1st Edition
Tensor Product Model Transformation in Polytopic Model-Based Control
Tensor Product Model Transformation in Polytopic Model-Based Control offers a new perspective of control system design. Instead of relying solely on the formulation of more effective LMIs, which is the widely adopted approach in existing LMI-related studies, this cutting-edge book calls for a systematic modification and reshaping of the polytopic convex hull to achieve enhanced performance. Varying the convexity of the resulting TP canonical form is a key new feature of the approach. The book concentrates on reducing analytical derivations in the design process, echoing the recent paradigm shift on the acceptance of numerical solution as a valid form of output to control system problems. The salient features of the book include:
- Presents a new HOSVD-based canonical representation for (qLPV) models that enables trade-offs between approximation accuracy and computation complexity
- Supports a conceptually new control design methodology by proposing TP model transformation that offers a straightforward way of manipulating different types of convexity to appear in polytopic representation
- Introduces a numerical transformation that has the advantage of readily accommodating models described by non-conventional modeling and identification approaches, such as neural networks and fuzzy rules
- Presents a number of practical examples to demonstrate the application of the approach to generate control system design for complex (qLPV) systems and multiple control objectives.
The authors’ approach is based on an extended version of singular value decomposition applicable to hyperdimensional tensors. Under the approach, trade-offs between approximation accuracy and computation complexity can be performed through the singular values to be retained in the process. The use of LMIs enables the incorporation of multiple performance objectives into the control design problem and assurance of a solution via convex optimization if feasible. Tensor Product Model Transformation in Polytopic Model-Based Control includes examples and incorporates MATLAB® Toolbox TPtool. It provides a reference guide for graduate students, researchers, engineers, and practitioners who are dealing with nonlinear systems control applications.
Introduction
Significant paradigm changes
Current computational methods and Applications
Role of the TP model transformation in control design
Part I: Tensor-product Model Transformation of Linear Parameter Varying Models
Higher Order Singular Value Decomposition of Tensors
Basic concept of tensor algebra
Higher Order Singular Value Decomposition (HOSVD)
Approximation trade-off by HOSVD
HOSVD-based canonical form of Linear Parameter-varying Models
Linear Parameter-Varying state-space model
HOSVD-based canonical form of LPV models
Numerical reconstruction of the HOSVD based canonical form
TP model transformation
Algorithm of the TP model transformation
Example of the TORA benchmark system
Computational relaxed TP model transformation
Column equivalence
Modified TP transformation
Evaluation of complexity reduction
Discretization complexity
Computational load of the HOSVD
Computational load of the tensor product
Numerical examples
A simple example
A more complex example
Convex TP model forms of Linear Parameter-varying Models
Convex TP model
Different types of convex TP models
Computation of different convex TP models
Methods for SN, NN and NO type matrices
Inverse, relaxed and normalized convex TP models (lNO, RNO)
The TORA benchmark example
Approximation and complexity trade-off by the TP model transformation
Approximation theory framework
No-where denseness
Examples
Part II: Control Design Examples
TP model transformation based design
Linear Matrix Inequality in system control design
Parallel Distributed Compensation based control design framework
Immediate link between the TP models and the PDC design framework
TP model transformation based control design methodology
Application to 2-D prototypical aeroelastic wing section with structural nonlinearity
Introduction to the prototypical aeroelastic wing section
Finite element convex TP model of the prototypical aeroelastic wing section
State-feedback control design
Observer based output-feedback control design
Application to 3 DOF helicopter with four propellers
Nomenclature
Equations of Motion of the RC Helicopter Dynamics
Finite element convex TP model of the -3-DOF RC helicopter
Control design of the3-DOF RC helicopter
Control results
Application to Parallel Double Inverted Pendulum
Nomenclature
Equations of Motion of the RC Helicopter Dynamics
Finite element convex TP model of the PDIP
Control design of the PDIP
Control results
Biography
Peter Beranyi, Ph.D, D. Sc, is head of the Computer and Automation Research Institute of the Hungarian Academy of Sciences and a professor at the Budapest University of Technology and Economics. He received his D.Sc in Informatics, his Ph.D. in Electrical Engineering, his M.Sc. in Education of Engineering Science, and his M.Sc. in Electrical Engineering at Budapest University of Technology and Economics. His research interest is on LPV- and LMI-based control design, modeling based on TP functions, fuzzy modeling, fuzzy rule interpolation, and calculation complexity reduction of various model types. He has written 48 journal papers for 262 publications.
Yeung Yam, is a professor in the Department of Mechanical and Automation Engineering at the Chinese University of Hong Kong. He obtained his B.Sc. from the Chinese University of Hong Kong, his M.Sc. from the University of Akron, Ohio, USA and his M.Sc., D.Sc. from the Massachusetts Institute of Technology, Cambridge, Massachusetts, USA. He has published over 100 technical papers in various areas of research, including human skill acquisition and analysis, dynamics modeling, control, system identification, fuzzy approximation, and intelligent and autonomous systems.
Peter Valarki, is a professor at the Budapest University of Technology and Economics. He graduated in mechanical engineering in 1971 at the Faculty of Transportation Engineering at the Technical University of Budapest, now the Budapest University of Technology and Economics. He also earned his Ph.D., his C.Sc. and his D.Sc. He is a founding member of the Hungarian Academy of Engineering and the main topics of his research field are the stochastic control theory, statistical system identification, and computational intelligency. He is the co-author of 10 books and more than 250 other scientific and technical publications.
"… well written and easily readable. … The examples and applications to 3 Degrees Of Freedom (DOFs) control schemes for helicopters, models for aeroelastic wing sections and models for controlling the behavior of suspension system in heavy trucks are the main strength of the book. … for control engineers with a solid mathematical formation as well as control theorists and even applied mathematicians."
—zbMATH 1308 in 2015"The book provides an introduction to a method that has potential to significantly advance the theory and practice of control system design. The modeling step is frequently the most time-consuming stage of practical control system design. The unifying TP representation of quasi LPV models described in this book has potential to make this stage more efficient as well as enabling many of the powerful LMI-based control design methods for LPV systems to be applied to practical problems."
—James Whidborne, Cranfield University, Bedfordshire, UK