CDSC - eng.newcastle.edu.au · Chalup. ARC Centre for Complex Dynamic Systems and Control – ANNUAL REPORT 2003 4 . CDSC STAFF DIRECTOR: Professor Richard H. Middleton ... Dr David

CCCDDDSSSCCC

Complex Dynamic Systems and Control

2003 Annual Report

Director’s Report ......................................................................... 3

2003 Highlights ........................................................................... 4

Staff............................................................................................. 5

Postgraduate Research Students ................................................. 6

Advisory Board ............................................................................ 9

Review Committee..................................................................... 10

Visitors ...................................................................................... 11

Short Courses/Workshops:........................................................ 13

Seminars: .................................................................................. 14

Research Programmes:.............................................................. 16

A. Control System Design........................................................ 16 A.1 Implementation of Model Predictive Control .......................... 16 A.2 Analysis of Optimal and Constrained Control Systems............. 18 A.3 Cross Directional Control .................................................... 20 A.4 Fundamental Design Trade-offs ........................................... 21 A.5 Finite Alphabet Systems ..................................................... 23 A.6 Duality of Constrained Control and Estimation ....................... 26 A.7 Development of Virtual Laboratory for Control System Design . 27 A.8 Automotive Power Train Control (General Motors, USA) .......... 29

B. Mathematical Systems Theory ............................................ 30 B.1 Graph Algebras: Operator Algebras We Can See .................... 30 B.2 Groupoid Models ............................................................... 31 B.3 Higher-Rank Graphs .......................................................... 31 B.4 Identification of Graph Algebras .......................................... 32 B.5 Symbolic Dynamics and Operator Algebras ........................... 32 B.6 Coupled Oscillators ............................................................ 33 B.7 Differential Algebraic Control............................................... 33 B.8 Nonlinear Analysis and Fixed-Point Theory ............................ 34

C. Bayesian Learning............................................................... 37 C.1 Markov Chain Monte Carlo (MCMC) Project: .......................... 37 C.2 Bayesian Modelling ............................................................ 38 C.3 Control Oriented System Estimation..................................... 39 C.4 Nonlinear and Mixture Modelling.......................................... 40 C.5 Robot Location and Vision................................................... 42

D. Signal Processing................................................................ 44

1 ARC Centre for Complex Dynamic Systems and Control – ANNUAL REPORT 2003

CCCDDDSSSCCC

D.1 Subband Equalization.........................................................44 D.2 Channel Modelling .............................................................45 D.3 Image Representation, Sampling and Reconstruction..............45 D.4 Dynamic Modeling of Turbo Decoding ...................................46

E. Process Control and Optimisation........................................ 48 E.1 Integrated Mine Planning (BHP Billiton).................................48 E.2 Optimisation Based Operator Guidance Schemes (BHP-Billiton Innovation) ............................................................................50 E.3 Next Generation Model Based Control Tools (Matrikon) ...........54 E.4 Nutating Mill Control (Hicom International Pty. Ltd.) ...............54

F. Mechatronics ....................................................................... 55 F.1 Dealing with Hystersis in Piezoelectric Transducers .................55 F.2 Vibration Control Using Shunted Piezoelectric Transducers .......57 F.3 Instrumentation for Smart Structures ...................................58 F.4 Electromagnetic Transducers ...............................................60 F.5 Nanopositioning Systems ....................................................61

Publications 2003....................................................................... 62 Books ....................................................................................62 Chapters in Books ...................................................................62 Plenary and Invited Addresses ..................................................62 Journal Papers (published) .......................................................63 Journal Papers (Accepted for publication) ...................................66 Conference Papers...................................................................67 Industry Technical Reports .......................................................70

Activity Plan for 2004................................................................. 72

Performance Indicators Report .................................................. 73 (P.1) Research........................................................................73 (P.2) Research Training and Professional Education......................74 (P.3) International, national and regional links and networks:........75 (P.4) End-user links:................................................................76 (P.5) Organisational Support: ...................................................76 (P.6) Governance:...................................................................76 (P.7) National Benefit: .............................................................77

Financial Statement ................................................................... 78

ARC Centre for Complex Dynamic Systems and Control – ANNUAL REPORT 2003 2

CCCDDDSSSCCC

DIRECTOR’S REPORT 2003 has been a year of significant change in the research group. We have evolved to a new Centre based on the previous work in the Centre for Integrated Dynamics and Control (CIDAC), and incorporating some excellent researchers from Mathematics and Statistics. In addition, we have strengthened ties with our major industrial partners, BHP-Billiton Innovation and Matrikon.

We were delighted therefore to be awarded a new ARC Centre for Complex Dynamic Systems and Control (CDSC), with an expanded research team and focus. This new focus extends our work to systems with complicated dynamic behaviours; and adds both fundamental mathematical analysis of dynamics and Bayesian statistical tools. In addition, we have a range of challenging new proposed applications such as:

Integrated Mine Planning (in conjunction with BHP-BI Melbourne) which involves high dimensional discrete variable optimisation to allow optimal scheduling of mining operations.

Control of Bioenhanced Heap Leaching (in conjunction with BHP-BI Newcastle) which involves modelling and control to achieve better yield, enhanced efficiency and reduced environmental impact in this promising new mineral extraction technique.

Next Generation Process Control Tools (in conjunction with Matrikon) that meet the modern challenges of optimal control of nonlinear processes with on-line learning/estimation tools.

Despite the ‘transient’ inherent in new funding arrangements, finalizing contracts and the like, we are proud of our progress, and look forward to even better things in 2004.

Rick Middleton

Director.


CCCDDDSSSCCC

2003 HIGHLIGHTS Summer Systems and Control Workshop (Mechatronic Systems).

In conjunction with the University of New South Wales, the 2003 Summer systems and Control Workshop was held at The University of Newcastle on Monday 17 February. A range of national and international experts gave talks on tensegrity structures, systems of systems, intelligent robotics, camless internal combustion engines, vibration control and high performance control of electric machines. (see http://murray.newcastle.edu.au/cidac/workshop/new.html).

Daniel Quevedo was awarded the 2003 IEEE Conference on Decision and Control Student Best Paper Prize for his paper (co-authored with G. Goodwin and J. Welsh), “Minimizing down-link traffic in networked control systems via optimal control techniques”.

Graham Goodwin was elected as a Member of the Swedish Academy of Science.

Graham Goodwin was also invited to present the prestigious Hurwitz Memorial Lectures at ETH, Zurich in December 2003.

We are also delighted to hear the announcement that Minyue Fu has been elected a Fellow of IEEE “for contributions to Robust Control and Signal Processing”.

It was announced that Iain Raeburn has been appointed as a Member of the Australian Research Council Mathematics, Information and Communication Sciences Expert Advisory Committee.

The US based NSF has sponsored a Regional Series of Lectures on Graph Algebras which Iain Raeburn has been invited to deliver in 2004, along with a research monograph based on the lectures.

The Newcastle University Robotics Team – the Nubots – was placed 3rd in the World Competition held in Padova, Italy in July 2003. Team members were Michaela Freeston, Craig Murch, Chris Seysener, Michael Quinlan, and Graham Shanks. Team leaders were Rick Middleton and Stephan Chalup.


http://murray.newcastle.edu.au/cidac/workshop/new.html

CCCDDDSSSCCC

STAFF DIRECTOR:

Professor Richard H. Middleton

RESEARCH DIRECTOR: Professor Graham C. Goodwin

PROGRAMME LEADERS: Professor Minyue Fu Professor Kerrie Mengersen Associate Professor Reza Moheimani Professor Iain Raeburn

CHIEF INVESTIGATORS: Dr Jose A De Dona Dr Richard Gerlach Associate Professor Brailey Sims

INDUSTRY PARTNER INVESTIGATORS: Dr Salvatore (Sam) Crisafulli (Matrikon) Dr Gary A. Froyland (BHP-Billiton Innovation) Dr Andrew A. Shook (BHP-Billiton Innovation) Mr Peter M. Stone (BHP-Billiton Innovation) Dr Robert O. Watts (BHP-Billiton Innovation)

ASSOCIATE INVESTIGATORS: Associate Professor Robert E. Betz Mr Tom Honeyands Dr Robert King Dr David Pask Dr Jacqui Ramagge Mr Paul Rippon Dr Wojciech Szymanski Mr Richard Thomas Mr Steven Vandenberg

RESEARCH ACADEMIC STAFF: Dr. Greg Adams Dr. Julio Braslavsky – Chief Investigator Dr. William P. Heath – Chief Investigator Dr. Katrina Lau Dr. Maria Seron – Chief Investigator

ENGINEERING STAFF: Mr. Adrian Bastiani Mr. Frank Sobora

SUPPORT STAFF: Mrs. Dianne Piefke – Executive Officer Mrs. Jayne Disney – Administrative Assistant.


CCCDDDSSSCCC

POSTGRADUATE RESEARCH STUDENTS JUAN CARLOS AGUERO

“Novel algorithms on system identification” Supervisor: G.C. Goodwin Degree: Ph.D.

SAM BEHRENS “Self sensing piezoelectric actuators”

Supervisor: S.O.R. Moheimani Degree: Ph.D.

SIMON DODDS “Online monitoring of quality in food extrusion: A spectroscopic approach”

Supervisor: W.P. Heath Degree: Ph.D.

ANDREW FLEMING (THESIS ACCEPTED) “Synthesis and implementation of sensor-less shunt controllers for piezoelectric and electromagnetic vibration control”


HERNAN HAIMOVICH “Nonlinear constrained control”

Supervisors: G.C. Goodwin, M.M. Seron Degree: Ph.D.

DUNANT HALIM – (GRADUATED 2003) “Vibration analysis of smart structures”


LIEF HANLEN – (THESIS ACCEPTED) “Channel modeling and capacity analysis for multi-input, multi-output wireless communication systems”

Supervisor: M. Fu Degree: Ph.D.

KATRINA LAU (GRADUATED 2003) “Time domain feedback performance limitations – extensions to nonlinear, switched, and sampled systems”

Supervisor: R.H. Middleton Degree: Ph.D.


CCCDDDSSSCCC

JOSE MARE (COMMENCED 2003)

“Dynamic programming solution to model predictive control” Supervisors: J.A. De Dona, G.C. Goodwin Degree: M.E.

DAMIEN MARELLI (GRADUATED 2003) “Subband approach to system identification”

Supervisor: M. Fu Degree: Ph.D.

ZEAD MUSTAFA “Generalised metric spaces”

Supervisors: Brailey Sims, Iain Raeburn Degree: Ph.D.

TRISTAN PEREZ (THESIS ACCEPTED) “Performance analysis and constraint control of fin and rudder-based roll stabilizers for ships”

Supervisor: G.C. Goodwin Degree: Ph.D.

DANIEL QUEVEDO “Control over network communication systems”


ALEJANDRO ROJAS (COMMENCED 2003) “Dynamics of coupled oscillators systems”

Supervisors: R.H. Middleton, J.A. De Dona Degree: Ph.D.

OSVALDO ROJAS “A frequency domain approach to constrained receding horizon

control” Supervisor: G.C. Goodwin Degree: Ph.D.

IAN SEARSTON (COMMENCED 2003) “Fixed-point theory in spaces of non-positive curvature”

Supervisor: B. Sims, George Willis Degree: Ph.D.

AIDAN SIMS (SUBMITTED 2003) “C*-algebras associated to higher-rank graphs”

Supervisor: I. Raeburn, W. Szymanski Degree: Ph.D.

MARK SMITH (COMMENCED 2003) “Ultraproduct methods in fixed-point theory”

Supervisor: Brailey Sims, George Willis Degree: Ph.D.


CCCDDDSSSCCC

ARIEF SYAICHU-ROHMAN “Feedback control of input constrained systems”


JASON TYLER (COMMENCED 2003) “Graph algebras and quantum deformations”

Supervisor: I. Raeburn, W. Szymanski Degree: Ph.D.

BEN VAUTIER “Change-driven piezoelectric actuators fan for structural vibration control”

Supervisor: S.O.R. Moheimani Degree: M.E.

WEI ZHUO XIANG (COMMENCED 2003) “Connections between constrained control and estimation”

Supervisor: J. De Doná Degree: M.E.

JAMES WELSH “Ill-conditioned inverse problems arising in closed loop system identification”


ADRIAN WILLS (GRADUATED 2003) “Barrier function based model predictive control”

Supervisor: W.P. Heath Degree: Ph.D.

DARREN WOODHOUSE (THESIS ACCEPTED 2003) “An evolution in earthing system testing: Refinement of earthing system current injection testing and its analysis, with emphasis on earth potential rise estimation”


TRENT YEEND “Groupoid models for graph algebras”

Supervisor: I. Raeburn, D. Pask Degree: Ph.D.

