FADR-MLP: A Lightweight Neural Network-Based Surrogate Model for Adaptive Data Rate Control in LoRaWAN
Resumo
The optimization of Adaptive Data Rate (ADR) mechanisms is fundamental to the energy efficiency and scalability of LoRaWAN networks. However, implementing intelligent control algorithms on resource-constrained Internet of Things (IoT) edge devices remains a significant challenge. This paper proposes and evaluates a novel mechanism based on a surrogate model for End Devices (EDs), which employs a Multi-Layer Perceptron (MLP) artificial neural network to adjust the following transmission parameters: Spreading Factor (SF), Transmission Power (TP), and the number of measurement packets, denoted as M. The MLP network was trained on a dataset generated by a pre-existing expert system, and its fidelity was rigorously validated via K-Fold cross-validation, achieving a Coefficient of Determination (R²) exceeding 0.92. Furthermore, a complexity analysis quantified the model’s low computational cost, revealing a memory footprint of 17.77 KB and an inference cost of 8,832 floating-point operations (FLOPs). The comparative analysis reveals that while the fuzzy inference system (FIS) it emulates may be more efficient in moderate-precision scenarios, the MLP’s architectural advantage for hardware execution positions it as a promising solution for implementation at the edge of LoRaWAN-based IoT systems.
Referências
G. A. Sarmento Neto, T. A. R. da Silva, P. F. F. Abreu, A. F. D. S. Veloso, L. H. D. O. Mendes, and J. V. dos Reis Junior, “A fuzzy logic approach to adaptive data rate for mobile lorawan environments,” in 2024 XIV Brazilian Symposium on Computing Systems Engineering (SBESC), 2024, pp. 1–6.
F. J. V. Santos, G. A. S. Neto, T. A. R. da Silva, P. F. F. de Abreu, L. H. de O. Mendes, and J. V. dos Reis Junior, “Adr based on fuzzy logic with dynamic number of uplink transmissions adjustment for lorawan networks in mobility scenarios,” in Proceedings of the Brazilian Conference on Intelligent Systems (BRACIS). Teresina, Piauí, Brazil: Federal University of Piauí(UFPI), 2025, aceito para publicação (Accepted for publication - BRACIS 2025).
H. Wang, P. Pei, R. Pan, K. Wu, Y. Zhang, J. Xiao, and J. Yang, “A Collision Reduction Adaptive Data Rate Algorithm based on the FSVM for a Low-Cost LoRa Gateway,” Mathematics, vol. 10, no. 21, p. 3920, 2022.
M. A. Lodhi, L. Wang, A. Farhad, K. I. Qureshi, J. Chen, K. Mahmood, and A. K. Das, “A contextual aware enhanced LoRaWAN adaptive data rate for mobile IoT applications,” Computer Communications, vol. 232, p. 108042, 2025.
A. Farhad and J.-Y. Pyun, “AI-ERA: Artificial intelligence-empowered resource allocation for LoRa-Enabled IoT applications,” IEEE Transactions on Industrial Informatics, vol. 19, no. 12, pp. 11 640–11 653, 2023.
J.-S. R. Jang, “Anfis: Adaptive-network-based fuzzy inference system,” IEEE Transactions on Systems, Man, and Cybernetics, vol. 23, no. 3, pp. 665–685, 1993.
Y. Cheng, D. Wang, P. Zhou, and T. Zhang, “A survey of model compression and acceleration for deep neural networks,” IEEE Signal Processing Magazine, vol. 37, no. 5, pp. 160–174, 2020.
M. González-Palacio, D. Tobón-Vallejo, L. M. Sepúlveda-Cano, M. Luna-Delrisco, C. Röhrig, and L. B. Le, “Machine-learning-assisted transmission power control for lorawan considering environments with high signal-to-noise variation,” IEEE Access, vol. 12, pp. 54 466–54 487, 2024.
S. A. Bhat, N.-F. Huang, I. Hussain, and U. Sajjad, “Correlating the ambient conditions and performance indicators of the lorawan via surrogate gaussian process-based bidirectional lstm stacked autoencoder,” IEEE Transactions on Network and Service Management, vol. 20, no. 3, pp. 3413–3427, 2023.
G. James, D. Witten, T. Hastie, and R. Tibshirani, An Introduction to Statistical Learning: with Applications in R. Springer, 2013, chapter 5: Resampling Methods.
