Files
Abstract
The advent of edge computing and IoT technologies has led to the ubiquitous presence of IoT-edge sensing systems, crucial for revolutionized data collection and analysis across numerous domains. However, the effective utilization and deployment of IoT-edge systems face significant challenges, particularly in terms of energy management and computational resource constraints. This dissertation addresses these challenges through a comprehensive approach that integrates three critical research areas to enhance the sustainability and computing capabilities of IoT-edge systems. Firstly, I introduce an innovative energy scheduling method for environmental sensors that harnesses solar power, enabling sustained operation in areas lacking conventional power infrastructures. It leverages an AI energy prediction model and a scheduler to optimize sensing activities in harsh, remote environments. Secondly, the study enhances edge computing capability by incorporating deep learning (DL) models and advanced system techniques leveraging AI multi-tenancy and heterogeneous AI accelerators. These techniques significantly increase the processing speed, inference throughput, and efficiency of data processing, reducing both network and computational delays. Finally, the research extends to the environmental monitoring of coastal wetlands with a tailored IoT-edge soil monitoring system that employs energy optimization and low-cost sensor calibration to effectively monitor soil properties under challenging field conditions.
This work integrated approaches improving the intelligent aspects of IoT-edge sensing systems and contributes to the effective monitoring and conservation of critical ecosystems, demonstrating a balanced focus on innovation and environmental sensing.