Directed information is an information-theoretic measure that can be interpreted as a directional information flow between random processes. The concept has been broadly used in causality analysis, as well as in the analysis of communication systems with feedback. In this talk, we discuss the fundamental trade-off between the best achievable control performance and the required feedback directed information, which turns out to be a central question in some engineering (e.g., networked control systems) and scientific (e.g., non-equilibrium thermodynamics) research domains. We discuss numerical solution algorithms, including (i) the semidefinite programming approach for linear-quadratic-Gaussian systems, and (ii) the forward-backward Arimoto-Blahut iteration for finite state systems. This is joint work with Takashi Tanaka (Univ. Texas, Austin) and Mikael Skoglund (KTH).
About the speaker
Henrik Sandberg is Professor at the Division of Decision and Control Systems, KTH Royal Institute of Technology, Stockholm, Sweden. He received the M.Sc. degree in engineering physics and the Ph.D. degree in automatic control from Lund University, Lund, Sweden, in 1999 and 2004, respectively. His current research interests include security of cyber-physical systems, power systems, model reduction, and fundamental limitations in control.