As a control theorist, I have been continually fascinated by the remarkable sensory-motor skills humans display when making decisions. My ongoing research focus has been on a mathematical characterization of efficient control systems from information-theoretic perspectives. Over the past decade, my primary scholarly goal has been understanding the secret of “efficient perception.” Specifically, I have been trying to quantify the minimum “data rate” at which an agent needs to gather information from the surroundings to keep making sensible real-time decisions. Answering these questions requires a unification of Information Theory and Control Theory at a fundamental level.
Although both of these fields are well-established on their own, it is surprisingly tricky to translate a concept from one field to another. One reason is that these subjects are rarely taught together in educational programs, resulting in distinct research cultures. To address this gap, I am developing a graduate-level course on Networked Control Systems, aiming to teach elements of these two subjects in a single semester.
Research portfolio: 2018 – 2023
In our 2018 publication [TAC18], we introduced the concept of minimum information control, which was motivated by a specific design challenge in Networked Control Systems (NCS). In this paper, we proposed an application of directed information, an information-theoretic measure of causality, to quantify the information flow in the perception-action cycle. This approach not only helps to identify the minimum amount of information needed for decision-making but also allows for the differentiation between task-relevant and task-irrelevant information. Subsequently, my research portfolio has been expanded into several directions exploring the interface between information and control theory.