CoSy Zoom seminar
- Date: 18 May 2021, 12:15–13:00
- Location: Zoom
- Type: Seminar
- Lecturer: Axel Ringh, postdoctoral researcher at the Department of Electronic and Computer Engineering at The Hong Kong University of Science and Technology
- Contact person: Benjamin Meco
Title: Control of ensembles via structured optimal transport
Register here
Abstract:
Optimal transport is an optimal allocation problem, which in its simplest form can be interpreted as a minimum-cost flow problem. Historically, it has been an important tool in economics and operations research, but more recently it has also been successfully applied in areas such as signal processing, imaging, statistics, machine learning, and systems and control. The latter is partly due to a rapid development over the last decades, related to both theoretical aspects and numerical solution methods for optimal transport, which has made it a readily available tool to tackle many problems.
In this talk I will show how optimal transport can be used to steer an ensembles of agents. More precisely, starting from a matching problem I will show how optimal transport can be used to formulate and solve optimal control problems for ensembles of dynamical systems. The resulting numerical optimization problems are large-scale problems, often with billions of variables. Nevertheless, we develop efficient numerical algorithms for (approximately) solving these problems. The methods build on using both the acclaimed Sinkhorn iterations, and on using the underlying graph-structure of the problems.