Strategic Information Transmission in the Control of Cyber-Socio-Physical Systems featuring Professor Cedric Langbort

Abstract: The efficiency of feedback control as a regulation mechanism is predicated on the engineer’s ability to observe the system’s behavior through sensors and correct it through actuators. What happens when these sensors cannot be trusted to faithfully transmit their measurements but, instead, may manipulate reports to their own advantage? Likewise, what happens when actuators, instead of obediently applying requested commands, must be persuaded to do so, lest they act selfishly?

These questions are becoming increasingly relevant in the context of cyber-socio-physical systems — in which sensors and databases may be attacked or subverted in cyber-attacks, and in which `actuators’ are often human agents who cannot be assumed to blindly implement control laws — and call for new viewpoints that go beyond traditional paradigms in controls such as fault-tolerance or robustness, which do not fully encapsulate information asymmetry and strategic intent.

In this talk, we show how such problems can be tackled under the conceptual umbrella of Strategic Information Transmission (SIT) first introduced in Economics (once models and assumptions are suitably modified to accommodate the engineering setting) and model the tension between the control designer’s goal and the self-interested sensors’ and actuators’ actions as Stackelberg games between (a) privately-informed leader(s) and a decision-maker whose belief about the state of the world is influenced by the messages it receives from the leader.  Particularizing this framework to the problems of state estimation with strategic sensors and road congestion control, we uncover a number of maybe surprising effects such as, e.g., the fact that  (1) it is always to the advantage of a strategic sensor to not “flat-out” lie, (2) providing complete (or no) information about the state of a road to all commuters can be socially inefficient, and (3) that a population of boundedly strategic sensors may result in better estimates than honest but misinformed ones.

We then discuss the implication of these findings for the design and control of CPS systems, and point to current extensions of SIT necessary to capture other challenging aspects of the control of cyber-socio-physical systems.

Biosketch: Cédric Langbort is an Associate Professor of Aerospace Engineering (with tenure) at the University of Illinois at Urbana–Champaign (UIUC), where he is also affiliated with the Decision & Control Group at the Coordinated Science Lab (CSL), and the Information Trust Institute.  Prior to joining UIUC in 2006, he studied at the Ecole Nationale Supérieure de l’Aéronautique et de l’Espace-Supaero in Toulouse (France), the Institut Non-Linéaire in Nice (France), and Cornell University, from which he received the Ph.D. in Theoretical & Applied Mechanics in January 2005. He also spent a year and a half as a postdoctoral scholar in the Center for the Mathematics of Information at Caltech.He works on applications of control, game, and optimization theory to a variety of fields; most recently to ”smart infrastructures” problems within the Center for People & Infrastructures which he co-founded and co-directs at CSL. He is a recipient of the NSF CAREER Award, the former advisor of recipients of the IEEE CDC Best Student Paper Award and the A2C2 Donald P. Eckman Award, and has been a subject editor for OCAM, and Systems & Control Letters. He is the PI of the recently announced ARO 2020 MURI Award on Information Exchange Network Dynamics.

Date/Time:
Date(s) - Mar 09, 2020
3:00 pm - 4:00 pm

Location:
38-138 Engineering IV
420 Westwood Plaza Los Angeles CA 90095