Kostas Tsakalis
School of Electrical Computer and Energy Engineering
Arizona State University
Tempe, AZ, USA
Email: [email protected]
Abstract
We discuss the use of H-infinity approximation for identification from input-output data and online adaptation. Standard least squares algorithms are common solutions for this problem, but their estimates exhibit a well-known strong dependence on the properties of the excitation. This can cause large identification errors and, ultimately, adaptation bursts when insufficient excitation is combined with disturbances or unmodeled dynamics. In an alternative formulation of the estimation problem, we use a filter-bank to decompose the error signal to different components and minimize approximately the H-infinity norm of the sensitivity-weighted error operator. This approach results in a more robust identification of the system and its controller tuning. While in general there are higher excitation requirements, the benefit is in the efficiency of data utilization, e.g., in the problem of closed-loop bandwidth maximization using either a single experiment as opposed to an iterative approach. A byproduct of this identification procedure is the computation of a ‘health indicator’ to describe the confidence in the estimated parameters and the identified system. This can have important practical implications in the implementation of high-level supervisory systems that monitor the performance of the identification and control loops. In a different application, the approximate H-infinity objective has led to an on-line tuning algorithm for PID tuning that has been shown to provide reliable controllers, even in cases of large mismatch between the target and the feasible loop shapes. The general theme of these results is that the formulation of a min–max optimization of an operator error provides a advantage over signal error optimization, in terms of the quantitative characteristics of system identification and controller tuning, which are in turn significant in practical applications Supported by NSF Grant ECCS-1102390.
Biography
Dr. Konstantinos Tsakalis is a Professor in the School of Electrical, Computer, and Energy Systems Engineering at Arizona State University since 1988. His work is in the theory and applications of control systems, adaptive control, system identification and optimization. Starting in 1995 and in collaboration with Semy Engineering, he developed an integrated identification and controller design procedure for the temperature control of diffusion furnaces, which was awarded 5 US patents and the 1998 Editor’s Choice, Best Products Award from Semiconductor International. Dr. Tsakalis has also worked on the application of robust control theory, system identification and optimization principles in various industrial problems in collaboration with Honeywell and EPRI. His recent activities include power system and biomedical applications, and in particular, prediction and control of epileptic seizures. He is also actively pursuing the transfer of industrial research experience in the classroom.
