Instituto de Investigación Aplicada

Permanent URI for this collectionhttps://repositorio.ucb.edu.bo/handle/20.500.12771/1426

Browse

Now showing 1 - 19 of 19

A Comparative Analysis of Empirical and Machine Learning Models for LoRa Path Loss Prediction at 915 MHz in an NLOS Urban Canyon
(CRC Press, 2026) Esprella Quiroga, Horacio Alessandro; Molina Silva, Marcelo
This study presents an evaluation of path loss behavior in a highly complex non-line-of-sight (NLOS) urban environment, focusing on the Obrajes district of La Paz, Bolivia, an urban area characterized by irregular topographic variation and densely built infrastructure. The analysis was conducted using LoRa technology operating at 915 MHz, with measurement campaigns carried out across three parallel avenues, each located at a distinct altitude and exhibiting descending elevation profiles along their paths. To assess propagation prediction accuracy in this challenging scenario, two well-established theoretical models were considered: Okumura–Hata, designed for urban environments, and the 3GPP UMi model, specific to NLOS urban microcells. The root mean square error (RMSE) was employed as the evaluation metric. Additionally, to improve prediction performance in such a terrain-constrained environment, two machine learning models, random forest (RF) and artificial neural networks (ANN), were trained using the field-collected experimental data. These data-driven models demonstrated superior adaptability and accuracy when compared to the classical theoretical approaches. The work highlights the importance of integrating detailed geographic and morphological characteristics into wireless propagation analysis and supports the integration of machine learning as a robust and scalable alternative for signal prediction in irregular urban topographies. The findings are especially relevant for future wireless planning in cities with complex terrain conditions across diverse global contexts
Advanced Path Loss Estimation at 5.8 GHz in Heterogeneous Urban Environments through Hybrid Modeling and Machine Learning
(CRC Press, 2026) Miranda Saravia, Alejandro Rommel; Miranda Saravia, Leoni Marti; Molina Silva, Marcelo; Gonsioroski, Leonardo H.
Accurate path loss estimation is a fundamental pillar for the design and optimization of wireless networks, especially in urban environments where signal propagation is significantly affected by complex and variable factors such as buildings, vegetation, and other obstacles. This comprehensive study evaluates the performance of a diverse set of path loss models by comparing their predictions with empirical data collected through extensive measurement campaigns conducted on the campus of the Pontifical Catholic University of Rio de Janeiro (PUCRio) at a central frequency of 5.8 GHz. The PUCRio campus, characterized by significant vegetation and six-story buildings, provides a realistic and challenging environment to assess the accuracy of these models under urban propagation conditions. The analysis covers a wide range of traditional deterministic models, including the free space path loss (FSPL) model, which assumes an unobstructed line of sight (LoS); the log-distance path loss model (LDPM), which considers distance-dependent attenuation; and the log-distance path loss model with shadowing (LDPM-S), evaluated both with fixed parameters (path loss exponent n = 3 and shadowing standard deviation σ = 6) and with empirically derived parameters adapted to the specific environment. In addition, the study includes multiple iterations of the two-ray model (DR), which accounts for both direct and ground-reflected paths; the DR with critical distances (DR-DC), introducing distance thresholds to refine predictions; and the sinusoidal DR, which incorporates sinusoidal variations to better capture environmental effects. Other traditional models evaluated include the original WINNER model (WNM), designed for urban scenarios; the modified WINNER model (M-WNM), adapted to specific conditions; the ITU-R P. 1411 model, addressing short-range outdoor propagation; and the ITU-R P. 1411 model with obstacles, which considers additional attenuation caused by physical barriers. Beyond these deterministic approaches, an advanced machine learning (ML) technique based on the adaptive moment estimation (ADAM) optimization algorithm was implemented to model path loss. This data-driven approach leverages adaptive learning capabilities to capture complex propagation patterns from empirical data, offering a more flexible and dynamic alternative to traditional models. The ADAM algorithm, known for its efficiency in optimizing neural networks, was applied to predict path loss with high accuracy, taking into account the complex interaction of environmental factors present on the PUCRio campus. The results of the comparative analysis reveal that the ML model based on ADAM consistently outperforms all traditional deterministic models in terms of accuracy. This superior performance is attributed to the algorithm’s ability to capture nonlinear relationships and adapt to the highly variable propagation conditions of urban environments—something that traditional models often fail to represent accurately due to their reliance on simplified assumptions. For instance, while the FSPL model provides a baseline for unobstructed environments, it does not account for obstructions or multipath effects prevalent in urban areas. Similarly, models such as LDPM-S and the two-ray variants improve upon FSPL by incorporating shadowing and ground reflections but still fall short of modeling the full complexity of real-world scenarios. The WINNER and ITU-R P. 1411 models, although designed for urban contexts, are limited by their dependence on predefined parameters that may not fully capture the site-specific characteristics. In contrast, the data-driven nature of the ADAM algorithm allows it to dynamically adjust to the unique propagation characteristics of the PUCRio campus, resulting in a significantly improved path loss estimation. These findings highlight the transformative potential of integrating advanced ML techniques, such as the ADAM algorithm, into the design and optimization of wireless networks. By achieving greater accuracy in path loss estimation, the ML approach enables more reliable network planning, better signal coverage, and enhanced communication performance in challenging urban environments. This study underscores the limitations of traditional deterministic models and advocates for the adoption of data-driven ML methods to address the complexities of modern wireless communication systems, paving the way for more robust and efficient network deployments.
