Self-Aware Computing for Cyber Physical Systems

Abstract

Novel cyber physical systems require digital systems that are very energy efficient. This program will develop a self-aware digital computation system that measures a range of static and dynamic variations, and adapts itself to them, so that performance, energy-efficiency, and reliability can be significantly increased. To enable this vision, we are investigating a range of approaches including the integration of a multitude of precision and flexible-supply-voltage analog sensors in the digital system but with a minimal area/testing overhead. We are further developing a framework to use the measured parameters, which can be imperfect, to effectively determine the optimal operating scenario. We strongly anticipate that such extensive instrumentation and control can advance the combination of performance, efficiency and reliability of digital computing systems beyond the current state of the art.