JUAN YUZ “Fundamental limitations in estimation and control”



CCCDDDSSSCCC

ADVISORY BOARD The Advisory Board meets once per year to offer advice on matters of concern to the Centre. The current membership of the Advisory Board is:

PROFESSOR B.D.O. ANDERSON Research School of Information Science & Engineering Australian National University, Canberra, ACT.

PROFESSOR A. CAREY Mathematical Sciences, Australian National University. Canberra, ACT.

DR. S. CRISAFULLI Matrikon, Mayfield, NSW.

DR. W.J. EDWARDS Industrial Automation Services Pty. Ltd., Teralba, NSW.

MR. R. HAYES Shell Refining (Australia) Pty. Ltd., Clyde Refinery, Rosehill, NSW.

PROFESSOR W. HOGARTH Pro. Vice-Chancellor, Faculty of Science and Information Technology The University of Newcastle, Callaghan, NSW.

PROFESSOR R. JARVIS Department of Electrical and Computer Systems Engineering, Monash University, Victoria.

PROFESSOR R.J. MACDONALD Deputy Vice Chancellor (Research), The University of Newcastle, Callaghan, NSW.

PROFESSOR I.M.Y. MAREELS Melbourne University, Melbourne, Victoria.

PROFESSOR A.W. PAGE Pro. Vice-Chgancellor, Faculty of Engineering and Built Environment, The University of Newcastle, Callaghan, NSW.

MR. R. PEIRCE Technical Systems, CSR Victoria Mill, Ingham, Queensland.

DR. S.J. SMITH TUNRA Limited, Callaghan, NSW.

DR. E.H. VAN LEEUWEN Exploration and Development, BHP Billiton, Innovation, Melbourne, Vic.

PROFESSOR I.R. PETERSEN (ARC OBSERVER) Australian Defence Force Academy, Canberra, ACT.


CCCDDDSSSCCC

Visiting Members to the Advisory Board Each year, several international visitors are appointed on a casual basis to the Centre’s Advisory Board. For 2003 the visiting members were:

PROFESSOR HUGUES GARNIER Centre de Recherche en Automatique de Nancy, Université Henri Poincaré, Nancy, France.

PROFESSOR BENGT LENNARTSON Control and Automation Laboratory, Department of Signal and Systems, Chalmers University of Technology, Sweden.

PROFESSOR SIRISH SHAH Department of Chemical and Materials Engineering, University of Alberta Edmonton, Canada.

REVIEW COMMITTEE A Review Committee (made up from a representative of each Research Partner, CDSC Programme Leaders and Deputy Programme Leaders) meets quarterly to maintain contacts between industry and the Centre and review programme budgets, etc. The industrial representative are:

PETER.M.STONE BHP-Billiton Innovations, Melbourne, Vic.

TOM.A.HONEYANDS BHP Billiton, Newcastle Technology Centre, Newcastle. NSW.

RICHARD.THOMAS Matrikon, Mayfield, NSW.


CCCDDDSSSCCC

VISITORS MR. JIM CUSTERS – OCTOBER 2002 – MARCH 2003.

Department of Mechanisal Engineering, Technical University, Eindhoven, THE NETHERLANDS

PROFESSOR JAIME GLARIA - FEBRUARY & MARCH Department of Electronic Engineering, Universidad Tecnica Federico Santa Maria Valparaíso, CHILE.

DR. JEONG HEE HONG - JUNE TO AUGUST Department of Applied Mathematics, Korea Maritime University, Busan, South Korea.

PROFESSOR SOMPONG DHOMPONGSA - MARCH University of Chiang Mai, Thailand,

PROFESSOR ARIE FEUER – JULY-SEPTEMBER Department of Electrical Engineering, Technion - Israel Institute of Technology.

PROFESSOR JIM FREUDENBERG – JANUARY – APRIL Department of Electrical Engineering and Computer Science, University of Michigan

PROFESSOR HUGUES GARNIER – SEPTEMBER 2003–AUGUST 2004 Centre de Recherche en Automatique de Nancy, Université Henri Poincaré, Nancy, France.

DR. ROB GORBET - JUNE University of Waterloo, Waterloo, ONTARIO, Canada

PROFESSOR ROLF JOHANSSON – JUNE - AUGUST Department of Automatic Control, Lund University, Sweden.

PROFESSOR OH-KYU KWON - FEBRUARY School of Electrical and Computer Engineerig, Inha Universiry, Incheon, Korea

PROFESSOR BENGT LENNARTSON – SEPTEMBER 2003 – APRIL 2004 Control and Automation Laboratory, Department of Signal and Systems, Chalmers University of Technology, Sweden.

PROFESSOR DAVID MAYNE – DECEMBER Department of Electrical Engineering, Imperial College of Science and Technology, London, UK.

MR. DOMINIK NIEDERBERGER – MARCH Automatic Control Laboratory, ETH – Swiss Federal Institute of Technology, Zurich, Switzerland

PROFESSOR JOHN QUIGG – JANUARY – APRIL Department of Mathematics, Arizona State University, from January to April.

MS. ANNA ROSEN – AUGUST 2003-FEBRUARY 2004 Department of Electrical Engineering, Linkoping University, Sweden.


CCCDDDSSSCCC

PROFESSOR BOB SKELTON – FEBRUARY Department of Mechanical and Aerospace Engineering University of California, San Diego, USA.

PROFESSOR ANNA STEFANOPOULOU – FEBRUARY Powertrain Control Laboratory, Mechanical Engineering Department, University of Michigan, USA.

PROFESSOR HITOSHI TAKATA - MARCH Department of Electrical and Electronics Engineering, Kagoshima University, Japan.

MR. HENRIK TIDEFELT – AUGUST 2003-FEBRUARY 2004 Department of Electrical Engineering, Linkoping University, Sweden.

PROFESSORT YUFENG ZHENG – AUSUT 2003 – AUGUST 2003-FEBRUARY 2004 Department of Electrical and Electronic Engineering, The University of Melbourne.


CCCDDDSSSCCC

SHORT COURSES/WORKSHOPS: Summer Systems and Control Workshop – Mechatronics: Each

year, the University of New South Wales and CDSC jointly organise a single day symposium with high level national and international speakers, on a specific topic related to systems and control. In February 2003, CDSC hosted this successful event at the University of Newcastle.

MPC Mini Symposium: During 1 December 2003 a mini symposium on model predictive control was organised. This included a range of speakers from both the University and Matrikon, and included a discussion session at the end to look at future directions for the Maitrkon Next Generation Process Control project (see page 53 for further details).

Workshop on Constrained Control & Estimation: This workshop was presented twice in the year at the University of Western Ontario (5 May) and at the University of Alberta (7 May).

IEEE Advanced Process Control Workshop: A tutorial workshop was presented at IEEE Advanced Process Control Workshop in Vancouver, Canada on 30 April.


CCCDDDSSSCCC

SEMINARS: Research students and staff from The University of Newcastle as well as Australian and international visitors participate in the Centre’ s seminar series. Seminars presented in 2003 (not including other general Electrical Engineering, Mathematics or Statistics seminars) are:

29 January Arief Rohman – Research student

“Multivariable algebraic loops in static anti-windup schemes and MPC”.

10 February Jim Freudenberg – University of Michigan, USA

“Fundamental design limitations of the general control configuration”.

26 February Anna Stefanopoulou – University of Michigan

“Control of fuel cell systems”.

17 March Terry Summers – Research Student

“Instantaneous real and imaginary power applied to the control of induction machines”

21 March Dominik Niederberger – ETH Swiss Federal Institute of Technology, Switzerland

“Active vibration control via shunted piezoelectric materials”

26 March Weizhou Su – Research student

“Performance limit and robustness of linear systems in tracking sinusoidal signals”

8 April Katrina Lau – Research student

“Time domain feedback performance limitations – Extensions to nonlinear-switched and sampled systems”

7 May Mirek Pawlak – University of Manitoba, Canada; Macquarie University.

“On non-parametric identification of multi-channel systems”


CCCDDDSSSCCC

21 May Minyue Fu – CDSC

“Control of linear systems using quantized feedback”

11 June Rob Gorbet - University of Waterloo, Canada

“An introduction to shape memory alloy actuators”

1 July Rob Gorbet – University of Waterloo, Canada

“Comments on regulation of actuators with hysteretic behaviour”

9 July B. Barmish – University of Wisconsin - Madison, USA

“On distributional robustness and Monte Carlo simulation”

11 July Rolf Johansson – Lund University, Sweden

“Observer-based strict positive real (SPR) feedback control system design”

22 October Rick Middleton – CDSC

“Feedback stabilization over signal to noise ratio limited channels”

27 October Peter Gawthrop – University of Glasgow, UK

“Indirect estimation of physical parameters within linear systems”

5 November Bob Bitmead – University of California, San Diego, USA; IEEE Distinguished Lecturer

“Model predictive control, state estimation and coordinated vehicles”

14 November Will Heath – University of Newcastle

“Non-parametric estimates for linear dynamic systems”

27 November Jonathon Keith – University of Queensland

“Bayesian algorithms for a novel sequencing technology”


CCCDDDSSSCCC

RESEARCH PROGRAMMES: A. Control System Design

Programme Leader: Graham Goodwin

Deputy Leader: Maria Seron

Programme Goals: Control System Design is a mature discipline. Surprisingly, however, the existing methodologies tend to be limited to relative standard problems – e.g. linear, unconstrained and with centralized architectures. As soon as one departs from these settings one soon finds oneself faced with severe difficulties. Unfortunately many real world problems fall into these, so called, “complex” problems. These problems include such features as nonlinear and non-smooth behaviour, high state dimension and lack of convexity. This programme of research is aimed at addressing these issues using alternative theoretical tools and in the context of modern computational methods.

A.1 Implementation of Model Predictive Control

Researchers: Jose De Dona, Graham Goodwin, Will Heath, Hernan Haimovich, David Mayne, Tristan Perez, Maria Seron, Adrian Wills

We have recently obtained exciting results on the connections between anti-windup control and MPC and on the geometric structure of receding horizon control for constrained linear systems. The latter allows for high-speed implementations of controls as off-line computation of the explicit solution replaces on-line optimisation. Explicit solutions to this problem were studied in parallel by other researchers such as Bemporad, Johansen and others. During 2003, substantial effort has been devoted to completing a comprehensive book on these topics. Details of individual topics are given below:

Geometry of Model Predictive Control

Model predictive control (MPC) has been widely adopted by industry as a means of dealing with constraints in control system design. However, the standard algorithms used for this problem require that a constrained optimization problem be solved, on-line, at each time step. This has restricted the application to low order systems or systems with long time constants. We have been studying methods for “speeding up” the on-line calculations needed to implement MPC. In particular, we have studied “closed form” solutions which can be implemented via a look-up table.


CCCDDDSSSCCC

This parallels work due by other research groups elsewhere. We have also studied the associated numerical issues and compared computational times for various methods. Related ideas have also been exploited by us to obtain results on enlarged terminal sets associated with stability of MPC algorithms.

Neural Approximation to the Explicit Solution of Constrained Linear MPC

This project has been motivated by the above work on explicit MPC. In particular, we have studied the use of Neural Networks as a way of approximating the explicit (piecewise affine) solution to the control policy for constrained linear systems.

Fast Algorithms for Constrained Control

As discussed previously, we have been studying alternative algorithms for obtaining fast on-line solution to QP problems. This has been motivated by very large dimensional systems, such as cross directional control problems, see Project A.3. We have developed a novel scheme for solving constrained control problems based on working in the singular value space of the system Markov parameter matrix. We have found this to lead to a very computationally efficient algorithm which is very close to optimal. On-going work includes a study of stability and robustness properties.

Interior Point and Barrier Function based MPC

Adrian Wills received his PhD for his thesis entitled “Barrier Function Based Model Predictive Control”. The work included the development of interior point algorithms for model predictive control, the stability analysis of state feedback model predictive control, and the practical implementation of output feedback model predictive control to an industrial plant. Papers concerning each aspect of this work have been submitted to journals. The implementation of the controller to an edible oil refining plant has also been completed and an industrial partner commissioned a final implementation of the controller in December 2003.

Dr Wills worked with the Centre as a post-doctoral researcher for a period in 2003. During that period we devised a novel strategy for handling soft constraints. A tutorial paper was also written on the state-of-the-art implementation of interior point methods for model predictive control.


CCCDDDSSSCCC

A.2 Analysis of Optimal and Constrained Control Systems

Researchers: Graham Goodwin, Hernan Haimovich, Rick Middleton, Tristan Perez, Maria Seron, Arief Syaichu-Roman

The widespread use of model predictive control and other techniques for constrained control gives rise to a number of issues in analysing and understanding such schemes. A range of topics are currently being studied in this area to advance our understanding and application of constrained and optimal control algorithms.

