IEEE Latin America Transactions

A GUI for the synthesis and design of analog filters based on Pascal and other classical approximations

Fri, 12 Jun 2026 00:00:00 +0000

A Graphical User Interface (GUI) called SAFIMAM (acronym of Synthesis of Analog Filters in MATLAB by Approximation Methods) developed in MATLAB App Designer, an interactive development environment for designing an app layout and programming its behavior, capable of carrying out the synthesis of analog filters based on classical approximation methods as well as the Pascal approach, starting from a few design specifications, is presented. Quantitative metrics regarding computational efficiency such as algorithmic scalability, in the range of milliseconds, and synthesis runtime, in a hundred seconds, confirm that the program maintains a good performance as workloads grow without significant slowdowns. Besides, according to the results of the many different tests applied, SAFIMAM has proved to be competitive when compared to some other synthesis tools reported in the literature. Two design examples synthesized with SAFIMAM were implemented: a Pascal filter optimized in the stopband for Electromyogram (EMG) signals, implemented with AN231E04 Field Programmable Analog Arrays (FPAAs) embedded in the Anadigm QuadApex Development Board; and a Chebyshev Substrate Integrated Waveguide (SIW) filter for the fifth generation of wireless cellular technology (5G). Experimental and synthesis results agreement demonstrate the SAFIMAM reliability. In addition, when compared to some other EMG and 5G filters reported elsewhere, it is evident that the performance of the synthesized filter structures produced by the proposed software are also feasible.

Robust Stability Analysis of DC Microgrids

Fri, 12 Jun 2026 00:00:00 +0000

Direct current (DC) microgrids integrate renewable energy sources, energy storage systems, and electronic loads in order to improve efficiency and flexibility in energy management. However, their stability can be compromised by dynamic interactions among subsystems under different operating conditions, particularly in the presence of constant power loads (CPLs). This paper proposes a robust stability analysis based on the $\mu$-analysis framework. The main contribution is a systematic and reproducible $\mu$-analysis procedure that quantifies robust stability margins under simultaneous load uncertainties applied to DC microgrids. To demonstrate its applicability, a microgrid architecture composed of a high-order Quadratic Buck converter with reduced redundant power processing interacting with a conventional Boost converter is analyzed. Validation is carried out using switched models implemented in the PowerSim simulator under time-varying load conditions. In addition, Monte Carlo simulations are performed by sampling the load parameters to evaluate robustness under simultaneous load uncertainties. The results confirm the effectiveness of the proposed methodology for validating time-domain behavior and assessing robust stability in DC microgrids with CPLs.

An Area-Aware Figure of Merit for Improved State-of-the-Art Comparison of Analog-to-Digital Converters

Fri, 12 Jun 2026 00:00:00 +0000

This paper introduces a new Figure-of-Merit (FoM) for Analog-to-Digital Converters (ADCs) that integrates silicon area alongside the traditional metrics of resolution, bandwidth, and power dissipation. As technology scaling and system-on-chip integration make area efficiency a critical design constraint, the established Walden and Schreier FoM's provide an incomplete comparison of state-of-the-art performance. The proposed Area-Aware FoM (FoM$_{\text{A}}$) enables a more integral and equitable benchmarking by directly quantifying the trade-off between dynamic performance and physical implementation cost. The validity and utility of the FoM$_{\text{A}}$ are demonstrated through a re-evaluation of over 40 published ADCs, revealing significant shifts in architectural ranking and offering designers a more relevant metric for advanced technology nodes.

Synthetic Dataset Generation for Tomato Ripening Stage Detection in Different Scenes

Fri, 12 Jun 2026 00:00:00 +0000

The development of intelligent robotic systems for agriculture depends on large and representative datasets, which are essential for training computer vision models. However, the availability of public datasets in this area is limited, hindering the implementation and improvement of these technologies. To address this problem, we propose a methodology for synthetic dataset generation. This methodology includes the automated creation of datasets optimized through evolutionary algorithms, thereby improving the quality and diversity of the generated data. To validate the method, we tested it in a case study: the detection of tomato ripening stages in greenhouses. The experiments showed that training a detector (YOLOv5m model) with this synthetic data significantly improves its performance in real scenarios, increasing detection from null to acceptable performance. These results validate the effectiveness of synthetic data generation as a viable and affordable alternative to compensate for the shortage of agricultural datasets.

