Research Statement

Understanding the nature of distributed decision-making systems is a universal challenge for scientists and engineers today. My research goal is to unveil fundamental principles of distributed systems and advance large-scale decision-making technologies for our future society. My research projects are categorized in terms of traditional disciplinary boundaries as (1) control-theoretic, (2) information-theoretic and (3) game-theoretic.

Control theoretic approach

 

Since the 1960s, control theorists have been studying how the difficulty of distributed decision-making problems depends on various factors, such as information structure (i.e., who knows what, when) and performance specifications (i.e., what needs to be achieved). While arbitrary combinations of these factors typically result in NP-hard problems, in certain exceptional and non-trivial cases distributed control problems admit computationally tractable solutions. Discovering such classes of distributed control problems is extremely valuable, since this type of knowledge directly contributes to our ability to develop scalable infrastructures.

Related research projects:

  • Distributed control of positive systems: [TAC2011][SCL2013][TAC2013]

  • Distributed LQG control under hierarchical information structure: [ACC2014]

Information-theoretic approach

 

When multiple individuals must perform a certain task in a distributed environment, the issue of communication costs comes into the picture. Understanding the interplay between Shannon's information theory and real-time decision-making theory is a key mission in this context. Although the challenge here is multi-spectrum and numerous research activities are going on, in our view the essence of the problem can be captured by the following “data-frugal” optimal control problem:

What is the minimal information that a decision-maker must acquire from the environment in order to complete tasks with some required accuracy?

Thorough understanding of this question leads to, for instance, the development of efficient real-time data-compression and encoding algorithms for networked control systems. It is also closely related to the issue of privacy and cyber-security in real-time environments. The main tool we use to analyze the information flow in real-time decision systems is the “directed information theory,” which has attracted much attention in neuro/cognitive science and statistical physics in recent years.

Related research projects:

Game-theoretic approach

 

In distributed systems, an additional difficulty arises when self-interested individuals with misaligned preferences are involved in the decision-making process. Game theory provides rich and elegant guidelines for understanding and resolving such situations.

Related research projects:

  • Distributed computation and incentives: [TCNS2016]

  • Real-time mechanism design for energy market: [ACC2012][CDC2016]