Output Feedback Issues in Model Predictive Control

Most theoretical studies of model predictive control assume that the full state of the plant is available. However, in practice, one usually only has available noisy output measurements. A common strategy is to estimate the states using some form of observer and then act as if the estimated states were the true states. (This is usually called Certainty Equivalence.) This project has studied the optimality of Certainly Equivalence (and related methods) in the control of constrained stochastic linear systems.

Singular Value Structure of Operators in Quadratic Optimization Problems

This research is aimed at gaining a better understanding of the singular structure for sampled data LQ problems of continuous linear systems, particularly the asymptotic behaviour when the sampling rate is increased. We have shown that there is a natural convergence between the finite set of singular values of the discrete time problem, and the (infinite) countable set in continuous time. Convergence of the singular structure for a particular example is illustrated below in Figure A.1.

Figure A.1: Convergence of the singular values of G∆


CCCDDDSSSCCC

Anti-windup and Model Predictive Control

We have considered a range of problems related to Anti-windup schemes and their close relatives, model predictive control. Based on our previous observation of the close link between anti-windup and model predictive control, we have been able to propose some iterative algorithms for solving MPC problems. These algorithms offer guaranteed linear convergence (including convergence rates); and by using primal/dual problem formulations, certificates of performance can also be established. Compared to other algorithms for solving QP problems, they have a very simple structure, with much simpler computations during each iteration. However, because of the linear convergence, the number of iterations required is typically larger. Example computational studies indicate that the computational cost is of the same order of magnitude as interior point or active set methods we have studied.

We have also examined the extension of optimisation based anti-windup design to the discrete time case. A fairly standard translation of previous continuous time algorithms is possible. In addition, however, we are able to generate a structure (see Figure A.2) which enforces a strictly proper formulation of the compensation, which gives us an explicit anti-windup scheme. LMI based synthesis of the anti-windup compensation is possible, with results showing little reduction in control performance compared to implicit anti-windup schemes, whilst having a dramatic reduction in the computational burden. These ideas have been tested in simulation on a process plant model supplied by Electricite de France (see for example Figure A.3). Additional practical problems of model uncertainty, saturation level uncertainty, and poor nominal control tuning need to be considered in this problem.

+ uu +

−

2Λ−

+

+

v1ζ

+

ex

e1−− )AzI(C CC

1Λ−

CB

CD

+dzΨ

1−z

+ u

Figure A.2: Block diagram of explicit Antiwindup

Compensation


CCCDDDSSSCCC

Figure A.3: Example Antiwindup responses for EDF Example: Left – without compensation, Right – with compensation

A.3 Cross Directional Control

Researchers: Iwan Akkermans, Tino Domanti, Graham Goodwin, Will Heath, Osvaldo Rojas

We have developed a novel MIMO anti-windup strategy for cross-directional control of web forming processes. The scheme relies on a basis function expansion of the system interaction matrix which allows one to analyse the problem in terms of the spatial frequency components of the web profile. Anti-windup protection is achieved by prioritizing the control directions in which the available control authority is employed. In particular, disturbance compensation is first sought in those spatial frequency components associated with the largest gain, that is, where the required control effort is smaller. This is closely related to the general algorithm described in the project on fast algorithms for constrained control (see pp 16). Special features of the cross-directional control problem include high levels of interactions and large numbers of inputs and outputs (typically several hundred). Stability of the scheme has been analysed for the case in which complete knowledge of the input and output rotation matrices is assumed and when there is gain uncertainty. Simulation results show that the steady-state performance obtained with the proposed strategy is superior to that achieved with traditional MIMO anti-windup schemes and is close to an optimal QP solution to the problem.

In another stream of this project we have proposed a novel controller for the cross-directional control problem. This combines the simple tuning of internal model control with optimal steady state performance (achieved via quadratic programming). Iwan Akkermans, visiting from Eindhoven


CCCDDDSSSCCC

University of Technology, studied the robustness properties of this and related controllers.

A.4 Fundamental Design Trade-offs

Researchers: Julio Braslavsky, Arie Feuer, Jim Freudenberg, Graham Goodwin, Rick Middleton, Tristan Perez, Mario Salgado, Maria Seron, Juan Yuz

The understanding of design trade-offs and associated fundamental design constraints is intrinsic to any design problem. It underlies estimation, signal processing, control system design, telecommunications and indeed all areas where complex dynamic systems arise. This topic has been well studied for low dimensional linear time invariant systems. Our aim has been to extend these results to new areas including nonlinear, switched, constrained, sampled data, multirate and infinite dimensional systems. Several topics in this project have been investigated in 2003:

Fundamental Constraints in Control over SNR limited Channels

A particular topic currently under investigation is the role played by communication constraints in networked control. Previous work by Nair, Evans and other researchers has examined the problem of stabilisation of an unstable system over a bit rate limited communication channel. In our work, we take a further step back, and examine stabilisation question for open loop unstable plants over a channel with a signal to noise ratio limitation. As a simple starting point, we consider the additive white Gaussian noise (AWGN) case, and show that there is indeed a limitation on the ability to stabilise. In particular, for a given noise level, and plant, there is a minimum power required to be transmitted that permits stabilisation. This problem can be most simply formulated in the case of a minimum phase plant, using H2 optimisation techniques. Furthermore, when compared to the bit rate limit for the same channel implied by Shannon theory, our results are equivalent to those based on bit-rate limitations. Extensions of this to stabilisation of plants which have both unstable open loop poles, and non-minimum phase zeros has been achieved by utilising periodically time varying feedback structures, such as Deadbeat state observers together with discrete time feedback, as illustrated in Figure A.4.


CCCDDDSSSCCC

Figure A.4: Feedback Stabilization over a SNR Limited Channel

Limits Due to Constrained Control

Whilst performance bounds have been extensively studied for linear control and estimation problems, there has been very little work done on performance bounds arising from actuator constraints. In this work we have made a first step toward quantifying the effect of actuator limits on the performance of control systems. Specifically, we have evaluated the cost associated with constraints via the L2 norm of the tracking error of constrained limiting optimal regulator.

Effect of Uncertainty on Control Law Performance

Our goal has been to gain a better understanding of fundamental performance limits for feedback control systems. In the literature to date on this topic, all available results assume that the designer has an exact model of the plant. Heuristically, however, one would expect that plant uncertainty will play a significant role in determining the best achievable performance. Our goal has been to quantity performance limits of linear feedback control systems in the presence of plant uncertainty. We have formulated the problem by utilizing stochastic embedding of the uncertainty. The results allow one to evaluate the best average performance in the presence of uncertainty. They also allow one to judge whether uncertainty or other properties, e.g., non-minimum phase behaviour, are dominant limiting factors.

Related work has also been carried out in an effort to understand the fundamental limits that apply to adaptive control laws and to quantify the performance loss due to the use of decentralized architectures.


CCCDDDSSSCCC

A.5 Finite Alphabet Systems

Researchers: Graham Goodwin, Daniel Quevedo, Osvaldo Rojas, Maria Seron

On-off and relay feedback control systems are widespread and have been studied extensively in the literature. An interesting generalization corresponds to the finite control set case. Here, the constraint set is allowed to contain only a finite number of elements. This situation arises in many practical contexts, e.g. when a finite number of control levels are available. The same problem arises when considering digital control systems affected by quantization and numerical overflow. In addition, finite alphabet control laws form a precursor to hybrid systems. They also have close links to the issue of control with communication constraints and predictive analog to digital data converters such as the Σ∆-Modulator.

We have studied the discrete time receding horizon quadratic control problem with a finite alphabet. The approach is in the spirit of recent results on Model Predictive Control, which are aimed at providing insights into the nature of the control law via an elucidation of the inherent structure of the mapping between plant states and optimal controls. In particular, for the case of constraints (e.g. saturation), one obtains a

characterization in terms of a partition of the state space into polytopal regions in which the control input is piece-wise affine in the state.

∞

Our key result has been to show that a similar structure holds for the case of control with a finite constraint set. Moreover, there exists a close connection between the partitions induced by the finite alphabet and the

-constrained solution. We have also studied the dynamics arising from

the use of controls derived from a finite alphabet. ∞

We have also studied many applications of these ideas including the following topics:

Receding Horizon Networked Control

We have designed a control strategy for multi-variable plants where controller and actuators are connected via a digital data-rate limited channel. In order to minimize bandwidth utilization, a communication constraint is imposed, which restricts all data transmitted to belong to a finite set and only permits one plant input to be addressed at a time. We have developed a new scheme, called Receding Horizon Networked Controller, which aims at optimizing quadratic performance under the above communication constraint. A key aspect of this contribution has


CCCDDDSSSCCC

been the application of the control scheme to a real laboratory-scale system.

Some experimental results obtained from the implementation to a five tank level control problem are described below. We transmit a total of only 4 bits per second to control all 5 tanks.

Once the tanks have reached the desired level (0V) after the initial start-up, an unmeasured inflow disturbance was introduced into Tank 2 at 35 seconds and an extra outflow valve was opened on Tank 3 at 1090 seconds.

Figure A.5 to Figure A.7 show the control increments sent by the controller to each plant. The control signal applied to the plant by the actuator is also shown together with the level measurement from the tank.

It can be seen from these figures that, when the disturbance occurs in Tank 2, most downlink bandwidth is dedicated to the control of level in this tank. Also, it is easily observed that the controller pays attention to the other tanks when the measured level in tank 2 approaches the desired level at around 210 seconds into the experiment. The same effect can be observed when tank 3 is disturbed.

This aspect of sharing attention is also apparent in the histograms contained in Figure A.8. Most control attention ( )1iu∆ = ± was applied to

those tanks, for the given time period, to which the disturbances occurred.

Figure A.5: Response for Receding Horizon Networked Control – Tanks 1 & 2

(Top: Transmission; Middle: Control Signal; Bottom: Output)


CCCDDDSSSCCC

Figure A.6: Response for Receding Horizon Networked Control – Tanks 3 & 4

(Top: Transmission; Middle: Control Signal; Bottom: Output)

Figure A.7: Response for Receding Horizon Networked Control – Tank 5 (Top: Transmission; Middle: Control Signal; Bottom: Output)

Figure A.8: Histogram of Control Increments


CCCDDDSSSCCC

Estimation using Filters Having Quantized Coefficients

A common requirement in digital signal processing is that the filter should be implemented with restricted coefficients. For example, if powers of 2 are used, then multiplication can be achieved by shifting and addition. However, this raises the question of what constitutes the best set of filter coefficients given that they need to be drawn from a finite alphabet of coefficients. It turns out that this question can be converted into an optimal control problem where the input is drawn from a finite alphabet. We are currently studying this problem using various cost functions, e.g. H2 and H∞ norms.

Digitization of Audio Signals

We have formulated the problem of quantization of audio signals as a deterministic finite set constrained quadratic optimization problem. This has led to a new scheme for quantization which we have called a Moving Horizon Optimal Quantizer (MHOQ).

The MHOQ includes a model of the ear-sensitivity to low level noise power and minimizes directly the perceived error over a finite prediction horizon. Feedback is incorporated by means of the moving horizon principle. With a prediction horizon N equal to 1, the MHOQ reduces to the psycho-acoustically optimal noise shaping quantizer, widely used in practical applications. We have shown that the performance achieved using N>1 is better than that traditionally achieved with the optimal noise shaping quantizer.

A.6 Duality of Constrained Control and Estimation

Researchers: Jose De Dona, Graham Goodwin, Maria Seron, Wei Zhou

The aim of this project is to investigate the implications of duality and other connections between constrained control and estimation. We believe that the research will result in a richer understanding of these problems. In particular, we envisage an impact in at least three areas: (i) Computational issues, i.e. development of more efficient algorithms for constrained problems. (ii) Geometry of constrained problems by extending recent results pertaining to constrained control to estimation problems. (iii) Problems with mixed constraints, for example, interval and finite set constraints.

The relationship between linear estimation and linear quadratic control is well known in the unconstrained case. Since the original work of Kalman and others, many authors have contributed to further understand this


CCCDDDSSSCCC

relationship. For example, Kailath, Sayed and Hassibi have explored duality in the unconstrained case using the geometrical concepts of dual bases and orthogonal complements. The connection between the two unconstrained optimisation problems using Lagrangian duality has also been established in, e.g., the recent work of Rao.

However, to the best of our knowledge, the duality between estimation and control remains an open question in the constrained case. We have derived the Lagrangian dual of a constrained estimation problem and shown that it leads to a particular nonlinear optimal control problem. We then show that the primal constrained estimation problem has an equivalent formulation as a nonlinear optimisation problem, exposing a clear symmetry with its dual.