Hierarchical Attention-Based Convolutional Neural Network Model for Intrusion Detection

Fri, 12 Jun 2026 00:00:00 +0000

The increasing scale and complexity of internet-connected systems demand robust intrusion detection under realistic traffic conditions. This study presents H.A.L.C.CO.N (Hierarchical Attention-based Loss Equalization with CatBoost-enhanced Convolutional Neural Network), a multiclass intrusion detection model evaluated on the real-world LITNET-2020 dataset. The model integrates convolutional feature extraction, hierarchical attention, CatBoost-based encoding for high-cardinality categorical features, and Equalization Loss V2 (EQLv2) to address severe class imbalance. Experimental results show strong performance, achieving a detection rate of 99.997%, an F1-score of 99.997%, an accuracy of 99.996%, and a false positive rate of 0.0135%. These findings indicate that H.A.L.C.CO.N is an effective and practical solution for real-world multiclass intrusion detection.

MFPAD: Memory–Forgetting Planning for Long-Horizon End-to-End Autonomous Driving

Yikai Wu, Qizhou Hu, Aiguo Lei, Ziying Song (Author) — Fri, 12 Jun 2026 00:00:00 +0000

Recent planning-oriented end-to-end autonomous driving methods have achieved competitive performance on short-horizon (3 s) trajectory prediction. However, their accuracy and stability degrade markedly when the prediction horizon is extended to long-horizon (6 s), leading to drifting trajectories. A key reason is that most existing frameworks adopt simple feed-forward multilayer perceptron regressors in the trajectory refinement head, which lack explicit modeling of long-term temporal dependencies and planning inertia. To address this limitation, we propose the Memory–Forgetting Planning for Autonomous Driving, a plug-in refinement head that combines an LSTM-based memory network with a Transformer-based forgetting network. The memory network autoregressively rolls out a coarse long-horizon trajectory and exposes a sequence of hidden states, while the forgetting network attends over these states and surrounding-agent features with token-level dropout to suppress outdated or noisy motion cues. A lightweight gating module fuses coarse and corrected trajectories at each time step, yielding temporally consistent, interaction-aware plans over a 6 s horizon. We evaluate our method on NuScenes, Adv-NuScenes, Bench2Drive, and NAVSIM, and the results demonstrate consistent improvements. Compared with baselines, it reduces the average collision rate on nuScenes by 11.1% and the 3 s collision rate on Adv-NuScenes by 29.2%, while achieving a 6 s collision rate of 2.01% on NuScenes. In addition, the closed-loop results on Bench2Drive and NAVSIM show that the proposed refinement head also improves downstream driving performance under feedback-driven evaluation. The source code is available at https://github.com/Y1Ka1/MFPAD

ImmersiveSHAP: Immersive analytics visualization system for XAI using SHAP scatter plot

Fri, 12 Jun 2026 00:00:00 +0000

This study presents ImmersiveSHAP, an immersive analytics visualization system for explainable artificial intelligence that leverages a modular pipeline to render SHAP scatter dependence plots in virtual reality. The research contribution is the integration of explainable artificial intelligence (SHAP Python library) with immersive analytics in a virtual reality environment (Unity). The system integrates a Python-based preprocessing module that handles data loading, model training, and SHAP explanation computation, and a Unity-based rendering and interaction module, implemented within a client–server architecture. The WebSocket protocol establishes communication between the Python server and the Unity client. The system extends traditional 2D SHAP plots into interactive 3D visualizations, designed to support immersive analytics with post hoc model explanations. Furthermore, a technical validation used the Iris, Breast Cancer, and California Housing datasets, covering point clouds from N =150 to N =20,640, and deployed the system on a Meta Quest 3. Results identify operational constraints, showing stable performance on small-to-medium datasets (N ≤ 2,000) with an average frame rate of approximately 70 FPS, close to the device’s refresh rate target and within acceptable ranges for virtual reality. These results indicate the system’s viability as a baseline architecture for immersive visualization of SHAP-based explanations.

