Computing and Telecommunications Research Laboratory (LCT), Postgraduate Program in Electrical Engineering, Federal University of Par, Belm, Pará, Brazil
Auto-generated profile, based on publicly available information
Caio Mateus Machado Cardoso
Computing and Telecommunications Research Laboratory (LCT), Postgraduate Program in Electrical Engineering, Federal University of Par, Belm, Pará, Brazil
The Amazon region wide forest extension and its limited infrastructure make it difficult to access essential services. In this context, efficient and reliable communication could reduce distances and facilitate access to these services. Considering infrastructural limitations, Low-Power Wide Area Networks (LPWAN), such as Long-Range (LoRa) networks, show promise for implementation in the region. However, it is important to understand how trees affect the propagation of electromagnetic waves to properly plan and install telecommunication services. Thus, this article aims to determine the attenuation coefficient for wooded environments in an Amazon region through a neural model. For this purpose, data were collected in a grove located at the Federal University of Pará (UFPA). The data were analyzed, and the impact of spreading factor and antenna polarizations on LoRa signal propagation was verified, with Vertical-Vertical (VV) polarization being the least affected by attenuation on average. Finally, the proposed neural model was compared with an adjusted version of the modified exponential decay model and with traditional models such as Weissberger, COST-235, ITU-R, FITU-R, ITU-R P.833-10, and Chen & Kuo. The neural model demonstrated its effectiveness, outperforming the others in all scenarios with an RMSE below 6.5 dB.
Reinforcement learning for uncrewed aerial vehicle (UAV)-assisted long range (LoRa) networks currently relies on model-free approaches utilizing generic propagation models. However, these studies leave model-based methods unexplored and frequently omit vegetation effects, which are essential for accurate forest deployments. This work compares the proximal policy optimization (PPO) (model-free) and DreamerV3 (model-based) algorithms for optimizing UAV-LoRa networks with vegetation-aware propagation. This study formulates the network optimization as a centralized control problem that integrates empirically-validated attenuation coefficients from Amazonian forest measurements, dynamic UAV positioning, and realistic payload constraints. Simulation results reveal fundamental tradeoffs: DreamerV3 achieves approximately 1.6× sample efficiency (150k vs 250k steps), reducing costly UAV testing, while PPO delivers 4× energy efficiency (4 · 107 vs 1 · 107 bits/Joule), and exhibits lower variance. DreamerV3’s higher coverage advantage comes at substantial energy cost. The results demonstrate the importance of vegetation-aware systems and provide deployment guidance for UAV-assisted networks in dense forest environments.
1 January 2026 in Journal of Microwaves, Optoelectronics and Electromagnetic Applications
The present research proposes a methodology to develop a model to predict Path Loss in a mixed environment comprising urban areas with tree-lined streets. Based on cross-validation and utilizing Artificial Intelligence (AI) and Machine Learning (ML) algorithms that employ various training functions, architectures, and numbers of neurons, the goal is to optimize the model’s performance in predicting Path Loss (PL) in mixed environments. The model is structured in layers, incorporating geolocation, non-line-of-sight loss, urban density, and vegetation coverage to predict PL. Within this model, Bayesian optimization is applied to select the best hyperparameters. Experimental results demonstrate its effectiveness, with a mean Root Mean Square Error (RMSE) per k-folds of 4.91 dB and Standard Deviation (STD) 4.76 dB for more forested areas.