Seminars by Carsten Scherer and Mark Cannon

Everyone is welcome

 

10:00-10:45

From Separation to Lyapunov Functionals

Carsten Scherer, University of Stuttgart

 

Abstract:

A classical approach to stability and performance analysis of feedback interconnections is based on graph separation. As a popular incarnation, the main result on integral quadratic constraints (IQCs) renders such separation based techniques computational for a linear system in feedback with a large variety of components in the loop. Despite its long history, the relation between general integral quadratic constraints and Lyapunov theory had not been established until recently. The purpose of this talk is to link both worlds through an encompassing result which is based on the new notion of a finite horizon IQCs with a terminal cost. We reflect on the general merits of this approach for inferring invariance properties on the basis of absolute stability criteria

 

 

10:45-11:30

Adaptive Model Predictive Control: Robustness, Performance Enhancement and Parameter Estimation
Mark Cannon, University of Oxford

 

Abstract:
Control algorithms that combine online model identification with optimization of predicted performance have been around since the origins of Model Predictive Control (MPC) some 40 years ago. However few control strategies based on MPC with online model identification provide guarantees of robust performance and constraint satisfaction. Recent developments in robust predictive control, set-based identification and convex optimization have led to a resurgence of interest in this direction. This talk will outline recent work on computationally tractable robust adaptive MPC formulations for linear systems with additive disturbances subject to state and control constraints. The approach has the potential to overcome a fundamental limitation of robust MPC, namely that the amount of uncertainty in system models and unknown disturbances restricts the achievable performance, even though model uncertainty can often be reduced by making use of information acquired during the closed loop operation of the controller. We will consider connections with the well-known dual control problem and discuss how to balance conflicting requirements on control signals for achieving good tracking performance and improving parameter estimates through convex constraints that ensure persistency of excitation.

 


Tidspunkt

fre 18 jan 19
10:00 - 11:30

Arrangør

DTU Compute

Hvor

Technical University of Denmark

Anker Engelunds Vej 1

Building 101, Room S10

DK-2800 Kgs Lyngby