A.7 Development of Virtual Laboratory for Control System Design

Researchers: Adrian Bastiani, Graham Goodwin, Jim Kusters, Rick Middleton, Frank Sobora, Bob Skelton

Our aim in this project is to capture some of the experiences we have gained from industrial control problems in the form of virtual laboratories for students. Currently, 6 laboratories are essentially complete. These are: i. Control of a continuous casting machine ii. Thickness control in a rolling mill iii. Dynamics a rocket iv. Control a rocket v. Control of a servo system vi. Control of coupled tanks

The following additional laboratories are currently under development vii. Control of a large radio antenna viii. Quantization of Audio via feedback ix. PH control


CCCDDDSSSCCC

Figure A.9: Example screens from virtual laboratories

These virtual laboratories are not intended to fully replace hardware based experiments. However, we believe that they can serve a significant

role in particular cases, e.g. as supplements with large classes, for short courses with industry, and as a mechanism for exposing students to real world problems which would be otherwise infeasible in a teaching environment.

Figure A.10 Satellite Tracking Virtual Laboratory


CCCDDDSSSCCC

Figure A.11 Rocket Control (left) and Dynamics (right) Virtual Laboratories

A.8 Automotive Power Train Control (General Motors, USA)

Researchers: Jose De Dona, Graham Goodwin, Hernan Haimovich, Rick Middleton, Maria Seron, James Welsh

Other Funding: General Motors Research

The automotive industries abound with complex dynamical problems. This project will focus on adaptive power train control, with the aim of reducing emissions, improving drivability and achieving greater efficiency. The project is fully funded by General Motors Research (USA) with a budget of approximately $110,000 per year.

One initial problem that we have focused on includes state estimation using the (highly nonlinear) exhaust gas oxygen sensor. In particular, we have studied particle filtering and related, more ad-hoc., methods. We have also studied combined state and parameter estimation for air/fuel control and have developed novel nonlinear control laws for this application.


CCCDDDSSSCCC

B. Mathematical Systems Theory Programme Leader: Iain Raeburn

Deputy Leader: Jose De Dona

Programme Goals: The object of the Programme is to investigate mathematical models of dynamic systems which exhibit complex behaviour, exploiting the expertise of the CI’s in modern functional analysis. The research will proceed in two broad directions, plus problems driven by a particular application.

B.1 Graph Algebras: Operator Algebras We Can See

Project leader: Iain Raeburn

Other Researchers: David Pask, Wojciech Szymanski

Other support: National Science Foundation (USA)

Directed graphs are simple mathematical structures which are used to model networks and Markov chains. When the network is large or infinite, Hilbert-space representations of the graph provide a powerful tool for analysing the long-term behaviour of the network; the C*-algebra of the graph provides a universal object for studying these Hilbert-space representations. Over the past decade, researchers have built up an elegant theory which describes the structure of the algebra in terms of the behaviour of cycles in the graph. Researchers in the Functional Analysis group at Newcastle have played a leading role in the study of graph algebras, but now many other groups all over the world are actively working in the area; there are several groups in each of Japan, Korea, the USA and Canada, for example.

The broad interest generated by this research has recently led the American NSF to commit funding for a Regional Conference on the subject, which will be held at the University of Iowa from 31 May to 4 June, 2004. The central feature of these Regional Conferences is a series of ten lectures delivered by a leading expert; in recognition of the pioneering role of the Newcastle group, Raeburn has been invited to give this series of lectures. Part of the agreement is that he will produce a research monograph based on the lectures, to be published by the American Mathematical Society. So a major focus for him in 2004 will be to write this monograph. In the meantime, of course, researchers in the Centre will continue to work on a variety of other projects involving graph algebras.


CCCDDDSSSCCC

B.2 Groupoid Models


Other Researchers: Cynthia Farthing (Iowa), Paul Muhly (Iowa), Trent Yeend (PhD student),

There are several ways in which one can approach the theory of graph algebras, and the analysis of graph algebras provides good test problems for the general theories which operator algebraists use to analyse crossed-product-like structures. The most potent of these general theories is based on groupoids, which provide an algebraic framework intermediary to the directed graph and the C*-algebra.

Groupoid models were used in the first analyses of the C*-algebras of infinite graphs in the mid-nineties, but these models were constructed in an ad hoc fashion. More recently, Paterson has developed a procedure for building groupoid models and has applied it to infinite graphs. The object of this project is to further develop Paterson's procedure and to apply to it to more general combinatorial structures, such as the higher-rank graphs of Kumjian-Pask and the continuous graphs of Deaconu and Katsura.

B.3 Higher-Rank Graphs


Other Researchers: David Pask, John Quigg (Arizona State U.) Aidan Sims (Research Assistant)

Higher-rank graphs are higher-dimensional analogues of the directed graphs used to model networks and Markov chains. They were introduced by Kumjian and Pask in 2000 as models for the operator algebras associated to actions of groups on trees and buildings. (In this context, the word "building" refers to one of a family of higher-order combinatorial structures; these are central objects in the study of algebraic groups, which is itself a huge and very active area of contemporary mathematics with deep links all over the discipline.) As for ordinary directed graphs, each higher-rank graph has a graph algebra which in universal for Hilbert-space representations of the graph.

Many of the properties of ordinary graph algebras have been extended to the algebras of higher-rank graphs, but in some respects the algebras of higher-rank graphs behave quite differently. Even for ordinary directed graphs, the ideal structure of graph algebras has only recently been completely understood (Hong-Szymanski, in press). In his thesis (submitted in November 2003), Sims has tackled these differences head


CCCDDDSSSCCC

on, and has made a good start on understanding the ideal structure of the algebras of higher-rank graphs. In this project, he will continue to work on this problem, and efforts will be made to understand the topological invariants of the algebras of higher-rank graphs.

B.4 Identification of Graph Algebras

Project leader: Wojciech Szymanski

Other Researchers: Jeong Hee Hong (Busan, S. Korea), David Pask, Iain Raeburn

Quite recently, major progress has been achieved in understanding of the basic structural invariants of the C*-algebras of infinite graphs, and in particular their ideal structure has been completely determined (Hong-Szymanski and others). This breakthrough combined with the previously obtained description of their K-theory yields a natural springboard for the beginning of a classification of these algebras.

This project concerns the possible classification of graph algebras. This is a vast programme, and a general solution is beyond reach at present. The present project will focus on tractable subclasses of special importance, such as the class of non-simple purely infinite graph algebras. It is likely that progress on the classification programme will lead to a better understanding of the behaviour of symbolic dynamic systems associated with infinite graphs.

B.5 Symbolic Dynamics and Operator Algebras

Project leader: David Pask

Other Researchers: Teresa Bates (UNSW), Wojciech Szymanski, Iain Raeburn

A symbolic dynamical system consists of a collection of infinite sequences of symbols from an alphabet, together with a shift map which moves the sequences from right to left. A special class of symbolic dynamical system consists of the subshifts of finite type, which are closely related to Markov chains. A subshift of finite type may be modelled by the infinite paths in a directed graph, and the graph algebra of this directed graph encodes the complex dynamics of the associated subshift of finite type. Many dynamical properties of a symbolic dynamical system are reflected in the structure of the associated graph algebra.

In this current project we seek to extend and expand this close relationship between symbolic dynamical systems and graph algebras. An arbitrary symbolic dynamical system over a given alphabet may be


CCCDDDSSSCCC

represented by a certain directed graph which carries a labelling by elements from the alphabet.

Our project aims to formulate the properties of a labelled graph algebra which subsume the properties of an ordinary graph algebra, and to use this structure examine how the properties of the symbolic dynamical system manifest themselves in the labelled graph algebra.

B.6 Coupled Oscillators

Project leader: Rick Middleton

Other Researchers: Jose De Dona, Alejandro Rojas

Motivated by a number of application areas, including biophysical systems, we wish to study properties of large interconnections of oscillatory dynamic systems. In particular, we consider systems where each ‘cell’ has a tightly defined oscillatory behaviour, the dynamics of each cell are similar, and where comparatively weak coupling occurs between cells. Our aim is to understand at a fundamental level the behaviour of such systems, and to develop tools to aid in their analysis.

B.7 Differential Algebraic Control

Project leaders: Jacqui Ramagge, Maria Seron Other Researchers: Udo Baumgartner, Jose De Dona, Hernan

Haimovich (PhD student), Iain Raeburn, Aidan Sims

This project is aimed at the investigation of the property of flatness of nonlinear control systems. It involves using differential algebra to characterise some control systems.

A common way of describing control systems is via a set of (linear or nonlinear) differential equations. In the simplest cases these involve two types of variables called the states and the inputs. The states are functions of time that represent the internal behaviour of the system, whereas the inputs are functions of time that we may alter in order to direct the behaviour of the system. When the inputs are selected as functions of the states, they constitute a feedback.

An essential property of linear control systems is that of controllability; this corresponds to being able to obtain a specific state behaviour by adjusting the inputs. Controllability of linear systems can be characterised using concepts from linear algebra. For nonlinear systems, some types of controllability can be associated with the property of flatness. From a “feedback system” perspective, flat systems are


CCCDDDSSSCCC

equivalent to linear ones via a special type of feedback called endogenous feedback. In other words, a flat system is a system that is “linearisable” via endogenous feedback. Much of the initial development of the concept of flatness by Fliess et al in the 1980's took place within the framework of differential algebra. Within this context, a system is flat if and only if it is equivalent to a purely differentially transcendental system.

This is a new area of research in Newcastle. We have initiated links with international experts, including Yufan Zheng, from the University of Melbourne. Prof. Zheng was invited to Newcastle in August 2003 to give a series of lectures on differential algebraic approach to nonlinear control systems.

We have made some progress at understanding the above concepts both from a mathematical and engineering perspective. In particular, we have identified some of the open problems of the topic, which include finding a systematic characterisation of flat systems and establishing whether or not a linearisable system via any kind of feedback (not necessarily endogenous) is flat. Our aim is to tackle some of these open questions while continuing to gain insight into the area by cross-fertilisation of mathematical and control engineering knowledge.

B.8 Nonlinear Analysis and Fixed-Point Theory

Project leader: Brailey Sims

Other Researchers: Tomas Benevides (Seville), Sompong Dhompongsa (Chiang Mai, Thailand), Enrique Llorens Fuster (Valencia), Art Kirk (Iowa), Zead Mustafa (PhD student), Ian Searston (PhD student), Mark Smith (PhD student).

Equilibria for discrete and continuous dynamical systems correspond to fixed points of nonlinear maps on infinite dimensional function spaces. The solution of nonlinear optimization and control problems lead to variational inequalities and hence ultimately to fixed points of a related nonlinear operator. When the system is conservative, the nonlinear mapping is nonexpansive with respect to an appropriate metric on the underlying function space. The convergence and ergodic structure of orbits and various iterative schemes, such as those of Ishikawa, relate to the stability and long-term average behaviour of the system.

The project continues to further our fundamental understanding of nonexpansive mappings, with an emphasis on identifying easily applied, yet widely applicable, criteria that ensure the existence of fixed points for


CCCDDDSSSCCC

such maps together with effective algorithms by which they can be approximated. Special emphasis is given to the more difficult cases, where the underlying space lacks nice geometric structure such as that exhibited by a Hilbert space, for example; or where linear structure is absent and so the usual tools of convex analysis are inapplicable. This work embodies the following three topics: Analysis in Hyperbolic Metric Spaces

In many situations the underlying space lacks linear structure. In this context, the existence and stability of fixed points for nonexpansive type depends on the special geometric properties of the space. Hyperconvex metric spaces; for example, the unit ball of l∞and R-trees, and classical hyperbolic n-space H∞ are examples of spaces on which nonexpansive mappings have nonempty fixed point sets. Recently it has been observed that geodesic metric spaces which are hyperbolic in the sense of Gromov, in particular the so called CAT(0) spaces, encompass many of these examples and provide a very general setting in which a rich fixed point theory paralleling that for Banach spaces seems possible. An important start has been made toward the development of such a theory. This project continues that development. Early in 2004 Professor Art Kirk, a pioneer in this development will visit CCDSC and deliver a series of lectures. That such spaces may provide a fertile setting in the absence of linearity for other problems arising in system dynamics and optimal control is also being investigated.

Generalized Metric Spaces

Classical metric fixed point theory relies on being able to measure the proximity of two point in the underlying space; typically, by a metric d(x,y). However, many results only rely on having some measure the relative proximity of three consecutive iterates of the mapping in question, D(x, Tx, T2x). Generalized metrics D(x,y,z) which provided such a measure were introduced in the 1990's, however the concept proved to be seriously flawed. A viable notion has been identified and a fruitful theory is under development.

