Edge intelligence for low-power IoT connectivity and operation
Thesis event information
Date and time of the thesis defence
Place of the thesis defence
Anttilansali (FY1103), Linnanmaa campus
Topic of the dissertation
Edge intelligence for low-power IoT connectivity and operation
Doctoral candidate
Master of Science (Technology) David Ernesto Ruiz-Guirola
Faculty and unit
University of Oulu Graduate School, Faculty of Information Technology and Electrical Engineering, CWC- Radio Technology
Subject of study
Green communications for low-power IoT
Opponent
Associate Professor Cicek Cavdar, KTH Royal Institute of Technology
Custos
Associate Professor Onel L. Alcaraz López, University of Oulu
Smarter Internet of Things Devices That Last Longer on Less Energy
The rapid growth of the Internet of Things (IoT) poses challenges to sustainability, including increased maintenance requirements and higher energy consumption. To address these issues, it is crucial to develop self-sustainable IoT ecosystems, minimizing energy consumption and optimizing resource use. This thesis addresses the design and development of energy-efficient optimization and machine learning (ML) mechanisms at the edge to support sustainable IoT deployments with minimal communication overhead. It combines theoretical performance analysis with algorithm development.
First, we characterize and predict IoT traffic patterns using ML. We identify key behaviors related to IoT traffic and develop reliable traffic models for both simulation and prediction. We validate these models against real-world datasets and design ML-based traffic prediction frameworks for low-power IoT device deployments. Next, we integrate lightweight ML at the edge, IoT device sensing and reporting operation optimization, discontinuous reception, and wake-up radio mechanisms to enhance energy efficiency in low-power IoT networks. We propose traffic prediction-assisted algorithms to save energy and improve the adaptability and effectiveness of these mechanisms in dynamic IoT environments. Finally, we focus on energy-neutral IoT scenarios, wherein IoT devices rely on energy harvesting to sustain operation, and propose ML-based edge mechanisms to reduce energy consumption while maintaining near-optimal performance in these scenarios.
These solutions leverage edge computing and predictive modeling to address the challenges and uncertainties of low-power IoT environments, providing scalable and sustainable energy management methods.
First, we characterize and predict IoT traffic patterns using ML. We identify key behaviors related to IoT traffic and develop reliable traffic models for both simulation and prediction. We validate these models against real-world datasets and design ML-based traffic prediction frameworks for low-power IoT device deployments. Next, we integrate lightweight ML at the edge, IoT device sensing and reporting operation optimization, discontinuous reception, and wake-up radio mechanisms to enhance energy efficiency in low-power IoT networks. We propose traffic prediction-assisted algorithms to save energy and improve the adaptability and effectiveness of these mechanisms in dynamic IoT environments. Finally, we focus on energy-neutral IoT scenarios, wherein IoT devices rely on energy harvesting to sustain operation, and propose ML-based edge mechanisms to reduce energy consumption while maintaining near-optimal performance in these scenarios.
These solutions leverage edge computing and predictive modeling to address the challenges and uncertainties of low-power IoT environments, providing scalable and sustainable energy management methods.
Last updated: 28.8.2025