Analysis of Path Loss in Narrowband for Millimeter-Wave Measurements Over a Frequency Range of 27 GHz to 29 GHz at PUC-Rio Campus, Brazil
(Institute of Electrical and Electronics Engineers Inc., 2026-02) Segaline Valle, Manuel Ernesto; Paredes Tamayo, Fátima Alejandra; Orihuela Carlos; Molina Silva, Marcelo; da Silva, Luiz; Verdugo, Elizabeth
This paper aims to evaluate the accuracy of Path Loss mathematical models using measurements taken at the Pontifícia Universidade Católica do Rio de Janeiro - PUC-Rio by comparing them with theoretical models such as the CI (Close-In) model and the Alpha-Beta model. Outdoor measurements were conducted with the transmitter installed on the Leme building at the PUC-Rio campus in Brazil. In total, 23 reception points were evaluated for 27−29GHz frequencies. Afterward, Path Loss was plotted as a function of distance for both theoretical models and empirical data. Finally, Root Mean Square Error (RMSE) was calculated to assess the efficiency of theoretical models regarding empirical measures.
Artificial Intelligence Enabled Radio Propagation: Path Loss Improvement and Channel Characterization in Vegetated Environments
(Brazilian Microwave and Optoelectronics Society (SBMO) / Brazilian Society of Electromagnetism (SBMag), 2024-03) Gonsioroski, Leonardo; Santos, Amanda; Viana, Jairon; Ferreira, Sandra; Silva, Rogerio; Da Silva Mello, Luiz; Matos, Leni; Molina Silva, Marcelo
In this paper, the application of AI and machine learning (ML) to the study of wireless propagation channels is investigated in two parts: first, an artificial neural network model is used to improve path loss prediction, and then, a pattern recognition model using multilayer perceptron (MLP) networks is used to identify and remove impulsive noise in power delay profiles (PDP). These studies were conducted based on field measurements in the 2400 MHz band in a public square with vegetation. The results are analyzed and compared with ordinary least squares (OLS) nonlinear regression results and results from similar studies. The Root Mean Square Error (RMSE) values between the experimental results of mean path loss and those provided by each propagation model are presented. The adjustment performed by OLS nonlinear regression and ANN significantly reduced the RMSE. The best results are those presented by artificial neural networks, with RMSE of 0.39 when using four neurons in the hidden layer of the ANN. The ANN used to identify and remove impulsive noise in power delay profiles (PDP) through pattern recognition proved to be more efficient than the CFAR technique. ANN technique found a larger number of valid multipaths compared to the CFAR technique
Caracterização do Canal de Propagação V2V em 5,8 GHz para Ambientes Urbano, Suburbano e Rural
(Sociedade Brasileira de Micro-ondas e Optoeletrônica (SBMO) / Sociedade Brasileira de Eletromagnetismo (SBMag), 2018) Molina Silva, Marcelo; Orihuela Vargas, Carlos; Abularach Arnez, Jussif Junior; Reis da Silva Mello, Luiz A.; Rodríguez R., Carlos V.; Gonzalez Castellanos, Pedro; Gonsioroski, Leonardo Henrique
Este artigo apresenta resultados preliminares de medidas do retardo médio e retardo RMS para o canal de comunicações veículo-a-veículo (V2V) em ambientes urbano, suburbano e rural na cidade do Rio de Janeiro Brasil. A sondagem do canal foi realizada na frequência de 5.8 GHz através da transmissão de um sinal OFDM com largura de banda de 20 MHz. Foram obtidas também as distribuições cumulativas de probabilidade do retardo RMS, do número e amplitude das componentes de multipercurso.
