Energy-Aware TinyML for Batteryless Fire Detection Nodes Using LoRaWAN and Ambient Energy Harvesting
Article Sidebar
Main Article Content
Abstract
Batteryless wireless sensor nodes promise maintenance-free environmental monitoring at large scale. However, their supply is intermittent and scarce. This conflicts directly with the low-latency, high-reliability requirements of early fire detection. This paper addresses that conflict. We present EAF-TinyML, an energy-aware tiny machine-learning framework for batteryless fire-detection nodes. The framework combines ambient photovoltaic and thermoelectric harvesting, supercapacitor buffering, and a Long-Range Wide-Area Network (LoRaWAN) backhaul. Three mechanisms are co-designed around a single energy budget: a gated inference pipeline, a state-of-charge-driven adaptive duty-cycle controller, and a confidence-aware transmission policy. We formalise the node energy balance and derive the energy-neutrality condition that ties the duty-cycle period to the harvested power. We evaluate the framework on the public Smoke Detection multi-sensor dataset and on an ARM Cortex-M4 hardware-in-the-loop prototype driven by replayed harvesting traces. The quantised model occupies 41 kB of flash and runs one inference in 18 ms at 73 µJ. Detection F1 is 0.967 ± 0.006 over five-fold episode-wise cross-validation with three random seeds. Against two node-level analogues re-evaluated under a common energy model, EAF-TinyML reduces average energy per cycle by 38–61% and lowers median alarm latency to 9.4 s. Energy-neutral operation is sustained in bench-replayed traces down to about 250 lx of sustained illumination; below roughly 120 lx for more than 36 h, the node degrades gracefully to intermittent operation rather than failing. The results indicate that system-level energy co-design, not smaller models alone, makes self-powered fire sentinels practical. Field validation remains future work
Article Details
Section
This work is licensed under a Creative Commons Attribution 4.0 International License.
How to Cite
References
[1] G. P. Acharya, L. Poluboyina, J. Veeragamoorthi, and C. Joydeb, “An efficient and low cost realization of LoRa-based real-time forest protection system,” Indonesian Journal of Electrical Engineering and Computer Science, vol. 34, no. 3, pp. 1452–1462, 2024, doi: 10.11591/ijeecs.v34.i3.pp1452-1462.
[2] M. Ouahsini et al., “Optimized wireless sensor network architecture for AI-based wildfire detection in remote areas,” Fire, vol. 8, no. 7, art. 245, 2025, doi: 10.3390/fire8070245.
[3] C. J. Briones-Suarez et al., “ForestProtector: An IoT architecture integrating machine vision and deep reinforcement learning for efficient wildfire monitoring,” arXiv:2501.09926, 2025.
[4] O. L. A. López et al., “Energy-sustainable IoT connectivity: Vision, technological enablers, challenges, and future directions,” IEEE Open Journal of the Communications Society, vol. 4, pp. 2609–2666, 2023, doi: 10.1109/OJCOMS.2023.3323832.
[5] M. Sidorov et al., “Towards mass-scale IoT with energy-autonomous LoRaWAN sensor nodes,” Sensors, vol. 24, no. 13, art. 4334, 2024, doi: 10.3390/s24134334.
[6] L. Mottola, K. S. Yıldırım, and J. Hester, “The Internet of Batteryless Things,” Communications of the ACM, vol. 67, no. 3, pp. 64–73, 2024, doi: 10.1145/3624718.
[7] A. Banotra et al., “Energy harvesting in self-sustainable IoT devices and applications based on cross-layer architecture design: A survey,” Computer Networks, vol. 236, art. 110011, 2023, doi: 10.1016/j.comnet.2023.110011.
[8] M. R. Islam et al., “Micro energy harvesting for IoT platform: Review analysis toward future research opportunities,” Heliyon, vol. 10, no. 6, art. e27778, 2024, doi: 10.1016/j.heliyon.2024.e27778.
[9] H. Han, S. Trimi, and S. M. Lee, “Tiny Machine Learning (TinyML): Research trends and future application opportunities,” Array, vol. 29, art. 100674, 2026, doi: 10.1016/j.array.2025.100674.
[10] V. Rajapakse, I. Karunanayake, and N. Ahmed, “Intelligence at the extreme edge: A survey on reformable TinyML,” ACM Computing Surveys, vol. 55, no. 14s, pp. 1–30, 2023, doi: 10.1145/3583683.
