Nowadays, lithium-ion is among the most used chemistry for batteries and shows an increasing market growth rate. However, to reduce failure or safety risks, the battery State-of-Charge (SoC) must be accurately monitored and predicted by a suitable Battery Management System (BMS). Artificial Intelligence (AI) techniques have been extensively applied to this field with good results. Typically, AIs are trained on dynamic profile data, emulating battery charging and discharging cycles related to the application under test. In this paper, a novel approach is presented: application independent constant current profiles are used to train a Support Vector Regression (SVR) algorithm. To enhance the estimation accuracy, the output of the obtained SVR model was post-processed. Finally, an error correction algorithm was applied to further reduce the estimation error. The system is validated over test cycles representing different application scenarios for the battery cell operations. For the development of the proposed approach, a total of 105 constant current discharge profiles for the training and 20 realistic test cycles for the validation have been considered, including standard automotive cycles and a generic battery powered power tool. The performance in the SoC estimation resulted in a Root Mean Square Error (RMSE) of 0.94% and a Mean Absolute Error (MAE) of 0.75% over all the test cycles. Error metrics are comparable to those obtained for SoC estimation AI algorithms based on traditional approaches using application dependent battery profiles for the training phase.

Monitoring the State of Charge (SoC) in battery cells is necessary to avoid damage and to extend battery life. Support Vector Machine (SVM) algorithms and Machine Learning techniques in general can provide real-time SoC estimation without the need to design a cell model. In this work, an SVM was trained by applying an Ant Colony Optimization method. The obtained trained model was 10-fold cross-validated and then designed in Hardware Description Language to be run on FPGA devices, enabling the design of low-cost and compact hardware. Thanks to the choice of a linear SVM kernel, the implemented architecture resulted in low resource usage (about 1.4% of Xilinx Artix7 XC7A100TFPGAG324C FPGA), allowing multiple instances of the SVM SoC estimator model to monitor multiple battery cells or modules, if needed. The ability of the model to maintain its good performance was further verified when applied to a dataset acquired from different driving cycles to the cycle used in the training phase, achieving a Root Mean Square Error of about 1.4%.

The potential of a voltametric E-tongue coupled with a custom data pre-processing stage to improve the performance of machine learning techniques for rapid discrimination of tomato purées between cultivars of different economic value has been investigated. To this aim, a sensor array with screen-printed carbon electrodes modified with gold nanoparticles (GNP), copper nanoparticles (CNP) and bulk gold subsequently modified with poly(3,4-ethylenedioxythiophene) (PEDOT), was developed to acquire data to be transformed by a custom pre-processing pipeline and then processed by a set of commonly used classifiers. The GNP and CNP-modified electrodes, selected based on their sensitivity to soluble monosaccharides, demonstrated good ability in discriminating samples of different cultivars. Among the different data analysis methods tested, Linear Discriminant Analysis (LDA) proved to be particularly suitable, obtaining an average F1 score of 99.26%. The pre-processing stage was beneficial in reducing the number of input features, decreasing the computational cost, i.e., the number of computing operations to be performed, of the entire method and aiding future cost-efficient hardware implementation. These findings proved that coupling the multi-sensing platform featuring properly modified sensors with the custom pre-processing method developed and LDA provided an optimal tradeoff between analytical problem solving and reliable chemical information, as well as accuracy and computational complexity. These results can be preliminary to the design of hardware solutions that could be embedded into low-cost portable devices.

This paper introduces a flexible multi-channel platform designed for integrating multiple sensors, among which electronic noses (e-nose) and electronic tongues (e-tongue), targeting food quality and safety assessments. The proposed device is designed aiming for maximum portability, allowing on-site analyses, and maximum modularity, to meet user requirements. The customizable configurations allow the user to adjust the number of channels, current ranges, and voltage settings and to perform different analyses simultaneously. A variable number of sensing modules with different features can simultaneously be interfaced with the same control board, equipped with a microcontroller unit and a wireless module to both process data for on-board analyses and send it to the cloud. The modular design of the system allows for interchangeable sensing modules, offering high flexibility in sensory analysis compared to the existing portable solutions. Two custom dual-channel module boards have been designed with different operating current ranges, minimizing power consumption for extended portability, requiring only 3.16mA per channel during operations of the high-current module and 2.11mA for the low-current module. Device characterization demonstrates high linearity, with an R-squared higher than 0.999 over the entire voltage range. The channel-to-channel gain mismatch is limited to 0.26%. Spectral analysis shows 20 μVrms of equivalent noise, leading to an equivalent minimum sensor resolution of 16 noise-free bits.

Cytokine detection—particularly interleukin-6 (IL-6)—is of growing interest due to its role in chronic inflammation and acute conditions like sepsis, where elevated concentrations above physiological levels require urgent clinical attention. Given the strong correlation between sweat and blood IL-6 levels, wearable sweat-based detection offers a promising non-invasive monitoring approach. Electrochemical sensors are especially well-suited for such applications due to their high sensitivity and accuracy. Among these, electrochemical impedance spectroscopy (EIS) provides excellent sensitivity but is challenging to implement in embedded, low-power systems. Conversely, cyclic voltammetry (CV) is more compatible with low-power, wearable devices but generally lacks the sensitivity needed for ultra-low concentration detection. To advance CV-based IL-6 sensing, this study establishes the core electrochemical and analytical performance of a novel CV-based sensor under controlled laboratory conditions. We developed a flexible electrochemical biosensor using a screen-printed carbon three-electrode system functionalized with gold nanoparticles and IL-6-specific aptamers. Electrochemical characterization was conducted using both EIS and CV with a benchtop potentiostat in 1x phosphate-buffered saline (PBS). To enhance threshold-based classification, machine learning (ML) was integrated with CV data analysis. Using a k-Nearest Neighbors (KNN) algorithm, we classified CV datasets with 96.7% accuracy, distinguishing IL-6 concentrations in 1x PBS as physiological (20 pg/mL). These findings demonstrate the potential of combining electrochemical sensing with ML for sensitive, threshold-based detection of ultra-low analyte concentrations, supporting future translation toward wearable health monitoring platforms.