Comparing Exponential Decay Models for Path Loss Prediction at 5.8 GHz Using Machine Learning Techniques
(Institute of Electrical and Electronics Engineers Inc., 2024-12) Miranda Saravia, Leoni Marti; Rommel Miranda Saravia, Alejandro; Molina Silva, Marcelo; Gonsioroski, Leonardo H.
Accurate path loss estimation is crucial for optimizing wireless networks, particularly in urban environments where signal propagation is complicated by factors such as buildings and vegetation. This study evaluates the performance of traditional path loss models against an AI-based approach utilizing the ADAM optimization algorithm. Data collected at 5.8 GHz from the Pontificia Universidade Católica do Rio de Janeiro (PUCRio) campus, characterized by significant vegetation and six-story buildings. These findings underscore the potential of integrating advanced techniques into the design and optimization of wireless networks to enhance accuracy and communication performance
Comparing WINNER Path Loss Models with Machine Learning at 5.8 GHz
(Institute of Electrical and Electronics Engineers Inc., 2024-12-01) Miranda Saravia, Alejandro Rommel; Miranda Saravia, Leoni Marti; Molina Silva, Marcelo; Rodríguez Ron, Carlos V.
In this study, the accuracy of various path loss models was evaluated for wireless networks, which are crucial for urban environments with variable propagation conditions. Traditional models such as the Free Space Path Loss (FSPL) model, the modified WINNER model, and the original WINNER model were compared with an advanced machine learning approach using the ADAM algorithm. Data were collected on the campus of the Pontifical Catholic University of Rio de Janeiro (PUCRio) at 5.8 GHz. The machine learning model demonstrated higher accuracy, surpassing the performance of traditional deterministic models.
Evaluation of Cooperative Cognitive Radio System for White Spectral Space Detection using the Covariance Detector
(Institute of Electrical and Electronics Engineers Inc., 2025-09) Ramirez Montecinos, Rafael G.; Luna Rico, Mateo I.; Molina Silva, Marcelo; Abularach Arnez, Jussif J.; Silva Mello, Luiz Da; Rodríguez R, Carlos V.; Gonsioroski, Leonardo H.
Wireless spectrum is increasingly scarce, which motivates the need for robust methods to detect unused bands—especially under challenging conditions like low SNR and fading. This study proposes integrating Spectral Covariance Sensing (SCS) into a cooperative cognitive radio framework, leveraging hard-decision fusion schemes (AND, OR, Majority) to enhance detection stability. Using real Advanced Television Systems Committee (ATSC) signal data, the detection performance was evaluated across various SNR levels. The results show that cooperative sensing significantly improves detection probability under low SNR, with the OR rule achieving the highest detection rate (e.g., ≈90% at -30 dB) and the majority rule providing the best overall trade-off between reliability and false alarms. These findings demonstrate the practical value of cooperative SCS systems in dynamic spectrum environments
Evaluation of OFDM Channel Sounding Techniques with Three Modulation Sequences
(Brazilian Microwave and Optoelectronics Society (SBMO) / Brazilian Society of Electromagnetism (SBMag), 2019-12) Molina Silva, Marcelo; Mello, Luiz da Silva; Ron, Carlos V. Rodríguez; Almeida, M. P. C.; Tacuri, Alexander B. Hilario; Castellanos, Pedro Gonzalez; Gonsioroski, L. H.
This paper presents an evaluation of multicarrier channel sounding techniques using different random and pseudo-random sequences to modulate the OFDM sounding signal. The Random (Rand), Pseudo-Noise (PN) and Zadoff-Chu (ZC) were tested, both in laboratory simulations and in field measurements. For the laboratory simulations Matlab routines were used to generate OFDM signals modulated with each of the three sounding signals, that were then convoluted with a synthesized transfer function of a test channel with six multipath components and added Gaussian noise and Doppler fading. The resulting signals are then correlated with a copy of the original signal to provide the multipath power delay profiles. The root mean square deviation (RMSD) and the relationship between peak power and mean power (PAPR) were used as metrics for the comparison between the test transfer function and the simulation detected multipath delay profiles, showing slight advantages of the ZC sequence. The three sequences were then used in field measurements to characterize an urban channel at 700 MHz. In the field measurements, the ZC sequence showed the lowest detection threshold, allowing for the detection of a larger number of multipath components
Evaluation of Propagation Models and Machine Learning Through Path Loss Analysis in Different Environments Using LoRa Technology at 915 MHz in The Department of La Paz, Bolivia
(Institute of Electrical and Electronics Engineers Inc., 2025-09) Esprella Quiroga, Horacio; Molina Silva, Marcelo; Condori Quispe, Hugo; Gonsioroski, Leonardo
The performance of Low-Power, Wide-Area Networks (LPWANs) is critical for long-range communication systems, but signal propagation in complex topographies limits their reliability. Theoretical propagation models often provide inaccurate path loss predictions in such environments, leading to suboptimal network design. This work presents an empirical evaluation of LoRa technology at 915 MHz in La Paz, Bolivia, and compares the accuracy of classical propagation models against data-driven machine learning (ML) techniques. A dataset was constructed from a field measurement campaign conducted in urban (NLOS), suburban (mixed LOS/NLOS), and rural (LOS) scenarios. The Okumura-Hata, 3GPP UMi, and COST231-WI models were evaluated. Among these, the 3GPP UMi model demonstrated the lowest error, yet yielded a Root Mean Square Error (RMSE) of 11.58 dB against measured data. In contrast, ML models trained on the same data showed higher accuracy. A Random Forest model achieved an RMSE of 3.49 dB, and an Artificial Neural Network model achieved an RMSE of 4.57 dB. The results indicate that ML models provide more accurate path loss predictions than theoretical models in complex terrains, offering an effective tool for planning LoRa-based communication networks
Evaluation of White Space Detection Algorithms for Cognitive Radio Applications
(Institute of Electrical and Electronics Engineers Inc., 2013) Molina Silva, Marcelo; Reis da Silva Mello, Luiz A.; Rodríguez Ron, Carlos V.
