High Performance Computing in Science and Engineering

High Performance Computing in Science and Engineering

Theme: System Identification and Control Software

The Department of Computing Sciemce and High Performance Computing Center North (HPC2N) invite to the following two seminars by Dr Vasile Sima, National Institute for Research & Development in Informatics, Bucharest, Romania.

Seminar 1: System Identification, November 1, 13.15 - 14.15 in room MA236, MIT-huset

Abstract
Basic algorithmic and computational issues involved in subspace-based linear multivariable system identification are described. A new system identification toolbox-SLIDENT-has been developed and incorporated in the freely available Subroutine Library in Control Theory (SLICOT). Reliability, efficiency, and ability to solve large, industrial identification problems received a special consideration. Two algorithmic subspace-based approaches (MOESP and N4SID) and their combination, and both standard or fast techniques for data compression are provided. Structure exploiting algorithms and dedicated linear algebra tools enhances the computational efficiency. Extensive comparisons with the available subspace techniques have been made. The numerical results show that the SLIDENT toolbox is highly performant and able to solve large identification problems. SLIDENT abilities will be demonstrated during the talk.

Seminar 2: Control Software, November 2, 10.15 - 12 in room MA236, MIT-huset

Abstract
The emphasis in this talk is on computer-aided control systems analysis and design (CACSD). Algorithms, and especially the associated software for CACSD have recorded significant advances in the last years. These advances were motivated by the increasing complexity of the control problems to be addressed; moreover, practical problems are frequently ill-conditioned or badly scaled, and traditional methods often fail. Efficiency and reliability issues are, therefore, very important. The freely available library SLICOT (Subroutine Library In COntrol Theory) addresses these issues. SLICOT is based on the state-of-the-art linear algebra package LAPACK and on the BLAS collections, which guarantee high performance, reliability, and portability. SLICOT extends the functionality of LAPACK for control systems-related problems, like Lyapunov and Riccati equations, model reduction, system identification, etc. To increase the user-friendliness, Matlab/Scilab interfaces have been designed and implemented for many basic computational problems in systems and control. The latest two releases of the SLICOT Library, as well as the associated toolboxes have been developed in the framework of the thematic network NICONET (Numerics in Control), funded by the European Commission starting with 1996. Umeå University has been strongly involved as a partner in this network.

Some abilities of the SLICOT toolboxes will be demonstrated during this talk.

For more information about NICONET, SLICOT and web-computing facilities see the following URLs:

Updated: 2017-11-24, 17:02