Numerous health conditions, such as obesity, diabetes, and cardiovascular diseases, require strict adherence to nutritional guidelines and accurate reporting of eating behaviors, making effective eating monitoring essential. A common approach to eating monitoring involves maintaining a food diary, where subjects manually self-report eating events, a process inherently prone to imprecision. Recent technological advances have enabled the development of passive, automatic eating detection systems, typically relying on data from wearable devices to identify eating events. In this context, the literature is vast on efforts that use machine learning methods for this purpose, with great success. However, most existing studies focus only on eating detection mechanisms and fail to offer an integrated solution with practical use cases. To address this gap, in this work, we present a cyber–physical systems approach to eating monitoring that integrates an eating event detection module with a cloud-based service-oriented backbone where numerous services are deployed, yielding an integrated solution for real-time eating monitoring.

This study presents the development of a data-driven model for the copper electro-refining process, designed to predict electrolyte composition and evaluate operational scenarios. The system was implemented using historical plant data and is based on a recurrent neural network with an LSTM encoder–decoder architecture. The model was trained to predict concentrations in the electrolyte and performance variables, employing root mean square error (RMSE) as the loss function. Results demonstrate that the proposed model delivers reasonably accurate predictions, provides a practical tool for scenario evaluation, supports decision-making, and facilitates the transition of copper electro-refineries to Industry 4.0.

As industrial cyber-physical systems (ICPSs) consolidate as mature automation solutions, questions about leveraging their flexibility and resilience to maintain performance have arisen. This article presents a methodology that employs industrial agents (IAs) to supervise and reconfigure service-oriented ICPSs, thereby enabling adaptability and improving performance. The main tasks of the IAs include: 1) selecting the best service for the ICPS, from a predefined component library, based on performance metrics and computational load; and 2) monitoring the performance of the chosen service online, carrying out a new selection process if poor performance is detected, thus enabling dynamic service selection. A key contribution is the introduction of a dynamic reconfiguration capability to the system, allowing it to adapt in real-time to changing conditions, thereby addressing a limitation of existing service-oriented architectures. Results show an improvement in system adaptability and performance, demonstrating the potential of agent-based supervision to support the operation of ICPSs and positioning the proposed methodology as an initial step toward the development of more resilient and efficient ICPSs.

Efficient complaint management is essential for service companies aiming to maintain high levels of customer satisfaction. Traditionally, complaint resolution has relied on human analysts performing manual processes, which are time-consuming and error-prone. As an alternative, rule-based systems have emerged as an initial step toward automation, seeking to make complaint resolution a reliable process independent of analysts’ judgment. However, these systems are often inflexible and limited in understanding and responding to complex, unstructured customer feedback, making analysts’ interventions unavoidable. In this context, recent advancements in natural language understanding by large language models (LLMs) have opened new opportunities for developing complaint management systems that leverage the capabilities of LLMs to classify, prioritize, and resolve complaints with minimal human intervention. This work addresses this opportunity by creating an integrated complaint management system, using complaints in the electric sector as a case study. Based on a service-oriented architecture, the proposed system integrates a data processing microservice, an artificial intelligence module deploying LLM-based virtual analysts, and a user-friendly web application. Extensive evaluations, including load testing and latency analysis, demonstrate the system’s performance, robustness, and scalability.

Noisy measurements are a common challenge in the process industry, often impacting the closed-loop performance of control systems. Traditional approaches, such as classical filters, can reduce noise but often introduce delays or artifacts that degrade overall system performance. Recently, learning-based denoisers have emerged as a powerful alternative. Among these, contrastive blind denoising autoencoders (CBDAEs) have shown promising results in cleaning noisy sensor data. However, their application in industrial control systems remains largely unexplored, and their potential to improve performance is not yet well understood. To address this gap, this work presents a systematic evaluation of a networked control system that incorporates a CBDAE in the loop. Copyright (c) 2025 The Authors. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0/)

Due to the inherent difficulties in accessing the state of a dynamical system, observers play a relevant role in modern state-feedback control schemes. Traditionally, observer synthesis has been tackled using models derived from first principles. Recently, data-driven methods capable of estimating the state directly from real data have gained relevance, although an important limitation lies in the physical interpretation of the variables. In this work, a data-driven observer based on an autoencoder neural network is proposed. By training the network using a (possibly noisy) database containing states and output measurements, the resulting observer is able to estimate the real states. Experimental results on a circuital model are presented to illustrate the potential of the proposed method.

Modularity, understood as the property that allows a complex system to be decomposed into well-defined compartments, is a distinctive element in modern engineering systems. Modularity increases scalability, facilitates mass production, and simplifies maintenance. Nonetheless, modular systems may exhibit complex dynamics, and the establishment of their structural properties can be challenging. In this brief, observability and observer-based control for a class of modular nonlinear systems are analyzed, with particular focus on the modular multilevel power converter. Analytical conditions to verify observability are given, a separation principle that enables observer-based control is formulated, and an experimental validation on a pilot-scale converter is conducted to corroborate the analytical findings.

