Alejandro Aragón-Zavala

Achieving the multi-layered spectral objectives of sixth-generation (6G) networks requires high-fidelity channel characterization, particularly for C-band deployments in complex vegetated environments supporting cellular vehicle-to-everything (C-V2X) and non-terrestrial networks (NTN). This study presents a physics-aware, heteroscedasticity-driven channel modeling framework that utilizes Variational Gaussian Processes (VGP) to quantify both aleatory and epistemic uncertainties, thereby characterizing the non-stationary dynamics of foliage-induced shadowing. By employing an Exponentiated Quadratic (EQ) kernel, the proposed stochastic architecture captures spatial correlation from real-world measurements at 4.9, 5.85, and 6.1 GHz. The results demonstrate a significant enhancement in predictive accuracy, achieving average Root Mean Square Error (RMSE) reductions of 73%, 76%, 82%, and 75% over traditional Floating-Intercept (FI), Close-In (CI), ABG, and CIF empirical benchmarks, respectively. These improvements are further validated through coverage probability and spread ratio metrics. Our findings establish variance-adaptive probabilistic regression as a robust alternative to the homoscedastic linear assumption, advancing 6G propagation modeling in high-entropy, vegetated scenarios.

Deterministic chaos in optically injected semiconductor lasers (OILs) has attracted significant attention due to its relevance in secure communications, entropy generation, and photonic applications. However, existing studies often rely on low-resolution parameter sweeps or include noise contributions that obscure the intrinsic nonlinear dynamics. To address this gap, we investigate a noise-free OIL model and construct high-resolution chaos maps across the injection strength and frequency detuning parameter space. Chaos is characterized using two complementary approaches for computing the largest Lyapunov exponent: the Rosenstein time-series method and the exact variational method. This dual approach provides reliable and reproducible detection of deterministic chaotic regimes and reveals a rich attractor landscape with alternating bands of periodicity, quasi-periodicity, and chaos. The novelty of this work lies in combining high-resolution mapping with rigorous chaos indicators, enabling fine-grained identification of dynamical transitions. The results not only deepen the fundamental understanding of nonlinear laser dynamics but also provide actionable guidelines for exploiting or avoiding chaos in photonic devices, with potential applications in random chaos-based communications, number generation, and optical security systems.

With the advancement of 5G and emerging wireless communication technologies, accurate modeling of wave propagation in indoor environments has become increasingly crucial. This study focuses on demonstrating how machine learning (ML) techniques can be applied to predict path loss within the millimeter wave (mmWave) spectrum in a specific indoor environment. We address high-frequency challenges like path loss and complex building layouts that impact signal propagation. We employ various ML models, including Artificial Neural Networks (ANNs), hybrid models integrating linear regression, ANNs, and Gaussian Processes, and Extreme Gradient Boosting (XGBoost), to predict and analyze the propagation loss in a controlled indoor setting. The models were trained and validated using data collected from a comprehensive measurement campaign at 28 GHz, which involved high precision radio equipment in a complex indoor environment. Our results demonstrate that while traditional models provide a baseline for understanding path loss, advanced ML models, particularly hybrid approaches, significantly enhance prediction accuracy and provide a deeper understanding of indoor propagation dynamics within this specific environment. The study highlights the potential of ML in overcoming the limitations of empirical models and showcases methodologies that can be adapted for similar indoor scenarios. This research advances our understanding of mmWave propagation indoors and sets a framework for utilizing ML in telecommunication system design and optimization in specific environments.

The Optimal Power Flow (OPF) problem, used to obtain efficient operation conditions with the lowest cost or the minimum power loss in electrical power systems, is a non-polynomial problem that becomes even harder to analyze when considering renewable energy sources (RES) with uncertain behavior. Therefore, establishing a manageable number of RES scenarios in the modeling is essential for optimizing cost-effective solutions, including those with constraints such as the valve-point effect and prohibited operational zones. This work compares three differential evolution algorithm (DEA) variants and four well-known metaheuristics: the Particle Swarm Optimization (PSO), the Bio-geographical Based Optimization (BBO), the Artificial Bee Colony Optimization (ABC), and the Non-dominated Sorting Genetic Algorithm II (NSGA-II). The metaheuristics are compared: 1) to determine the one with the best performance considering RES; 2) to establish an approach to minimize and find the best set of scenarios representing variable RES; 3) to compare the success rate of convergence of the penalized function against the real objective function. Results show that BBO and PSO optimization are the best choices for solving the classic objective function of OPF. On the other hand, the DE/best/1 (DEAB) algorithm demonstrates the best performance when the valve-point effect with prohibited zones is considered. DEAB presents the largest weighted cumulative rating (WCR) and the second-best weighted cumulative successful rate (WCSR) for all the evaluated criteria.

This paper presents an in-depth investigation into path loss modeling for wireless communication systems, emphasizing the critical role of Exploratory Data Analysis (EDA) and accurate field data collection. It reveals the enhanced efficacy of theoretical models when synergized with specific, real-world data, highlighting the necessity for empirical grounding. Central to our research is the demonstration of the indispensability of thorough EDA in path loss modeling. This phase uncovers key trends and anomalies, setting a solid foundation for subsequent Machine Learning (ML)-based predictive modeling. Deep dataset insights allow for the effective tuning of algorithms to reflect the complexities of signal propagation environments accurately. The study further underscores the impact of the quality of field data collection on model accuracy. Meticulous data gathering that accurately mirrors environmental conditions is essential, providing insights into signal behavior and validating theoretical models. Our research also explores the importance of spatial pattern analysis in understanding path loss. Advanced visualization and clustering methods are employed to illustrate how spatial variability significantly influences path loss, shedding light on the effects of environmental factors like obstructions and topography on signal propagation.