kolmogorov–Arnold Networks (KANs) represent a promising machine learning architecture that leverages univariate functional decomposition to model complex phenomena using compact and interpretable structures. These characteristics make KANs especially attractive for deployment in TinyML environments, where memory, processing power, and energy consumption are strictly constrained. This paper evaluates the feasibility and trade-offs of using a KAN model to estimate the State of Charge (SoC) in electric vehicle batteries. We design a KAN tailored for embedded systems and compare its performance with a conventional Multilayer Perceptron (MLP) baseline under identical training and deployment conditions. Our evaluation includes predictive accuracy, training cost, model size, inference speed, and energy consumption on microcontrollers. Results show that the KAN model achieves a nearly ten times smaller memory footprint than the MLP (693 bytes vs. 6807 bytes) and maintains comparable energy consumption and inference speed across different embedded platforms. Although the MLP outperforms the KAN during training with faster convergence and lower energy requirements, the KAN demonstrates competitive predictive performance at inference time while significantly reducing deployment costs in terms of memory usage and energy efficiency at the edge. Furthermore, the KAN model produces symbolic mathematical expressions, offering direct interpretability and facilitating analytical validation — a critical advantage for embedded battery diagnostics and safety-critical applications.

The persistence of wildfires in natural landscapes calls for innovative early detection methods that leverage cutting-edge technologies. Traditional approaches, which rely solely on visual sensors or isolated devices, while valuable, often fall short in terms of accuracy, cost, scalability, and contextual adaptability. In response to these challenges, this paper introduces a novel fire detection system that integrates a sensor-based model with a dynamically triggered visual analysis module at edge devices. Central to our approach is a multi-sensor monitoring architecture that employs a TinyML classifier to continuously monitor environmental conditions under strict energy constraints. Upon detecting potential fire indicators, the system promptly activates a visual sensor that uses a camera platform to adjust its orientation based on the target position, capturing and analyzing images through a lightweight Convolutional Neural Network (CNN). This proposed system achieves an accuracy of up to 92%, while the quantized CNN models deliver an 83% reduction in inference time and a 74% and 70% decrease in peak RAM and Flash usage, respectively. Simulations also demonstrated that the system reduced the false-positive rate with minimal power increase.

This study addresses the challenges of dependability in Wireless Sensor Networks by proposing a Machine Learning-based approach using Convolutional Neural Networks for network planning for smart cities. Simulated scenarios were used to train the model, which predicts sensor placement and communication configurations to optimize coverage and availability. Results show significant improvements, including an average of 10.7% increase in dependability index and a rise in area coverage from 59% to 73% in 7-node networks, while reducing path failure rates by 27.6%. The method proves effective for enhancing WSN performance and adaptability in safety-critical applications.

The Internet of Things (IoT) has made it possible to include everyday objects in a connected network, allowing them to intelligently process data and respond to their environment. Thus, it is expected that those objects will gain an intelligent understanding of their environment and be able to process data more efficiently than before. Particularly, such edge computing paradigm has allowed the execution of inference methods on resource-constrained devices such as microcontrollers, significantly changing the way IoT applications have evolved in recent years. However, although this scenario has supported the development of Tiny Machine Learning (TinyML) approaches on such devices, there are still some challenges that require further investigation when optimizing data streaming on the edge. Therefore, this article proposes a new unsupervised TinyML regression technique based on the typicality and eccentricity of the samples to be processed. Moreover, the proposed technique also exploits a Recursive Least Squares (RLS) filter approach. Combining all these features, the proposed method uses similarities between samples to identify patterns when processing data streams, predicting outcomes based on these patterns. The results obtained through the extensive experimentation utilizing vehicular data streams were highly encouraging. The proposed algorithm was meticulously compared with the RLS algorithm and Convolutional Neural Networks (CNN). It exhibited significantly superior performance, with mean squared errors that were 4.68 and 12.02 times lower, respectively, compared to the aforementioned techniques.

The Internet of Things (IoT) paradigm has fostered several transformations in various industrial sectors, with important improvements in the automotive industry. Actually, the number of sensors and the computational power of modern vehicles have grown significantly, providing an opportunity for instrumentation, monitoring, and creation of increasingly efficient diagnostic algorithms. In fact, it is known that diagnosis is an essential requirement since the way of driving may have significant impacts in different contexts, such as traffic safety, fuel consumption, emissions, and maintenance, among others. Furthermore, solutions generally available in the literature for analyzing drivers’ behavior have focused on supervised offline learning models, fed with an entire dataset for training and testing. In this context, this paper proposes an approach for detecting drivers’ driving events, exploiting for that unsupervised online data flows and a specialized machine learning algorithm. The validation of the proposal was carried out with a case study in a real scenario with different conditions, which allowed the identification of daily driving operations. The results demonstrated the feasibility of the proposal as well as the identification of the different intended events.