This article presents results of evaluations and comparisons of two detection algorithms for cognitive radio, the energy detection (ED) and spectral covariance sensing (SCS), using experimental data obtained in field measurement campaigns. The algorithms were tested to evaluate their performance in terms of probability of false alarm, probability of detection and probability of misdetection. The experimental data consists of received signal measurements at geo-positioned points along different routes in an urban area
Initial Results of Millimeter Wave Outdoor Propagation Measurements in a Campus Environment
(Institute of Electrical and Electronics Engineers Inc., 2018-09) Orihuela Vargas, Carlos Eduardo; Molina Silva, Marcelo; Abularach Arnez, Jussif Junior; Da Silva Mello, Luiz
This paper presents results of outdoor measurements for a range of frequencies from 27 GHz to 40 GHz with steps of 1 GHz. Measurements were performed in PUC-RIO campus. Using a signal generator, a CW at 0 dBm was transmitted. Received power was measured with a spectrum analyzer which allows a maximum path loss measurement of 140 dB. Two 20 dBi horn antennas were used, one at transmitter and one at receiver. The transmitter was positioned at Leme Building's roof with an elevation of 50 meters from ground. 23 receiver points were selected for measurements and received power level at each point and each frequency was obtained. Results indicates that foliage caused weak signals in some points. As expected, lower frequencies have better coverage range
Integration of AI and Deterministic Models for Accurate Path Loss Prediction at 5.8 GHz
(Institute of Electrical and Electronics Engineers Inc., 2026-02) Miranda Saravia, Alejandro Rommel; Molina Silva, Marcelo; da Silva Mello, Luis; Verdugo Elizabeth; Miranda Saravia, Leoni Marti
This paper evaluates classical deterministic path loss models and an Artificial Intelligence (AI)-based approach using empirical measurements collected at 5.8 GHz on the PUCRio campus. Free Space Path Loss (FSPL), Two-Ray models and their variants are compared against a neural network optimized with the ADAM algorithm. Results show that the AI model significantly outperforms deterministic approaches, achieving an R2 of 0.98 and an RMSE of 1.20 dB, demonstrating its effectiveness for accurate path loss prediction in complex environments.
Modeling Propagation Loss at 5.8 GHz: A Comparison of Log-Distance Models and Machine Learning with Empirical Measurements
(Institute of Electrical and Electronics Engineers Inc., 2024) Miranda Saravia, Leoni Marti; Miranda Saravia, Alejandro Rommel; Molina Silva, Marcelo; Gonsioroski, Leonardo H.