[11] N. N. Alajlan and D. M. Ibrahim, “Advancements in TinyML: Applications, limitations, and impact on IoT devices,” Electronics, vol. 13, no. 17, art. 3562, 2024, doi: 10.3390/electronics13173562.
[12] N. Schizas, A. Karras, C. Karras, and S. Sioutas, “TinyML for ultra-low power AI and large-scale IoT deployments: A systematic review,” Future Internet, vol. 14, no. 12, art. 363, 2022, doi: 10.3390/fi14120363.
[13] D. L. Dutta and S. Bharali, “TinyML meets IoT: A comprehensive survey,” Internet of Things, vol. 16, art. 100461, 2021, doi: 10.1016/j.iot.2021.100461.
[14] S. Umesh and S. Mittal, “A survey of techniques for intermittent computing,” Journal of Systems Architecture, vol. 112, art. 101859, 2021, doi: 10.1016/j.sysarc.2020.101859.
[15] U. Dampage, L. Bandaranayake, R. Wanasinghe, K. Kottahachchi, and B. Jayasanka, “Forest fire detection system using wireless sensor networks and machine learning,” Scientific Reports, vol. 12, art. 46, 2022, doi: 10.1038/s41598-021-03882-9.
[16] T. Brito et al., “Data acquisition filtering focused on optimizing transmission in a LoRaWAN network applied to the WSN forest monitoring system,” Sensors, vol. 23, no. 3, art. 1282, 2023, doi: 10.3390/s23031282.
[17] R. Mukhia, K. G. S. Jayarathna, and A. Lertsinsrubtavee, “Performance evaluation of LoRaWAN forest fire monitoring network in the wild,” in Proc. 18th Asian Internet Engineering Conf. (AINTEC ’23), 2023, pp. 96–104, doi: 10.1145/3630590.3630602.
[18] A. De Rango, L. Furnari, F. Cortale, A. Senatore, and G. Mendicino, “Wildfire early warning system based on a smart CO₂ sensors network,” Sensors, vol. 25, no. 7, art. 2012, 2025, doi: 10.3390/s25072012.
[19] S. N. Saydirasulovich, M. Mukhiddinov, O. Djuraev, A. Abdusalomov, and Y.-I. Cho, “An improved wildfire smoke detection based on YOLOv8 and UAV images,” Sensors, vol. 23, no. 20, art. 8374, 2023, doi: 10.3390/s23208374.
[20] R. H. Jhaveri, H. R. Chi, and H. Wu, “TinyML for empowering low-power IoT edge consumer devices,” IEEE Transactions on Consumer Electronics, vol. 70, no. 4, pp. 7318–7321, 2024, doi: 10.1109/TCE.2024.3482353.
[21] P.-E. Novac, G. Boukli Hacene, A. Pegatoquet, B. Miramond, and V. Gripon, “Quantization and deployment of deep neural networks on microcontrollers,” Sensors, vol. 21, no. 9, art. 2984, 2021, doi: 10.3390/s21092984.
[22] E. Liberis and N. D. Lane, “Differentiable neural network pruning to enable smart applications on microcontrollers,” Proc. ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, vol. 6, no. 4, pp. 1–19, 2023, doi: 10.1145/3569468.
[23] P. Mittal et al., “Deploying TinyML for energy-efficient object detection and communication in low-power edge AI systems,” Scientific Reports, vol. 15, art. 27818, 2025, doi: 10.1038/s41598-025-27818-9.
[24] J. Moosmann, M. Giordano, C. Vogt, and M. Magno, “TinyissimoYOLO: A quantized, low-memory footprint, TinyML object detection network for low power microcontrollers,” in Proc. IEEE 5th Int. Conf. on AI Circuits and Systems (AICAS), 2023, pp. 1–5, doi: 10.1109/AICAS57966.2023.10168657.
[25] Y. Abadade, N. Benamar, M. Bagaa, and H. Chaoui, “Empowering healthcare: TinyML for precise lung disease classification,” Future Internet, vol. 16, no. 11, art. 391, 2024, doi: 10.3390/fi16110391.
[26] A. Tschand et al., “MLPerf Power: Benchmarking the energy efficiency of machine learning systems from microwatts to megawatts for sustainable AI,” arXiv:2410.12032, 2024.
[27] A. Sabovic, A. K. Sultania, C. Delgado, L. De Roeck, and J. Famaey, “Towards energy-aware tinyML on battery-less IoT devices,” Internet of Things, vol. 22, art. 100736, 2023, doi: 10.1016/j.iot.2023.100736.