Research Vision
The swift progression of Internet of Things (IoT) techniques has driven a surge applications in precision agriculture, smart cities, and environment monitoring. These applications, which encompass vast numbers of wireless devices, pose a significant challenge to sustainability, as powering large quantities of devices over extensive areas via power grids or batteries is not scalable. Energy harvesting emerges as a promising solution. As a viable alternative, energy harvesting technology presents a sustainable and scalable solution, utilizing renewable energy sources in the environment such as solar, RF energy, and vibration energy to power IoT devices. Although, this Sustainable IoT (SIoT) approach allows for near-perpetual operation without battery replacements, it necessitates vigilant oversight to ensure efficiency and security. In light of these challenges, my research aims to tackle issues of sustainability, efficiency, and security within the realm of SIoT networks.
Another research direction for Dr. Pu's group is the development of robust edge computing solutions tailored for edge devices with constrained resources. This involves addressing challenges related to energy, computing, and networking limitations. The team's approach includes creating specialized feature map compression algorithms, implementing environment-aware model partitioning strategies, and employing advanced techniques in post-training quantization as well as quantization-aware training. These efforts aim to optimize the performance of edge computing devices within their operational constraints.
Another research direction for Dr. Pu's group is the development of robust edge computing solutions tailored for edge devices with constrained resources. This involves addressing challenges related to energy, computing, and networking limitations. The team's approach includes creating specialized feature map compression algorithms, implementing environment-aware model partitioning strategies, and employing advanced techniques in post-training quantization as well as quantization-aware training. These efforts aim to optimize the performance of edge computing devices within their operational constraints.