Industrial Cyber-Physical Systems (ICPSs) are regarded as the enabling technology of the ongoing fourth industrial revolution. Despite their recognized advantages, ICPSs face challenges that need to be addressed to unleash their full potential. Among these challenges is the standardization of information models. Currently, various industry standards are used for modeling information. However, these efforts primarily focus on describing the engineering and logic aspects of industrial processes, without considering all dimensions of an ICPS. As a step towards closing this gap, this work proposes a novel approach based on multilayer graphs to generate a flexible information model that addresses three key aspects of an ICPS: engineering, production flow, and automation. To illustrate the effectiveness of the proposed approach, a specific case study on a mineral processing plant is presented.

Accurate short- and mid-term blood glucose predictions are crucial for patients with diabetes struggling to maintain healthy glucose levels, as well as for individuals at risk of developing the disease. Consequently, numerous efforts from the scientific community have focused on developing predictive models for glucose levels. This study harnesses physiological data collected from wearable sensors to construct a series of data-driven models based on deep learning approaches. We systematically compare these models to offer insights for practitioners and researchers venturing into glucose prediction using deep learning techniques. Key questions addressed in this work encompass the comparison of various deep learning architectures for this task, determining the optimal set of input variables for accurate glucose prediction, comparing population-wide, fine-tuned, and personalized models, and assessing the impact of an individual's data volume on model performance. Additionally, as part of our outcomes, we introduce a meticulously curated dataset inclusive of data from both healthy individuals and those with diabetes, recorded in free-living conditions. This dataset aims to foster research in this domain and facilitate equitable comparisons among researchers.

Average consensus is a fundamental problem in distributed control that still lacks a general solution when the agents face nonideal conditions and uncertain environments. This note contributes to the topic by addressing average consensus in networks of agents that: 1) interact over a time-varying and nonbalanced environment and 2) face parametric and nonparametric disturbances. It is shown that an adaptive strategy, which combines surplus variables and virtual agents, effectively solves the exact average consensus problem in this nonideal and uncertain setup. A numerical example is presented to illustrate the potential of the proposed approach.

Stability guarantees in networked control systems (NCSs) have traditionally been addressed in the deterministic setting using the concepts of maximum allowable transmit interval and maximum allowable delay (MAD). Motivated by the widespread use of nondeterministic networks in modern industrial NCSs, this note aims to extend the analysis to the stochastic setting by considering a setup where the NCS faces stochastic communication delays, which take values over the MAD with positive probability. Conditions for uniform asymptotic stability in probability are given and their practical relevance is illustrated through a set of experiments performed on a real testbed.

This article presents the design and analysis of a switching high-gain adaptive control scheme for a class of nonlinear systems. Adaptation is included in the scheme to estimate the controller gains, using differential equations whose order can switch between $1$ (integer order) and some real number (fractional order) in the interval $(0,1)$ , depending on the error level. This switching strategy permits obtaining lower values for controller gains due to fractional orders, resulting in improved robustness, while simultaneously guaranteeing fast convergence of the state to the origin due to the integer order, leading to a better balance between system behavior and control energy efficiency. Applications to multiagent systems are presented to illustrate the potential of the proposed scheme.

In the pursuit of a personalized healthcare experience, data-driven decision-making has become increasingly relevant. In this context, there is a growing need for technological systems specifically tailored to efficiently manage vast amounts of healthcare data. To address this need, in this work, we contribute by presenting a cyber-physical solution for monitoring and analyzing physiological data obtained from wearable devices. The proposed system is designed following a service-oriented architecture, which promotes modularity and enables efficient data access and analysis. The system is capable of ingesting and consolidating wearable data produced by a variety of commercial devices, scaling to accommodate a large number of data producers, and accepting queries from a multitude of consumers through various mechanisms. Performance tests conducted in various scenarios using real data demonstrate the system’s effectiveness in maintaining real-time data access.

In an industrial IoT setting, ensuring the quality of sensor data is a must when data-driven algorithms operate on the upper layers of the control system. Unfortunately, the common place in industrial facilities is to find sensor time series heavily corrupted by noise and outliers. This work proposes a purely data-driven self-supervised learning-based approach, based on a blind denoising autoencoder, for real time denoising of industrial sensor data. The term blind stresses that no prior knowledge about the noise is required for denoising, in contrast to typical denoising autoencoders. Blind denoising is achieved by using a noise contrastive estimation (NCE) regularization on the latent space of the autoencoder, which not only helps to denoise but also induces a meaningful and smooth latent space that can be exploited in other downstream tasks. Experimental evaluation in both a simulated system and a real industrial process shows that the proposed technique outperforms classical denoising methods.

