Programa del congreso

This paper presents the development of KARNAV, an autonomous mobile robot (AMR) prototype designed for indoor navigation with real-time obstacle detection and avoidance capabilities. The system leverages a ROS~2-based architecture to integrate LiDAR and inertial measurement unit (IMU) data, supporting advanced sensor fusion and SLAM-based (Simultaneous Localization and Mapping) localization. The design prioritizes the use of low-cost, accessible components without compromising operational reliability, demonstrating both technical and economic feasibility for industrial and academic applications. Experimental results validate the platform's ability to achieve robust navigation, energy efficiency, and adaptability within dynamic environments. This prototype provides a versatile foundation for further research in industrial robotics and the deployment of autonomous systems in complex indoor scenarios.

Volt/Var control is a fundamental strategy for maintaining power quality within predefined limits, reducing energy losses, and improving the operational efficiency of distribution feeders. However, the increasing penetration of distributed energy resources (DERs) and electric vehicles (EVs) has imposed significant challenges on traditional Volt/Var control schemes. In this regard, this work presents a PRISMA-aligned review of Volt/Var control strategies applied to distribution feeders with high DER/EV penetration. This review employs a six-stage workflow encompassing database searches Scopus and The Lens, corpus screening eligibility criteria and citation tracking, and evidence synthesize through temporal trends, VOSviewer-based keyword co-occurrence mapping, and a narrative assessment structured around seven research questions. The results reveal steady growth in the field since 2014, with a marked acceleration driven by rising DER penetration and new inverter grid-support requirements. Four main research fronts emerge: deterministic Volt/Var optimization using traditional devices; inverter-based reactive-power and power-quality control; PV-centric voltage regulation; and AI-enabled coordination strategies for EV charging management. In addition, the literature converges on hierarchical, multi-timescale control schemes that integrate OLTCs and capacitor banks with smart inverters, energy storage systems, and reactive-capable EV chargers, all supported by advanced computational and optimization tools.

This study presents the development of a platform for real-time simulation and optimization of power distribution networks. The proposed system integrates OPAL-RT hardware and the ePHASORSIM module as the simulation tool, enabling dynamic modelling of network behaviour, while a phasor-based framework facilitates real-time phase analysis under variable loading conditions. A high-speed fibre optic communication infrastructure operating under the DNP3 protocol ensures reliable synchronization and data transmission from two substations to the simulation environment. In addition, artificial intelligence techniques are incorporated through large language models and reinforcement learning algorithms to generate automated operational recommendations aimed at optimizing power flow and improving decision-making processes. Software-based analysis is employed for voltage profile assessment, contingency scenario simulation, and operational stability evaluation in low-inertia distribution networks. By simulating multiple operating conditions, the platform enhances early fault detection and improves overall system reliability, reducing the impact of external disturbances on network performance. The proposed methodology is validated through technical studies conducted using the CYMDIST power distribution simulation software and automated diagnostics performed through specialized Python scripts for signal processing and data evaluation. The obtained results accurately reproduce network dynamics and optimal operational strategies, demonstrating the feasibility and robustness of the proposed approach. Furthermore, an interactive dashboard and a geographic information system interface are integrated to centralize real-time data visualization, providing operators with geospatial insight required for comprehensive network supervision, preventive maintenance planning, and rapid emergency response. The platform establishes a framework for future integration of advanced control strategies, intelligent monitoring systems, and large-scale distribution network applications.

The increase in the share of renewable energy sources, driven by CO₂ emission reduction targets, lower generation costs, and energy policies, has introduced a significant level of uncertainty into the operation of electrical systems. This variability generates discrepancies between scheduled generation and dispatched generation, increasing the need for backup systems to ensure the security and stability of supply, which makes the operation of the electrical system vulnerable. This paper develops a cost optimization model for economic dispatch based on the gravitational search algorithm (GSA), which considers the uncertainty associated with a renewable generation source using Monte Carlo simulation. Thirty scenarios are generated over a 24-hour horizon, from which the minimum, average, and maximum costs per hour are obtained. A performance comparison of each algorithm is also carried out with the particle swarm optimization (PSO) method, widely used in the literature. The performance comparison shows that both methods produce high-quality solutions; however, the GSA obtains lower costs in most scenarios and presents less dispersion of results, demonstrating greater robustness and stability in the face of renewable uncertainty

Abstract– Reliable precipitation sampling is essential for stable isotope analysis in hydrology and climate studies, particularly in remote regions where frequent access to monitoring stations is limited. Under such conditions, accumulated sampling over several days is commonly adopted, increasing the risk of cross-contamination and reducing the temporal resolution of isotopic data. This work presents the design and experimental evaluation of an automated precipitation sampling prototype intended to explore the feasibility and limitations of reducing cross-contamination through sequential sample separation. The system integrates commercially available water-level sensors, electromechanical actuators, and a microcontroller-based control architecture, and is evaluated under controlled laboratory conditions. Experimental results show that the conductive water-level sensor provides stable and repeatable measurements once a measurable water level is established, with no significant dependence on water pH or conductivity within the tested range. However, a significant dead volume driven by container geometry and sensor placement introduces systematic detection delays at low flow rates. At the system level, partial automation is achieved, but fully autonomous operation is limited by actuator performance and fixed-threshold control logic. The main contribution of this work lies in the experimental identification of practical limitations and failure modes associated with affordable automated precipitation sampling systems, providing guidance for future designs aimed at improving reliability, autonomy, and suitability for isotopic precipitation monitoring in resource-constrained environments.

Epileptic seizure prediction remains a critical challenge in clinical neurology, particularly for patients with drug-resistant epilepsy. Recent advances in deep learning have improved predictive performance; however, the lack of interpretability and reliability limits their adoption in real-world healthcare settings. This paper proposes a trustworthy Internet of Medical Things (IoMT) framework for automated seizure prediction from multi-channel EEG signals, integrating explainable artificial intelligence techniques with a hybrid deep learning architecture. The proposed approach employs a CNN–BiLSTM model integrated with a channel-wise attention mechanism to enhance EEG preprocessing and feature extraction across various domains. SHapley Additive exPlanations (SHAP) and Gradient-weighted Class Activation Mapping (Grad-CAM) are used to show how the model makes decisions on both a global and a local level. The framework is evaluated using the benchmark CHB-MIT scalp EEG dataset through patient-wise cross-validation to mitigate data leakage. The average accuracy of the experiments was 94.7%, the sensitivity was 95.6%, and the AUC was 98.2%. Also, the calibration analysis shows a very small Expected Calibration Error of 0.018, which means the probability predictions are good. These results show that the new method does a great job of balancing accuracy, clarity, and trustworthiness. This makes it a good choice for helping doctors figure out when someone might have a seizure.