This study evaluates several propagation loss models by comparing them with empirical data collected from measurement campaigns at the Pontifical Catholic University of Rio de Janeiro (PUCRio) at a central frequency of 5.8 GHz. The analysis encompasses five different models: the Free Space Path Loss (FSPL) model, the Log-Distance Path Loss Model (LDPM), the Log-Distance Path Loss Model with Shadowing (LDPM-S) using fixed parameters (n=3, σ=6), LDPM-S with empirically obtained parameters, and the ADAM algorithm for machine learning. The findings reveal that the ADAM algorithm achieved significantly higher accuracy compared to the traditional deterministic models, underscoring its potential to improve path loss estimation performance
Optimized Mammography Preprocessing for Tumor Detection with YOLO11-Seg in Young Women from the Bolivian Altiplano
(Institute of Electrical and Electronics Engineers Inc., 2025-09) Miranda Saravia, Leoni Marti; Miranda Saravia, AlejandrRommel; Garrido Vargas, Alexandra Jahel; Molina Silva, Marcelo; Conde Canaviri, Yancarla Mary; Mendez Conde, Mauricio Abdel; Leyva Lamas, Leonardo
Early diagnosis of breast cancer in young women presents a critical clinical challenge, particularly in geographic contexts such as the Bolivian Altiplano, where high breast density and limited access to specialized technologies hinder detection. This study evaluates the impact of various image preprocessing techniques on the performance of an automatic detection model based on YOLO11-seg. Using a dataset of mammograms annotated by certified radiologists, transformations such as CLAHE, histogram equalization, Canny filtering, wavelets, and anisotropic diffusion were applied. Standard metrics (mAP, precision, recall) were measured and results were compared in a real clinical setting. Findings show that CLAHE significantly improves the model's ability to detect lesions in dense breasts, achieving a mAP of 71.8%. The results suggest that combining enhancement techniques with AI models can strengthen early detection in high-risk populations, offering a viable and scalable alternative for resource-limited settings
Prediction of Path Loss in Real-World Environments: Comparison of ITU-R P.1411 Models and Deep Learning
(Institute of Electrical and Electronics Engineers Inc., 2024-12) Miranda Saravia, Leoni Marti; Miranda Saravia, Alejandro Rommel; Molina Silva, Marcelo; Gonzalez Castellanos, Pedro Vladimir
Accurate path loss estimation is essential for the design and optimization of wireless networks, especially in urban environments where propagation conditions can be extremely variable. This study compares traditional models such as the Free Space Path Loss (FSPL) model, ITU-R P.1411 model, and ITU-R P.1411 model with obstacles with an advanced machine learning approach using the ADAM algorithm. Empirical data were collected on the campus of the Pontifical Catholic University of Rio de Janeiro (PUCRio) at a frequency of 5.8 GHz. The comparison between the methods revealed that the machine learning approach offers greater accuracy, surpassing traditional deterministic models in terms of path loss estimation accuracy
Preliminary Results of the Indoor Coverage Field Tests of the Advanced ISDB-T System in Brazil
(Institute of Electrical and Electronics Engineer, 2023) Santos, Amanda B.; Gonsioroski, Leonardo H.; Oliveira, Rodrigo; Da Silva Mello, Luiz; Botelho, Alberto; Akamine, Cristiano; Fernandes, Natália C.; Molina Silva, Marcelo
This article presents the preliminary results of field tests with the Advanced Integrated Services Digital Broadcasting for Terrestrial Television Broadcasting (ISDB-T) system in indoor environments. Received channel power and error-free system reception threshold records were performed at four different measurement sites. Five different H/V dual-polarized MIMO antennas are used in measurements to evaluate system performance when using antennas from different manufacturers. The results show that the advanced ISDB-T system is robust to indoor penetration losses and when subjected to the use of different antennas
Urban Mobile Channel Delay Spread Measurements at 700 MHz, 2.5 GHz and 3.5 GHz
(Institution of Engineering and Technology (IET), 2018-04) Molina Silva, Marcelo; Da Silva Mello, Luiz; Rodríguez R., Carlos V.; Almeida, M. P. C.; Gonzalez Castellanos, Pedro
This paper presents results of delay spread measurements at 700 MHz, 2.5 GHz and 3.5 GHz in an urban environment in Rio de Janeiro, Brazil. The measurements were made in the same routes for all three bands, allowing the comparison of channel parameters and the investigation of their dependency on frequency. Results include the average delay, the RMS delay spread, and its cumulative probability distribution
Wideband Channel Sounding Using Modulated OFDM Signals
(Institute of Electrical and Electronics Engineers Inc., 2018) Molina Silva, Marcelo; Mello, Luiz da Silva; Rodriguez R., Carlos V.; Castellanos, Pedro Gonzalez
This work presents a comparison of the results of measurements of average delay and RMS delay spread in an urban environment using different sequences of information for an OFDM channel sounding signal. The sequences evaluated are a random sequence, a Pseudo-Noise sequence (PN) and the Zadoff-Chu sequence (ZC). The measurements were carried out in urban areas of Rio de Janeiro - Brazil, with a 20 MHz bandwidth signal with a central frequency of 773 MHz. The results are influenced by the autocorrelation properties of each sequence. In order to obtain compare the values of average delay and RMS delay spread obtained with each sequence, the measurements were performed under the same conditions.