Real-time Optimization as a Cyber-Physical System
Event details
| Date | 16.02.2021 |
| Hour | 09:00 › 10:00 |
| Speaker | Dr. Dominic Liao-Mc Pherson, Automatic Control Laboratory, ETH Zürich |
| Location | Online |
| Category | Conferences - Seminars |
Abstract: Accelerating advances in computing, sensors, and data generation are beginning to profoundly impact society through the control and automation of complex systems such as electrical grids (Smart Grid), manufacturing processes (Industry 4.0), and the increasing ubiquity of drones, self-driving cars, and other autonomous systems. Due to their complexity, these intelligent systems are controlled by algorithms running in real-time and in closed-loop with the system, rather than closed-form feedback laws, resulting in a tightly coupled cyber-physical systems (CPS). Developing reliable real-time algorithms and the theoretical tools needed to understand the complex feedback interactions between algorithms and the physical systems they control is challenging but essential for realizing the potential of ubiquitous closed-loop control and for building the public trust necessary for widespread deployment.
In this talk, I argue for a cyber-physical systems view of optimization/algorithm based control wherein the physical system and the algorithms controlling it are designed and analyzed together. First, I present Time-distributed Optimization (TDO), a unifying framework for studying the system theoretic consequences of computational limits in the context of Model Predictive Control (MPC). I show that it is possible to recover the stability and robustness properties of optimal MPC despite limited computational resources and illustrate how this CPS viewpoint can be exploited to certify the closed-loop system. Further, I illustrate the applicability of these methods in the real-world through diesel engine, and autonomous driving examples. Second, I discuss ongoing work on feedback equilibrium seeking wherein the goal is to regulate a physical system to the solution of a time-varying variational inequality. I present tracking error, stability, and robustness results and provide motivating examples in smart buildings and DC power grids. Finally, I briefly discuss how a CPS viewpoint can be used for certification, health monitoring, and supervision of algorithmic controllers and provide future perspectives on extensions to equilibrium seeking and learning algorithms.
Bio: Dominic Liao-McPherson obtained his BASc (with High Honours) in Engineering Science (Aerospace Option) from the University of Toronto in 2015 and his PhD in Aerospace Engineering and Scientific Computing from the University of Michigan in 2020. His research interests lie at the interface of systems theory, optimization, variational analysis, and numerical methods with applications in autonomous vehicles, smart grids, and advanced manufacturing. He is a recipient of the Prof. Pierre T. Kabamba award, the Richard and Eleanor Towner prize, the Sir James Lougheed award, and was a finalist in the 2019 ECC best student paper competition.
In this talk, I argue for a cyber-physical systems view of optimization/algorithm based control wherein the physical system and the algorithms controlling it are designed and analyzed together. First, I present Time-distributed Optimization (TDO), a unifying framework for studying the system theoretic consequences of computational limits in the context of Model Predictive Control (MPC). I show that it is possible to recover the stability and robustness properties of optimal MPC despite limited computational resources and illustrate how this CPS viewpoint can be exploited to certify the closed-loop system. Further, I illustrate the applicability of these methods in the real-world through diesel engine, and autonomous driving examples. Second, I discuss ongoing work on feedback equilibrium seeking wherein the goal is to regulate a physical system to the solution of a time-varying variational inequality. I present tracking error, stability, and robustness results and provide motivating examples in smart buildings and DC power grids. Finally, I briefly discuss how a CPS viewpoint can be used for certification, health monitoring, and supervision of algorithmic controllers and provide future perspectives on extensions to equilibrium seeking and learning algorithms.
Bio: Dominic Liao-McPherson obtained his BASc (with High Honours) in Engineering Science (Aerospace Option) from the University of Toronto in 2015 and his PhD in Aerospace Engineering and Scientific Computing from the University of Michigan in 2020. His research interests lie at the interface of systems theory, optimization, variational analysis, and numerical methods with applications in autonomous vehicles, smart grids, and advanced manufacturing. He is a recipient of the Prof. Pierre T. Kabamba award, the Richard and Eleanor Towner prize, the Sir James Lougheed award, and was a finalist in the 2019 ECC best student paper competition.
Practical information
- General public
- Free