Applications of Ultra-Methods to Nonlinear Analysis

Banach space ultra-products, and more recently ultra-powers of metric spaces, have proved a powerful and important tool in both linear and nonlinear analysis. And represent a common meeting ground between standard and non-standard analysis. They have proved the major tool for establishing the existence of fixed points for nonexpansive and asymptotically nonexpansive maps in the absence of normal structure, and are particularly suited to the analysis of attainment problems. By


CCCDDDSSSCCC

lifting the problem to an ultra-power approximate solutions become a solution. For example, an approximate eigenvalue of a linear operator corresponds to an eigenvalue of the lifted operator. The further development of these techniques and their potential application to a wide variety of problem is the focus of this project. The first half of 2004 will see the completion of a research monograph (to be published by Springer) on this topic by the project leader and one of his students, Mark Smith.


CCCDDDSSSCCC

C. Bayesian Learning Programme leader: Kerrie Mengersen

Deputy Leader: Will Heath

Programme Goals: The Bayesian Learning programme comprises researchers from both Engineering and Statistics backgrounds, reflecting the strong interdisciplinary nature of the Centre. Five main areas of research are identified in the programme: system identification, Markov Chain Monte Carlo (MCMC) methods, Bayesian modelling, nonlinear and mixture modelling, and robot location and vision.

C.1 Markov Chain Monte Carlo (MCMC) Project:

Project Leader: Kerrie Mengersen

Other Researchers: Richard Gerlach, Robert King, Darfiana Nur

Figure C.1: The movement of MCMC algorithms over multi-modal likelihoods

The analysis of complex systems requires computational algorithms for solving high-dimensional integrals. These occur naturally in Bayesian modelling, where interest focuses on estimation of posterior expectations and determination of posterior distributions of parameters of interest. The MCMC Project focuses on the development of efficient simulation methods for this purpose. A review of available MCMC methods is to be published in The Encyclopedia of Biostatistics in 2004. New MCMC algorithms, and new methods of assessing their performance in terms of convergence, have been developed and evaluated. It is anticipated that this project will feed directly into other projects and programmes that require advanced computational methods.


CCCDDDSSSCCC

MCMC Algorithms

Population Monte Carlo algorithms have potential to better cover complex, high-dimensional, multi-modal parameter spaces. These algorithms combine the benefits of parallelised computing with Markov chain Monte Carlo methods. A new algorithm, the Pinball Sampler, has been developed as part of this project. The algorithm comprises population Monte Carlo, adaptive rejection sampling and a repulsive mechanism that discourages clustering of particles around a local mode. The algorithm has been applied to a high-dimensional hidden Markov model (HMM) problem that has recently been addressed with Bayesian particle filters. This work has led to the publication of an invited book chapter and international invited presentations.

MCMC Convergence

As with most numerical solutions, the issue of convergence is important in MCMC analysis. Questions of ‘burn-in’, adequate exploration of the parameter space and the number of iterations required for estimation must be addressed. A new diagnostic for assessing convergence has been developed as part of this project. This approach uses phase randomisation, or Fourier bootstrapping, to assess the behaviour of the third cumulant and make inferences about the stationarity of the chain. This is shown to complement existing diagnostics that assess first-order stationarity properties alone. The project has led to the submission of two papers to refereed journals in 2003; one of these is now accepted subject to revision and will appear in 2004.

C.2 Bayesian Modelling

Project Leader: Kerrie Mengersen

Other Researchers: Richard Gerlach, Graham Goodwin.

The overall aim of this project is to develop Bayesian theory and methodology as a mechanism for updating of opinion or state. Focus will be on semi-parametric hidden Markov models (HMM) via Markov chain Monte Carlo (MCMC). Particular problems include spatio-temporal models, choice of states in HMM, clustering and loss functions to enhance multi-modal exploration, elicitation of expert opinion to inform priors, and perfect sampling and convergence in MCMC. Links with other projects include constrained estimation, inverse problems, optimisation based operator guidance systems and robust control for spatially distributed systems.


CCCDDDSSSCCC

C.3 Control Oriented System Estimation

Researchers: Minyue Fu, Richard Gerlach, Graham Goodwin, Will Heath, Bob Skelton, James Welsh

An important class of learning and adaptation tools are known by the generic title of system estimation methods. These tools are well understood for linear systems with idealised experimental conditions. In practice experiments are often performed with limited excitation, necessitated by both physical constraints on actuators and economic constraints on performance. Similarly, experiments are often performed on plants running in closed-loop; in this case the control action constrains the signal space of the excitation signal. The aim of this project is to better understand the implications of such constraints, and to develop novel analysis tools for estimation under such circumstances. We are also interested in developing tools for non-linear and spatially distributed systems. Within this research project several topics have been studied:

Closed Loop Identification

Closed look identification has been a topic of interest for several decades. This is due to many factors including the fact that many industrial processes require that data be collected under closed loop conditions for safety or other operational reasons. We have studied aspects of both non-parametric and parametric closed loop identification.

In the case of non-parametric estimation, we have studied the, so-called, indirect method in which one first estimates a model for the closed loop and then inverts this model to obtain the open-loop frequency response. A well known difficulty with this procedure is that the inverse is ill-conditioned and this leads, in general, to infinite variance for the resulting estimates due to a singularity at unity. We have overcome this ill-conditioning by using an exclusion zone around unity. This has led to novel results quantifying the bias and variance of frequency domain estimates obtained from closed loop data. More recently we have studied an alternative regularization procedure based on Tikhanou regularization. We have shown that this leads to the same qualitative conclusions as obtained earlier via the use of an exclusive zone.

We have additionally characterised the probability density function of such closed loop estimates under the assumption that the corresponding closed loop system estimate has complex normal distribution in the frequency domain. The probability density function can be described as a horseshoe encircling the inverse of the controller, with a global maximum


CCCDDDSSSCCC

on the line between the true value and the inverse of the controller. The expected value of the absolute value of such estimates is finite, and we have proposed it as a measure of variation. We have also derived and discussed new expressions for the variance when an exclusion zone is introduced around the singularity.

Similar results have been obtained for errors-in-variables estimates in open-loop, where there is assumed to be additive Gaussian noise on both inputs and outputs. In these circumstances the variance of so-called frequency response function measurements is infinite, but can be rendered finite with an exclusion zone. An expression for the variance has been derived which includes the case where the noise sources are correlated. The expression is useful for the estimate without exclusion zone over a wide range of input signal to noise ratio. We have also constructed novel confidence regions for such measurements via the Minkowski division of circular confidence regions for the output and input spectra.

Local Maxima of Likelihood Functions

We have been studying the conditions under which the maximum likelihood algorithm for ARMAX models leads to a local maxima. We have found necessary and sufficient conditions such that the gradient of the (asymptotic) likelihood function takes one closer to the true parameters. Extensions to the multi-input, multi-output case have also been studied.

System Identification for Wide Band Systems

We have been interested in developing novel identification algorithms aimed at special applications. One particular problem that we have studied is the identification of highly resonant systems over a wide-bandwidth (say ten decades of frequency). For this purpose, we have developed a method based on, so-called, Frequency Localizing Basis Function, which seems to outperform all other methods for this application.

C.4 Nonlinear and Mixture Modelling

Project Leader: Richard Gerlach

Other Researchers: Robert King, Kerrie Mengersen, Darfiana Nur, Paul Rippon

Nonlinear systems characterise many aspects of control. The Nonlinear and Mixture Modelling Project aims to pursue research into methods for describing such complex behaviour. Mixture models provide parametric or semi-parametric frameworks for modelling these and other systems


CCCDDDSSSCCC

and, because of the delicacy required in estimation, they are ideal for practising new estimation methods. Work has commenced on transferring to other control problems the technique of decomposing mixtures using latent variables. A Research Associate position is about to be appointed to this component of the project.

Timeno

Value

Den

sity

20 40 60 80 100

0.0

0.01

0.02

0.03

0.04

0 50 100 150 200 250

Figure C.2: Examples of mixture models arising from nonlinear representations of digitised images (left) and realisations from multiple latent groups (right)

Nonlinear Modelling

Nonlinear models are encountered in a variety of disciplines, from finance to control. Bayesian approaches to estimation and prediction from such models have been suggested in a correspondingly wide range of literature. This project aims to amalgamate this body of research in order to construct improved approaches relevant to control problems. This is currently being applied to the problem of variable selection, that is, determination of the number of modes or other characteristics of the nonlinear model.

New methods for nonlinear modelling are also under consideration. For example, Bayesian mixtures of triangular distributions have been developed by researchers on this project and extensions to this approach for goodness of fit are under investigation. A paper is in preparation.

Generalised Distributions

Quantile-based generalised distributions, including the g-and-k and g-and-h, contain additional parameters that control skewness and kurtosis, and therefore provide flexible parametric frameworks for describing a wide variety of distributional shapes.

A new Bayesian approach to estimation of these distributions has been developed as part of this project. This has been applied to process control problems, resulting in a draft paper that is expected to be completed in early 2004.


CCCDDDSSSCCC

Bayesian approaches to regression estimation using these generalised distributions allow for parametric representation, estimation and testing of nonlinear systems, accommodating a wide range of non-normal residual behaviour. A paper on this approach is almost complete.

Mixture Models

Digitised images can be represented on a grey scale and cast as a nonlinear model (Figure C.2a). The analysis of such images is important in robotic visualisation and optical control situations. Bayesian mixture models are being applied to this problem, with particular focus on accommodating spatial information. Extensions to allow modelling of such images over time are also being considered. This research has lead to the publication of one refereed paper in 2003, the acceptance of a second paper to appear in 2004, the submission of a third paper to a refereed journal, and invited and contributed presentations at international conferences.

Research has been undertaken into multivariate mixtures, with applications in meta-analysis. In this context, information about multiple (correlated) responses is combined both across responses and over separate (independent or dependent) studies. Consideration of two particular approaches resulted in a paper published in a refereed journal (2003). Methods for accommodating heterogeneous study quality have also been developed and applied to a case study, resulting in acceptance, subject to revision, of a refereed journal paper, expected to appear in 2004.

Research in Bayesian methods and Bayesian mixture modelling has led to the presentation of an invited four-day short course on Bayesian methods and mixtures in February 2004. Invitations to repeat this course have been received separately from groups in Canberra, Cairns and Perth. These will take place during the course of 2004.

C.5 Robot Location and Vision

Project Leader: Rick Middleton

Other Researchers: Stephan Chalup, Robert King

The Robot Location and Vision Project aims to develop methods and algorithms for problems of localisation and vision in robots. Using the University of Newcastle NUBOTS robo-dog soccer team as a platform for practical development of this research, questions such as “Where am I?”, “Which way am I facing?”, “Where is the ball?”, “Where is the goal?” can be addressed.


CCCDDDSSSCCC

The information available to answer such questions is in the form of priors (field dimension and shape), dynamics (odometry from locomotion) and sensors (vision, distance, communicated information). Uncertainty, in the form of measurement errors, odometry errors, probabilities of kicking the ball and so on, can be represented statistically. Other issues such as bias, non-linear constraints, non-unique solutions (competitor or team-mate tackling) can also be considered in this concrete realm, providing a platform for practical evaluation and modification in preparation for more general applications.

The world-class standard of robotics research undertaken through this project is reflected by the achievement of third place by the NUBOTS team in Italy in 2003, beating Carnegie-Mellon.

A PhD student has commenced study of localisation as part of this project, under the supervision of Middleton, King and Chalup.

Figure C.3: Game setup for Robot Soccer Competition

Figure C.4: Example Vision Data used for Localisation


CCCDDDSSSCCC

D. Signal Processing Programme Leader: Minyue Fu

Deputy Leader: Graham Goodwin

Programme Goals: This programme focuses on model-based signal processing. Research problems include physical modelling, system identification, model validation, prediction, filtering, and signal recovery. Examples of this type of signal processing are adaptive control, Kalman filtering, communications channel equalisation, and multi-user detection for wireless communications. Much of the fundamental research for model-based signal processing is related to other programmes. But the aim of this programme is to promote applications of modelling, control and estimation in various signal processing problems.

D.1 Subband Equalization

Project Leader: Minyue Fu

Other Researchers: Damien Marelli (Ph.D. Student)

This project is concerned with the application of a newly developed method for system identification, called subband identification, to the equalisation problem for broadband wireless communications, which is a key problem in such a communication technique. In particular, we study new equalisation methods for communications with orthogonal frequency-division multiplexing (OFDM) modulation. The aim is to achieve faster equalisation compared with known methods such as the cyclic prefix equalization (CPE) method. A key problem in equalisation is channel estimation or system identification.