Visual Question Answering (VQA) models fail catastrophically on questions related to the reading of text-carrying images. However, TextVQA aims to answer questions by understanding the scene texts in an image–question context, such as the brand name of a product or the time on a clock from an image. Most TextVQA approaches focus on objects and scene text detection, which are then integrated with the words in a question by a simple transformer encoder. The focus of these approaches is to use shared weights during the training of a multi-modal dataset, but it fails to capture the semantic relations between an image and a question. In this paper, we proposed a Scene Graph-Based Co-Attention Network (SceneGATE) for TextVQA, which reveals the semantic relations among the objects, the Optical Character Recognition (OCR) tokens and the question words. It is achieved by a TextVQA-based scene graph that discovers the underlying semantics of an image. We create a guided-attention module to capture the intra-modal interplay between the language and the vision as a guidance for inter-modal interactions. To permit explicit teaching of the relations between the two modalities, we propose and integrate two attention modules, namely a scene graph-based semantic relation-aware attention and a positional relation-aware attention. We conduct extensive experiments on two widely used benchmark datasets, Text-VQA and ST-VQA. It is shown that our SceneGATE method outperforms existing ones because of the scene graph and its attention modules.

Industrial cyber-physical systems (ICPSs) are widely regarded as the next generation industrial control systems and as one of the core technologies of the ongoing fourth industrial revolution. Despite its advantages, ICPSs are heavily dependent on the underlying physical process and their synthesis is a customized effort, demanding in terms of resources, which if not conducted carefully may impact the performance of the system. This work proposes a methodology to tackle ICPS synthesis in a systematic way, by using a set of industrial agents that take as input and standardized process description file and automatically deploy a modular ICPS from predesigned functional containers. Concrete examples on a tanks system and an industrial paste thickener are presented to illustrate the potential of the proposed methodology.

The combination of rapid urbanization, population growth, and the hydric stress due to climate change effects demand innovative, optimized approaches to the operation and supervision of urban water services. In Chile, the Superintendency of Sanitary Services has underscored the need for automatized data integration and analysis tools that foster an evidence-based, preventive approach to the supervision of urban water systems. This motivates the development of a pilot supervision and early-warning system, conceived as a cybernetic entity whose objective is to enable efficient access, analysis, and predictive modeling of the data provided by water service companies, so as to identify risks and inefficiencies in water services, monitor their evolution, and anticipate possible failures. This article discusses the development and implementation of a prototype system that provides tools for visualization, statistical and temporal analysis of georeferenced data on water pressures, network disruptions, and client complaints and deploys machine learning capabilities for predicting the quality of service indicators at different locations. The initial implementation in one region of Chile has been shown to expedite the exploitation of data on urban water services, reduce time lags in the detection of service disruptions, and generate evidence for the planning and execution of supervision activities. Based on this successful pilot, a roadmap for geographical and technological expansion is formulated, including other technological, organizational, and regulatory gaps that must be addressed to establish a data-driven framework for the supervision of urban water services.

Timeliness of information is critical for accurate decision making in data-driven systems. As large-scale systems based on Internet of Things (IoT) technologies start populating shop floors, their performance in terms of timeliness must be assessed. This work analyzes age of information (AoI) in IoT-based control systems, where independent control loops share a wireless network. Analytical expressions for AoI under two typical MAC schemes used in IoT systems: CSMA/CA and TDMA are given and validated by simulations. The significance of the results is illustrated in the context of a design example.

Modular multilevel converters (MMCs) have become one of the most popular power converters for medium/high power applications, from transmission systems to motor drives. However, to operate the MMC, typical control schemes use a voltage measurement per submodule (SM), which increases dramatically the number of sensors required to build an MMC, adding complexity in terms of communications and increasing costs, hence limiting its applicability. As an effort to overcome these issues, this article presents a technique based on Kalman filtering for estimating the capacitor voltage at each SM and the converter current. The proposed approach operates both in open and closed-loop, during transients and steady-state, enabling the use of estimator-based state feedback control without the need of a voltage sensor per SM, and filtering the electromagnetic interference from voltage and current sensors. Experiments conducted in a three-phase MMC with 24 SMs confirm the effectiveness of the proposed approach during transients and steady-state operation.

Mobility and transportation services in modern large-scale cities face traffic congestion as one of the main sources of discomfort and economic losses. In this context, taking preventive measures based on traffic predictions looks like an appealing alternative to mitigate congestion. The increasing availability of detectors in the transportation infrastructure has allowed tackling the traffic prediction problem by using a purely data-driven approach, where deep learning models have excelled. Unfortunately, the implementation of these techniques in real time is still under development. This work presents the implementation of a real-time traffic prediction application in the Las Vegas, NV, USA, urban area, built as a cyberphysical system with real-time data streaming from field sensors to a cloud-like environment where deep learning-based traffic predictors are hosted. Implementation results show the feasibility of doing traffic prediction in real time with the current technology and the usefulness of periodic retraining to maintain prediction accuracy.