Optimized implementation methodology for HIL FPGA simulations for power converters

Fri, 12 Jun 2026 00:00:00 +0000

The growing complexity of power electronic systems demands simulation methods capable of high computational speed and accuracy. Traditional software‑based simulations exhibit long execution times and limited ability to reproduce critical operating conditions, while hardware experimentation may expose physical components to extreme scenarios that can compromise their integrity. This work presents an optimized methodology for modeling, discretizing, and implementing Boost, Buck, and Boost‑Buck converters on FPGA using LabVIEW. Mathematical models are derived from Kirchhoff’s laws and discretized through Euler’s method, followed by an algebraic conditioning stage that minimizes arithmetic operations and reduces clock‑cycle latency. Open‑loop tests comparing FPGA execution with PSIM simulations at 25% and 75% duty cycles demonstrate high fidelity in capacitor‑voltage and inductor‑current responses. Error growth at higher duty cycles is attributed to the UQ10.32 fixed‑point format. Overall, the proposed methodology offers an efficient and accurate alternative for real‑time‑oriented converter simulation, supporting safer and faster validation

FPGA-Based Control of an Extendable Bidirectional DC-DC Converter for EV Application

Kumaravel, Anjana, Seshagiri Rao (Author) — Fri, 12 Jun 2026 00:00:00 +0000

This paper proposes an extendable bidirectional DC–DC converter (E-BDC) for medium- and high-voltage DC applications requiring high voltage gain, reduced device stress, and bidirectional power flow. The topology ensures that each switch conducts only one inductor current per n-stage implementation, reducing conduction losses and improving component utilization. The converter inherently provides voltage self-balancing across switches and maintains continuous low-voltage port current, minimizing ripple and making it well suited for battery-integrated systems. Steady-state analysis is carried out under synchronous and phase-shifted switching schemes, where phase-shifted operation significantly reduces capacitor voltage ripple and capacitance requirements. A comprehensive small-signal model including parasitic elements is developed for both operating modes. The control-to-output transfer function reveals non-minimum phase behavior in step-up mode, which is addressed in the PI controller design. A 700-W, 800-V prototype operating at 50 kHz and implemented on a Zynq-7000 FPGA validates the analysis, demonstrating high gain, voltage self-balancing, reduced stress, and stable dynamic performance.

Impact of inertial control on battery energy storage requirements in wind-integrated power systems

Brian Loza, Luis Ismael Minchala (Author) — Fri, 12 Jun 2026 00:00:00 +0000

The increasing penetration of wind energy in modern power systems has intensified the need for advanced frequency regulation strategies capable of preserving grid stability under low-inertia conditions. In this context, inertial control of wind turbines, complemented by energy storage systems (ESS), represents a promising approach to enhance frequency response during contingency events. This paper investigates, through detailed time-domain simulations, the impact of advanced inertial control strategies on the energy storage requirements associated with frequency regulation tasks. Two control schemes are analyzed: (i) an extended optimized power point tracking (OPPTE) method and (ii) an adaptive OPPTE strategy enhanced with a fuzzy logic controller (OPPTE-FLC). Both approaches are coordinated with a battery energy storage system (BESS) equipped with state-of-charge feedback control (SOC-FB) to ensure operational sustainability. Simulation results demonstrate that the coordinated OPPTE-FLC and BESS framework significantly enhances dynamic performance during frequency disturbances. In particular, the proposed strategy increases the peak active power injection during contingencies by approximately 15\% and yields a notable improvement in frequency nadir compared to the conventional OPPTE-based approach.

Table of Contents August 2026

Daniel Ulises Campos Delgado — Fri, 12 Jun 2026 00:00:00 +0000