Figure D.1: Block Diagram for Subband Equalisation


CCCDDDSSSCCC

In broadband wireless communications, the channel tap size is typically very large. For such a system identification problem, the subband identification approach is known to be much more numerically efficient. Hence, by adopting the subband identification approach to equalisation, we expect to come up more efficient algorithms. Alternatively, with the same computational complexity, we expect the come up algorithms achieving better equalisation.

D.2 Channel Modelling


Other Researchers: Lief Hanlen (Ph.D. Student)

In this project, we study the problems of channel modeling and channel capacity estimation for wireless communications systems with multi-antennas. Multi-antenna design has been used recently in conjunction with space-time coding to significantly improve the channel capacity of a wireless communication system. However, the known theoretical estimates of channel capacities are based on channel models with ideal assumptions on multipath propagation. In this project, we propose to model a multi-antenna system using the fundamental physical principles on radio wave propagation. The aim is to derive an accurate yet practical multi-input-multi-output model for communications between arbitrary antenna arrays in two volumes in a scattering environment. This model can then be used to compute the channel capacity of a multi-antenna system.

Figure D.2: Volumetric Representation of Multi-antenna Systems

D.3 Image Representation, Sampling and Reconstruction

Project Leader: Graham Goodwin

Other Researchers: Arie Feuer (The Technion, Israel)

A core question in signal processing is that of signal representation with respect to different functional bases and the role that sampling plays in forming such bases. The inverse problem of signal reconstruction also hinges on the use of basis representations. These questions can have difficult manifestations, for example in video imaging with camera


CCCDDDSSSCCC

movement. We plan to address this amongst other problems in signal representation and reconstruction. Two particular topics were studied in 2003:

Signal Sampling and Recovery

An important question in signal processing is the capacity to recover signals from samples. Many applications involve motion between the object and the sensor. We have studied necessary and sufficient conditions for signal recovery from samples collected with motion.

State Estimation Using Quantized Measurements

Many Problems naturally lead to measurements that are quantized. Two specific examples that we have studied are:

Exhaust gas O2 sensors

Coarsely quantized data as used in control over communication networks

We have developed novel algorithms motivated by both deterministic and stochastic noise models for state estimation using quantized coefficients.

Digital Elevation Modelling

We have been studying a number of applications of Total Least Squares and Structured Total Least Squares. One particular problem that has been studied is that of fitting a Regular Triangular Network in Digital Elevation Modelling to data having errors in x, y and z directions.

D.4 Dynamic Modeling of Turbo Decoding


One of the major breakthroughs since the pioneering work of Shannon in coding theory is the invention of turbo codes. These codes are so powerful that they are only a fraction of dB away from the Shannon limit and yet the decoding complexities are well within the reach of today's computing power. A similar type of codes, called low-density parity check codes, which use a decoding algorithm similar to those for turbo codes, are even more powerful. They are shown to be within 0.01 dB of the Shannon limit!

Despite of the remarkable properties of these codes, the amazing power of the decoding algorithm remains one of the major mysteries in the coding theory. This project aims to provide a suitable dynamic model for turbo decoding.


CCCDDDSSSCCC

In our initial investigation, we have successfully developed a stochastic framework to model the dynamics of turbo decoding. Using this model, we are able to accurately predict the behaviour of the decoding process. This allows us to understand the amazing power of turbo decoding. The model explains well why certain types of turbo codes work well and but others, which are seemingly very similar, do not work. The model also predicts very well how close a given turbo code can potentially be to the Shannon limit. Most importantly, the model can be used as a convenient tool to design new turbo codes.

Figure D.3: Dynamic Model for Turbo Decoding


CCCDDDSSSCCC

E. Process Control and Optimisation Programme Leader: Rick Middleton

Deputy Leader: Julio Braslavsky

Programme Goals: The Partnerships between researchers and industry enable reciprocal transfer of knowledge and new ideas of great potential impact on the community and economy. This research programme encompasses four research projects motivated by and in collaboration with industrial partners. The main underlying theme of these projects is the application of advanced control and optimisation techniques to maximise asset utilisation and production in selected industrial processes of significant complexity. The complexity of the dynamics of such processes arise from factors including model errors, unknown disturbances, nonlinearities, distributed parameter systems, elements of Human Machine Interaction and hybrid (Discrete and Continuous State) components. Expected outcomes of the programme include high quality research solutions and human resources tailored to the needs of the Australian industry.

E.1 Integrated Mine Planning (BHP Billiton)

Project leader: Will Heath

Other researchers: Greg Adams, Gary Froyland, Graham Goodwin, Merab Menabde, Kerrie Mengersen, Rick Middleton, Peter Stone, Robert Watts, Karsten Weihe.

This project involves investigation and algorithm development for a particularly complex operation, including stochastic elements (to model uncertainty in the ore grade, and in the mineral and metal prices) and complex finite state optimization problems (where a very large number of discrete decisions can be made). The aim is to combine the background of the University of Newcastle academics in systems theory, stochastic modelling and finite horizon optimization with BHP Billiton’s detailed problem knowledge to design new algorithms yielding improved and more valuable mining operations.


CCCDDDSSSCCC

Figure E.1: BHPB Operating Diamond Mine (photo courtesy of BHPB Diamonds Inc.)

Much of 2003 has been spent bringing project members up to speed with both the strategic and technical issues, as well as defining a manageable problem for study. To this end there have been several discussions as well as meetings at which third parties were also invited, including Dr Natashia Boland (University of Melbourne) and Jim Netterfield (Asset Development Services for BHP Billiton Carbon Steel Materials).

The name of the project reflects the outcome of these discussions. The project will develop strategic mine planning and business valuation tools that can be used by mine planners and management to add value to operations.

These tools will provide life-of-mine plans (extraction schedules) and infrastructure configurations that have been jointly optimised. The tools will be designed to assist decision makers in making the correct decisions by providing the ability to rapidly assess the best plan/configuration over all reasonable scenarios in a single step.

Figure E.2: Example of optimal pushbacks for a BHPB mining operation


CCCDDDSSSCCC

To this end we have drafted an initial confidential report which articulates the basic mine scheduling problem as a mixed integer linear programme. The optimization takes into account extraction and processing costs together with ore prices discounted over time. Our longer term goal is to design and implement algorithms that determine a mine and infrastructure configuration together with a strategic plan that maximizes net present value (NPV). The design will account for a range of feasible future price scenarios, allow for stockpiling and also account for uncertainty (for example mine grade uncertainties and variable processing rates). We are investigating the use of branch and cut methods for solving such problems. Recently we have been joined by Prof. Karsten Weihe, who brings an expertise on integer programming and operations research. Karsten will take over as project leader for 2004.

E.2 Optimisation Based Operator Guidance Schemes (BHP-Billiton Innovation)

Project leader: Julio Braslavsky

Other researchers: Greg Adams, Richard Gerlach, Tom Honeyands, Katrina Lau, Kerrie Mengersen, Rick Middleton, Paul Rippon, Steve Vandenberg

This project is funded by a partnership of the Centre with BHP Billiton Innovation (Newcastle) and encompasses two main Industrial Case Studies described below:

Optimisation of the HBI Plant

BHP Billiton s Hot Briquetted Iron Plant at Port Headland in WA produces iron briquettes from fine iron ore using the novel FINMET® technology. This Industrial Case Study aims at developing a better understanding of the FINMET® technology in order to optimise the overall HBI process performance.


CCCDDDSSSCCC

Figure E.3: Port Headland HBI Plant

Initial work within this Case Study involved the identification of problems of common interest in this process and the organisation of the consequent research work. Some of the problems under current analysis are:

Modelling and Quantification of a single process train Intensity.

Modelling and Quantification of a single process train Capacity.

Modelling and Optimisation of the multitrain process.

Figure E.4: Schematic of the HBI Process


CCCDDDSSSCCC

A CDSC Research Fellow will be appointed in 2004 to work full time in this Industrial Case Study. Planned activities for 2004 include an HBI Workshop in conjunction with the Bayesian Learning CDSC programme.

Modelling and Control of Copper Heap Leaching

Heap Leaching is a technology for copper extraction from low copper content ores. This technology is attracting increasing interest in the last years as it appears as a simple, efficient and environmentally friendly alternative to standard technologies based on pyrometallurgy for copper extraction.

The heap leaching technology presents significant room for improvement with major potential economic benefits. Therefore, BHP Billiton is interested in further developments in this area and has recently signed an agreement with the Chilean state company CODELCO for the common exploitation of copper heap leaching facilities in Chile. The aims of this Industrial Case Study are to provide the modelling and control tools necessary in the optimisation of these processes.

Figure E.5: Copper extraction via smelting and leaching technologies

In the leaching process, the ore or concentrate is brought into close contact with a leach solution (frequently sulfuric acid) that dissolves the copper and leaves a residue of gangue. Key catalysts in the leaching process are bacteria like the thiobacillus ferrooxidans, which convert the metal compounds into their water-soluble forms.


CCCDDDSSSCCC

Figure E.6: Schematic of a heap leaching process

Figure E.7: Thiobacillus ferrooxidans (blue) attached to copper mineral

crystals

Efficient control of these processes will play a crucial role in the improvement of the heap leaching technology, although appears as a research topic largely unexplored in the literature.

Initial work in this Case Study involved the compilation of relevant background information about these processes, including a review of the current research advances in modelling and control in the area. Background information and a proposal for the development of a preliminary mathematical model and control scheme have been documented in Technical Report EE03029. A PhD student will start working in this topic in March 2004.


CCCDDDSSSCCC

E.3 Next Generation Model Based Control Tools (Matrikon)

Project leader: James Welsh

Other researchers: Sam Crisafulli, Graham Goodwin, Rick Middleton, Maria Seron, Richard Thomas

To kick-start the project and initiate interaction between CDSC and Matrikon a one-day Mini-Symposium on Model Predictive Control (MPC) was conducted at the University of Newcastle, December 1. There were 15 attendees from CDSC, Matrikon and the School of Electrical Engineering and Computer Science, Newcastle University.

Presenters at the symposium were representatives from both CDSC and Matrikon. Presentations from CDSC staff outlined their current research interests in MPC whereas Matrikon staff based their presentations on practical experiences obtained during design and deployment of MPC in industry.

At the conclusion of the presentations a round table discussion was used to brainstorm ideas for the next generation model predictive control toolbox. This discussion was fruitful in generating a wish list of desirable features that should become part of a MPC toolbox.

E.4 Nutating Mill Control (Hicom International Pty. Ltd.)

Project leader: Graham Goodwin

Other researchers: Greg Adams, Rick Middleton

This project is fully funded by Hicom International. During 2003, work progressed in completing a detailed functional specification and proposal for the advanced control system. Technical details of this project are confidential.


CCCDDDSSSCCC

F. Mechatronics Programme Leader: Reza Moheimani

Deputy Leader: Rick Middleton

Programme Goals: Many technical processes and products in the area of mechanical and electrical engineering show an increasing integration of mechanics with electronics and information processing. This integration is between the components (hardware) and the information-driven functions (software), resulting in integrated systems called mechatronic systems. The development of mechatronic systems involves finding an optimal balance between the basic mechanical structure, sensor and actuator implementation, automatic digital information processing and overall control, and this synergy results in innovative solutions. The practice of mechatronics requires multidisciplinary expertise across a range of disciplines, such as: mechanical engineering, electronics, information technology, and decision-making theories. These complicated interactions generate a rich and complex set of dynamic behaviours to be analysed and controlled. This research programme is aimed at investigating such analysis and control questions in emerging mechatronic systems.

F.1 Dealing with Hystersis in Piezoelectric Transducers

Project Leader: Reza Moheimani

Other Researchers: Ben Vautier (M.E. student)

Piezoelectric transducers are often driven using voltage amplifiers. When operated at high voltages, these transducers display a significant level of hysteresis. Therefore, at such levels, they can no longer be viewed as linear devices. The mainstream approach to addressing this issue is centred at modelling the hysteresis using the Preisach operator, and then obtaining an inverse of the model, which can be used to render the transducer linear. Apart from being a tedious task, the inherent assumption in this technique is that the hysteresis associated with piezoelectric transducers is rate independent, which is not true in many circumstances.

An interesting property of piezoelectric devices, which has been known to the scientific community since late 1980’s is that when driven by current or charge amplifiers, piezoelectric transducers display little hystersis. This is clearly illustrated in Figure F.1 and Figure F.2 which compare experimentally obtained hysteresis plots associated with a piezoelectric


CCCDDDSSSCCC

transducer when driven by voltage and charge. Despite this interesting, and useful property, hardly anyone has ever used charge or current amplifiers to drive piezoelectric loads in a systematic way. This is due to the highly capacitive nature of piezoelectric loads, which makes the operation of standard current and charge amplifiers rather difficult.

