Inverse Parametric Programming and Control of Hybrid Systems

Andreas Hempel (ETHZ)

SYSTEMS AND CONTROL SERIES

DATE: 2013-10-18
TIME: 11:00:00 - 12:00:00
LOCATION: RSISE Seminar Room, ground floor, building 115, cnr. North and Daley Roads, ANU
CONTACT: JavaScript must be enabled to display this email address.

ABSTRACT:
Many applications use ideas from the inverse optimization literature, for example expert learning and economic policy analysis. The main assumption is that a given function is assumed to be the explicit solution to a parametric optimization problem. The goal in inverse optimization is then to identify the generating problem data from the given function. We consider the case of piecewise affine (PWA) functions, which are of particular interest because the explicit solutions to parametric linear and quadratic programs have this structure. In this talk, we show that every PWA function can be obtained by a simple linear transformation from the solution to a linear or quadratic program and how the generating problem data can be found. The main step in the construction is the decomposition of the given PWA function into a convex and a concave part. This result expands our previous contributions considerably and enables us to model every PWA dynamical system in terms of an optimization problem. Using this optimization model for control input computations seems to be beneficial when compared to currently standard approaches.

BIO:
Since Feb 11, Andres Hempel has been a PhD student in Control at Automatic Control Laboratory at ETH Zurich (Switzerland), research topics include inverse optimization, hybrid system modeling and control, and nonconvex optimization Previously, he studied Mechatronics (MSc and BSc) at Hamburg University of Technology(Germany) and Technology Management (MBA) at Northern Institute of Technology Management (Hamburg, Germany).