The Green IC group aims to develop cutting-edge research on extremely energy-efficient integrated circuits for high-performance, mobile and ubiquitous micro and nano systems. The group focuses on the following fundamental energy-centric research thrusts that cover the entire range from physical sensing to data sensemaking:
• untethering all human senses, from sight to hearing to smell, and enable truly distributed sensing via innovative energy-autonomous systems and wireless sensors with nearly perpetual operation (e.g., ultra-low power analog interfaces, digital processing, power management). This is far more difficult (and interesting) than existing sensors measuring traditional physical quantities (e.g., environmental, strain, etc.), getting closer to human capabilities while enabling the unprecedented ability to have it distributed and ubiquitous. Applications involve biomedical, distributed surveillance, smart cities, smart buildings, structural monitoring, smart cameras, among the others
• near-threshold VLSI circuits and systems for at least 10X improvement in energy efficiency for mobile and wearable applications (both error-free and energy-quality scalable computing are of interest)
• circuit- to system-level techniques to enhance the energy efficiency of high-speed VLSI circuits to enable continued performance enhancements under a tight power budget
• circuit and system integration with emerging and post-CMOS technologies for green computing
• enable ubiquitous data sensemaking via innovative ultra-lightweight approaches to embed machine learning on a chip, tackling the unaddressed challenge of retaining excellent accuracy of deep learning (and other techniques) while fitting the typical energy and area requirements of single chips. With the ultimate goal of on-chip learning and inference, we are approaching this problem from the collective perspectives of algorithm, circuit and architecture, while leveraging their inherent error resiliency to apply our strong expertise in energy-quality scalable circuits
• assuring hardware security ubiquitously and continuously (i.e., down to very low-end devices, always-on), through highly area/energy-efficient techniques to authenticate chips, and protect data from intrusion (including physical intrusion).