The new generation of charge and current amplifiers developed in the Laboratory for Dynamics and Control of Smart Structures enables us to operate piezoelectric loads effectively over a wide bandwidth. Thanks to this development, we are now able to use charge and current amplifiers to operate piezoelectric transducers. This calls for a study of the dynamics of flexible structures with embedded charge driven piezoelectric actuators, and its control related issues. This project is aimed at developing a framework for modelling and control of this specific class of structures.

Figure F.1: Hysteresis plot of a piezoelectric transducer when driven by voltage

Figure F.2: Hysteresis plot of a piezoelectric transducer, when driven by charge


CCCDDDSSSCCC

F.2 Vibration Control Using Shunted Piezoelectric Transducers


Other Researchers: Sam Behrens, Andrew Fleming, Manfred Morari (ETHZ), Dominik Nederberger (ETHZ),

This has been an ongoing research project for several years, which among other things has resulted in identification of the feedback control structure associated with shunt damped systems, the development of a class of resonant controllers for piezoelectric shunt damping and development of some ad-hoc, yet effective piezoelectric shunts.

More recently, the focus of the project has shifted towards developing tools for synthesis of piezoelectric shunts, which are capable of maintaining performance, and stability, in presence of varying resonance frequencies of the structure. This is quite an important issue since resonance frequencies of flexible structures are highly sensitive to environmental changes such as a drift in temperature, etc. In this regard, recent progress has been made in extending the technique of piezoelectric shunt damping to synthesize “active” piezoelectric shunts, and in collaboration with colleagues from the ETHZ, to design and implement multi-modal adaptive shunts.

The active shunts that have been designed and experimentally implemented appear to be quite promising. In particular, we have managed to achieve significant damping with these shunts as illustrated in Figure F.3, which plots experimental results obtained from a cantilevered beam with bonded piezoelectric transducers. Performance of the shunts appears to be quite robust with respect to changing resonance frequencies of the base structure. In fact we have demonstrated that if resonance frequencies of the structure are shifted up or down significantly (up to 14%), as a result of adding an extra mass to the structure, the closed loop system remains stable and achieves satisfactory performance. This appears to be due to the nature of the shunts. Careful examination of the Bode plot of the shunt reveals that the it attempts to mimic the properties of a negative capacitor – an ‘ideal’ shunt - around each resonance frequency of the base structure.

We also have ongoing collaborations with colleagues from the Swiss Federal Institute of Technology in Zurich. As a result of this collaboration we have developed and experimentally implemented multi mode adaptive resonant piezoelectric shunts, which operate on the basis of online optimal tuning of circuit component values by minimizing the relative phase difference between a vibration reference signal and the shunt circuit. The adaptive law converges quickly and maintains optimal


CCCDDDSSSCCC

performance in the presence of environmental uncertainties that change resonance frequencies of the base structure.

Figure F.3: Open loop vs. closed loop response of a two mode active LQG shunt

F.3 Instrumentation for Smart Structures


Other Researchers: Andrew Fleming, Clint Lawrence

Piezoelectric and electromagnetic transducers require dedicated drive and instrumentation electronics to fully utilize their dynamic range, noise performance, and bandwidth capabilities.

As the need has arisen, a range of optimized drivers, instrumentation, and processing electronics has been developed for specialized applications ranging from high-power structural control to nanometer precision positioning devices.

Custom drives include: differential voltage amplifiers, dynamic current sources, charge amplifiers, and current-fluxion drives. Such devices are typically employed together with instrumentation and processing packages to implement passive, semi-active, or active piezoelectric and electromagnetic shunt control systems.

Implementation of the structural and drive control algorithms is performed by either self-calibrating analog processors or block-diagram programmable DSP subsystems. The most recent drive generation contains optional integrated processing electronics to implement default control algorithms for ‘off the shelf’ structural vibration control.

Current research includes the design of low-noise, ultra-low frequency, and DC accurate charge amplifiers for piezoelectric loads. The next


CCCDDDSSSCCC

generation of DSP subsystems is also under development. Peripherals include high-resolution signal conversion, digital IO, broadband host-interface, and shaft encoder. An Analog-Devices super-scaler DSP will provide 1200 Million Floating Point Operations Per Second.

Figure F.4: 400V Power amplifier with integrated charge, current, current fluxion, and voltage instrumentation.

Figure F.5: Synthetic impedance with interchangeable analog signal processor installed vertically.

Figure F.6: Precision switch-mode dynamic current amplifier for piezoelectric loads.

Figure F.7: Digital Signal Processing subsystem.


CCCDDDSSSCCC

F.4 Electromagnetic Transducers


Other Researchers: Sam Behrens, Andrew Fleming

The purpose of this research programme is to extend control design methodologies developed for piezoelectric transducers to electromagnetic transducers. Although quite effective for specific applications, piezoelectric transducers are incapable of offering high strokes, which is needed in applications such as active vibration isolation systems. Electromagnetic transducers, on the other hand, appear to be ideal devices for applications of this nature.

Similar to piezoelectric transducers, electromagnetic transducers are capable of transforming electrical energy into mechanical energy and vice versa. However, they are highly inductive loads, which differentiates them from piezoelectric devices that electrically function as a large capacitance.

An electromagnetic transducer can be used in active structural control applications by shunting its terminals to an electrical impedance. If the impedance is designed properly, certain properties of the structure can be controlled in an effective way. This approach has the advantage that the need for a sensor, such as an accelerometer, is removed.

To test this technique an electromagnetic transducer has been designed and built in the Laboratory for Dynamics and Control of Smart Structures. A picture of the device and its schematics are shown in Figure F.8 and Figure F.9, respectively. A number of passive and active shunts have been designed and successfully implemented on this device in the Lab.

Figure F.8 An electromagnetic transducer


CCCDDDSSSCCC

Figure F.9 Schematics of the electromagnetic transducer

F.5 Nanopositioning Systems


Other Researchers: Andrew Fleming

This project has started recently, and is aimed at developing a new generation of fast nanopositioning systems.


CCCDDDSSSCCC

PUBLICATIONS 2003 Books

S.O.R. Moheimani, D. Halim and A. J. Fleming. Spatial Control of Vibration: Theory and Experiments . World Scientific, 2003; ISBN 981-238-337-9.

Chapters in Books

G.C. Goodwin, T. Perez and J.A. De Doná. “An introduction to constrained control”, Chapter in Modelling Control and Optimization of Complex Systems, Kluwer Academic Press, 2003.

K.L. Mengersen and C.P. Robert. “Population Markov Chain Monte Carlo: the Pinball Sampler”. Chapter in Bayesian Statistics 7. Editors, J.O. Berger, A.P. Dawid, A.F.M. Smith, Oxford University Press, 2003.

A.G. Wills, W.P. Heath and M. Perrier. “Commande prédictive linéare: application au raffinage d’huiles alimentaires”. Chapter in J.-M. Flaus and L. Boillereaux (editors): Automatique pour les procédés agroalimentaires; volume 2, commande et supervision, Lavoisier, 2003.

Plenary and Invited Addresses

G.C. Goodwin, Hurwitz Memorial Lecture, ETH Zurich, December 16, 2003. “Finite alphabet control and estimation”.

G.C. Goodwin, Hurwitz Memorial Lecture, ETH Zurich, December 17, 2003. “Constrained estimation”.

G.C.Goodwin and D.E. Quevedo, Plenary Address, International Conference on Control, Automation and Systems, Gyeongju, Korea, October 22-25, 2003. “Finite alphabet control and estimation”.

K. Mengersen, Invited Speaker, Complex Systems International Workshop, Melbourne, November 2003. “Population Monte Carlo: Estimating hidden Markov models using the pinball sampler”.

K. Mengersen, Invited Speaker, Queensland Statistical Society Conference, Toowoomba, October 2003. “Adjusting for quality in meta-analysis”.

K. Mengersen, Invited Speaker, Australian Youngstats Conference, Newcastle, September 2003. “Bayesian meta-analysis”.


http://www.wspc.com/books/engineering/5246.html

http://www.wspc.com/books/engineering/5246.html

CCCDDDSSSCCC

K. Mengersen, Invited Speaker, Bioinformatics Australia Conference, Sydney, September 2003. “Bayesian methods for QTL mapping”.

K. Mengersen, Invited Speaker,Bayesian Case Studies International Workshop, Santa Cruz, U”SA, August 2003. “The pinball sampler”.

K. Mengersen, Invited Speaker, New Zealand Statistical Society, Palmerston North, July 2003. “Multivariate meta-analysis”.

K. Mengersen, Invited Discussant, Objective Bayes Workshop, Aussois, Switzerland, June 2003. “Why use full Bayesian models? (Why useful Bayesian models?)”.

R.H. Middleton, Invited Presentation, Young Statisticians Conference, Newcastle, September 2003. “NUBots: Robot Soccer – modelling and uncertainty”.

B. Sims, Invited Speaker, International Conference on Fixed Point Theory and its Applications, July 13-19, Valencia, Spain. “Some thoughts (probably random) on generalized metrics”.

B. Sims, Plenary Lecture, Władysław Orlicz Centenary Conference and Function Spaces VII, July 21-25, Poznań, Poland. “Some new and old perspectives on metric fixed point theory”.

W. Szymanski, Invited Lecture, Fifth International Congress on Industrial and Applied Mathematics, Sydney, July 2003. “Quantum spaces and graph algebras”.

Journal Papers (published)

S. Behrens, A.J. Fleming, and S.O.R. Moheimani. “A broadband controller for shunt piezoelectric damping of structural vibration”, Smart Materials and Structures, Vol.12, No.1, pp.36-48, February 2003.

S. Behrens, S.O.R. Moheimani and A.J. Fleming. “Multiple mode current flowing passive piezoelectric shunt controller”, Journal of Sound and Vibration, Vol.266, No.5: pp.929-942, October, 2003.

K. Deicke, J.H. Hong and W. Szymanski, “Stable rank of graph algebras. Type I graph algebras and their limits”, Indiana Univ. Math. J. Vol.52, pp.963-980, 2003

K. Deicke, D. Pask and I. Raeburn, “Coverings of directed graphs and crossed products of C-algebras by coactions of homogeneous spaces”, Internat. J. Math. Vol.14, pp,773-789, 2003.

A. Feuer and G.C. Goodwin. “Linear deterministic adaptive control – fundamental limitation?”, Systems and Control Letters, Special Issue on Adaptive Systems, Issue 1, Vol.49, 2003.


CCCDDDSSSCCC

A. J. Fleming and S.O.R. Moheimani. “Spatial system identification of a beam and a trapezoidal plate”, IEEE Transactions on Control Systems Technology, Vol.11, No.5, pp.726-736, September 2003.

A.J. Fleming and S.O.R. Moheimani. “Precision current and charge ampliers for driving highly capacitive piezoelectric loads”, Electronics Letters, Vol.39, No.3, pp.282-284, February 2003.

A.J. Fleming, S.O.R. Moheimani and S. Behrens. “Reducing the inductance requirments of piezoelectric shunt damping circuits”, Smart Materials and Structures, Vol.12, No.1, pp:57-64, February 2003.

A.J. Fleming and S.O.R. Moheimani. “Adaptive piezoelectric shunt damping”, Smart Materials and Structures, Vol.12, No.1, pp:18-28, February 2003.

N.J. Fowler, P.S. Muhly and I. Raeburn. “Representations of Cuntz-Pimsner algebras”, Indiana Univ. Math. J. Vol.52, pp.569-605, 2003.

J.S. Freudenberg, C.V. Hollot, R.H. Middleton and V. Toochinda, "Fundamental Design Limitations of the General Control Configuration", IEEE Transactions on Automatic Control, Vol.48, No.8, pp.1355-1370, August 2003.

G.C. Goodwin and D.E. Quevedo. “Finite alphabet control and estimation”, International Journal of Control, Automation, and Systems, Vol.1, No.4, 2003.

G.C. Goodwin, D.E. Quevedo and D. McGrath. “A moving horizon quantizer for audio signals”, Journal of the Audio Engineering Society, Vol.51, No.3, March 2003.

G.C. Goodwin, M. Salgado and J. Yuz. “Performance limitations for linear feedback systems in the presence of plant uncertainty”, IEEE Trans. on Auto. Control, Vol.48, No.8, pp.1312-1319, 2003.

P. Graham and K.L. Mengersen, “Confidence limits for the ratio of two rates based on likelihood scores: non-iterative method”, Statistics in Medicine, Vol.22, pp.2071-2083, 2003.

D. Halim and S.O.R. Moheimani. “An optimization approach to optimal placement of collocated piezoelectric actuators and sensors on a thin plate”, Mechatronics, Vol.13, No.1, pp:27-47, February 2003.

W.P. Heath. “The variation of non-parametric estimates in closed-loop”. Automatica, Vol.39, pp.1849-1863, 2003.

J.H. Hong and W. Szymanski, “Purely infinite Cuntz-Krieger algebras of directed graphs, Bull. London Math. Soc. Vol.35, pp.689-696, 2003.


CCCDDDSSSCCC

J.H. Hong and W. Szymanski, “Quantum lens spaces and graph algebras”, Pacific J. Math. Vol.211, pp.249-263, 2003.

P. Kuhnert and K.L. Mengersen, “Reliability Measures for local nodes assessment in classification trees”, Comput. Statist. Data Anal. Vol.12 No.2, pp.398-416, 2003.

P. Kuhnert and K.L. Mengersen, “Bridging the gap: an integrated approach to modelling”, International Statistical Review, 2003.

P. Kuhnert and K.L. Mengersen, “Reliability Measures for local nodes assessment in classification trees. Comput. Statist. Data Anal. Vol.12, No.2, pp.398-416, 2003.

K. Lau, R.H. Middleton and J.H. Braslavsky. “Undershoot and settling time trade-offs for nonminimum phase systems”, IEEE Transactions on Automatic Control,Vol.48, No.8, pp.1389-1393, August 2003.

S.O.R. Moheimani. “A survey of recent innovations in vibration damping and control using shunted piezoelectric transducers”, IEEE Transactions on Control Systems Technology, Vol.11, No.4, pp.482-494, July 2003.

S.O.R. Moheimani, A.J. Fleming, and S. Behrens. “On the feedback structure of wideband piezoelectric shunt damping systems”, Smart Materials and Structures, Vol.12, No.1, pp:49-56, February 2003.

L.H.E. Morawska, J. Hitchins, K. Mengersen and D. Gilbert, “Characteristics of particle number and mass concentrations in residential houses in Brisbane, Australia. Atmospheric Environment, Vol.37, pp.4195-4203, 2003.

I.-S. Nam, K.L. Mengersen and P. Garthwaite, “Multivariate meta-analysis”, Statistics in Medicine, Vol.22, pp.2309-2333, 2003.

T. Perez, G.C. Goodwin and M.M. Seron. “Performance degradation in feedback control due to constraints”, IEEE Trans. on Auto. Control, Vol.48, No.8, pp.1381-1385, 2003.

I. Raeburn, A. Sims and T. Yeend, “Higher-rank graphs and their C*-algebras”, Proc. Edinburgh Math. Soc. Vol.46, pp.99-115, 2003.

J.A. Rodriguez, J.A. Romagnoli and G.C. Goodwin. "Supervisory multiple model control", Journal of Process Control, Vol.13, 2003, pp.177-191.

M.M. Seron, G.C. Goodwin and J. De Doná. “Characterizations of receding horizon control for constrained linear systems”, Asian Journal of Control, Vol.5, No.2, pp.271-286, 2003.


CCCDDDSSSCCC

B.J.G. Vautier and S.O.R. Moheimani. “Vibration reduction of resonant structures using charge controlled piezoelectric actuators”, Electronics Letters, Vol.39, No.14, pp.1036-1038, July 2003.

J. Yuz and G.C. Goodwin. “Loop performance assessment for decentralized control of stable linear systems”, European Journal of Control, Vol.9, No.1, pp.118-132, 2003.

P.V. Zhivoglyadov and R.H. Middleton. "Networked Control Design for Linear Systems", Automatica, Vol.39 No.4, pp.743-750, 2003.

P.V. Zhivoglyadov and R.H. Middleton. “Stability and switching control design issues for a class of discrete time hybrid systems”, Automatica Vol.39, No.6, pp.981-987, June 2003

Journal Papers (Accepted for publication)

T. Bates and D. Pask, “Flow equivalence of graph algebras”, to appear in Ergodic Theory & Dynamical Systems.

T. Dominguez, H. Hudzik, G. lópez, M. Mastylo and B. Sims. “Complete characterizations of Kadec-Klee properties in Orlicz spaces”, to appear in Houston J. Math.

A.J. Fleming and S.O.R. Moheimani. “Control oriented synthesis of high performance piezoelectric shunt impedances for structural vibration control”, to appear in IEEE Transactions on Control Systems Technology.

A.J. Fleming and S.O.R. Moheimani. “Improved current and charge amplifiers for driving piezoelectric loads, and issues in signal processing design for synthesis of shunt damping circuits”, to appear in Journal of Intelligent Material Systems and Structures.

D. Halim and S. O. R. Moheimani. “Reducing the effect of truncation error in spatial and pointwise models of resonant systems with damping”, to appear in Mechanical Systems and Signal Processing. W.P. Heath and A.G. Wills. “Design of cross-directional controllers with optimal steady state performance”, to appear in European Journal of Control.

J.H. Hong and W. Szymanski, “The primitive ideal space of the C*-algebras of infinite graphs”, to appear in J. Math. Soc. Japan.

R.H. Middleton, J. Chen and J.S. Freudenberg, "Sensitivity, Robustness and Achievable Hinfinity Performance in Preview Control", to appear, Automatica.


CCCDDDSSSCCC

S.O.R. Moheimani and S. Behrens. “Multi-mode piezoelectric shunt damping with a highly resonant impedance”, to appear in IEEE Transactions on Control Systems Technology.

S.O.R. Moheimani, A.J. Fleming and S. Behrens. “Dynamics, Stability and Control of Multivariable Piezoelectric Shunts”, to appear in IEEE/ASME Transactions on Mechatronics.

S.O.R. Moheimani and D. Halim. “A convex optimization approach to the mode acceleration problem”, to appear in Automatica.

I. Raeburn, A. Sims and T. Yeend, “The C*-algebras of finitely aligned higher-rank graphs”, to appear in J. Funct. Anal.

I. Raeburn and W. Szymanski, “Cuntz-Krieger algebras of infinite graphs and matrices”, to appear in Trans. Amer. Math. Soc.

K. Salehi-Rad, and K. Mengersen. “Reservicing some customers in M/G/1 queues, under three disciplines”, to appear in J. Math. Manag. Decision Sci.

A. G. Wills and W. P. Heath. “Barrier function based model predictive control”, to appear in Automatica.

D.J. Woodhouse and R.H. Middleton, "Consistency in ground potential rise estimation utilising fall of potential method data", to appear, IEEE Transactions on Power Delivery.

Conference Papers

J.C. Aguero, A. Feuer and G.C., Goodwin. “Terrain modeling via triangular regular networks”, MODSIM, Townsville, July 2003.

S. Behrens, A.J. Fleming, and S.O.R. Moheimani. “Electromagnetic shunt damping”, Proc. IEEE/ASME Conference on Advanced Intelligent Mechatronics, Kobe, Japan, July 2003.

S. Behrens, A.J. Fleming and S.O.R. Moheimani. “Electromagnetic vibration suppression”, Proc. SPIE Symposium on Smart Structures and Materials - Damping and Isolation, March 2003, San Diego, CA, U.S.A.

D.F. Coutinho, M. Fu and A. Trofino. “Guaranteed cost analysis and control for a class of uncertain nonlinear discrete-time systems”, Proc. American Control Conference, Denver, June 2003.

A.J. Fleming, S. Behrens and S.O.R. Moheimani. “Robust piezoelectric passive shunt dampener”, Proc. SPIE Symposium on Smart Structures and Materials - Damping and Isolation, March 2003, San Diego, CA, U.S.A.


CCCDDDSSSCCC

A.J. Fleming and S.O.R. Moheimani. Active H-2 and H-infinity shunt control of electromagnetic transducers, Proc. IEEE Conference on Decision and Control, Hawaii, USA, December 2003.

A.J. Fleming and S.O.R. Moheimani. “Improved Current and Charge Amplifiers for Driving Piezoelectric Loads”, Proc. SPIE Symposium on Smart Structures and Materials - Damping and Isolation, March 2003, San Diego, CA, U.S.A.

A.J. Fleming and S.O.R. Moheimani. “An autonomous piezoelectric resonant shunt damping system”, Proc. SPIE Symposium on Smart Structures and Materials - Damping and Isolation, March 2003, San Diego, CA, U.S.A.

J.S. Freudenberg, C.V. Hollot and R.H. Middleton. “A tradeoff between disturbance attenuation and stability robustness”, Proc. American Control Conference, Denver, June 2003.

G.C. Goodwin, J.C. Aguero and R.E. Skelton. “Conditions for local convergence of maximum likelihood estimators for ARMAX models”, IFAC Symposium on System Identification, Rotterdam, September 2003.

G.C. Goodwin and D.E. Quevedo. “An application of receding horizon control to estimation with quantized coefficients”, Proc. IEEE American Control Conference, Denver, June 2003.

H. Haimovich, G.C. Goodwin and D.E. Quevedo. “Rolling horizon monte carlo state estimation for liner systems with output quantization”, Proc. IEEE Conference on Decision and Control, Hawaii, USA, December 2003.

H. Haimovich, T. Perez and G.C. Goodwin. “On optimality and certainty equivalence in output feedback control of constrained uncertain linear systems”, Proc. European Control Conference, Cambridge, September 2003.

H. Haimovich, M.M. Seron, G.C. Goodwin and J.C. Aguero. “A neural approximation to the explicit solution of constrained linear MPC, Proc. European Control Conference, Cambridge, September 2003.

W.P. Heath. “The variance of non-parametric errors-in-variables estimates”, Proc. IEEE Conference on Decision and Control, Hawaii, December 2003.

W.P. Heath and A.G. Wills. “Design of cross-directional controllers with optimal steady state performance”, Proc. European Control Conference, Cambridge, UK, September 2003.


CCCDDDSSSCCC

W.P. Heath. “Confidence regions for non-parametric errors-in-variables estimates” SYSID’03, Rotterdam, August 27-29, 2003.

K. Lau and R.H. Middleton. “Switched Integrator Control Schemes for Integrating Plants”, Proc. 2003 European Control Conference, Cambridge, September 2003.

D. Marells and M. Fu. “Subband methods for OFDM equalization”, Proc. International Conference on Communications, 2003.

R.H. Middleton, K. Lau and J.H. Braslavsky. “Conjectures and Counterexamples on Optimal L2 Disturbance Attenuation in Nonlinear Systems”, 42nd IEEE Conference on Decision and Control, Hawaii, USA, December 2003.

Z. Mustafa and B. Sims. “Some remarks concerning D-metric spaces”, Proc. International Conference on Fixed Point Theory and its Applications, Valencia - 2003.

Pask and S.J. Rho. “Some instrinsic properties of simple graph C*-algebras”, Operator Algebras and Mathematical Physics, Constanza (2001). Theta Foundation, Romania, 2003, pages 325-340.

T. Perez, G.C. Goodwin and R.E. Skelton. “Analysis of performance and applicability of rudder-based stabilizers”, 6th IFAC Symposium on Manoeuvring and Control of Marine Craft, Spain, September 2003.

T. Perez and G.C. Goodwin. “Constrained control to prevent dynamic stall of ship fin stabilizers”, 6th IFAC Symposium on Manoeuvring and Control of Marine Craft, Spain, September 2003.

D.E. Quevedo and G.C. Goodwin. “Audio quantization from a receding horizon control perspective”, Proc. IEEE American Control Conference, Denver, June 2003.

D.E. Quevedo and G.C. Goodwin and J.S. Welsh. “Minimizing down-link traffic in networked control systems via optimal control techniques”, Proc. IEEE Conference on Decision and Control, Hawaii, USA, December 2003.

O.J. Rojas, G.C. Goodwin, A. Feuer and M.M. Seron. “A suboptimal receding horizon control strategy for linear systems”, Proc. IEEE American Control Conference, Denver, June 2003.

O.J. Rojas, M.M. Seron and G.C. Goodwin. “SVD Based receding horizon control for constrained linear systems: stability results”, Proc. IEEE Conference on Decision and Control, Hawaii, USA, December 2003.


CCCDDDSSSCCC

A. Syaichu-Rohman, R.H. Middleton and M.M. Seron. “A multivariable nonlinear algebraic loop as a QP with application to MPC”, Proc. 2003 European Control Conference, Cambridge, September 2003.

B.J.G. Vautier and S.O.R. Moheimani. “Avoiding hysteresis in vibration control using piezoelectric laminates”, Proc. IEEE Conference on Decision and Control, Hawaii, USA, December 2003.

J.S. Welsh and G.C. Goodwin. “Frequency localizing basis functions for wide-band identification”, Proc. European Control Conference, Cambridge, September 2003.

A.G. Wills and W.P. Heath. “An exterior/interior-point approach to infeasibility in model predictive control”, Proc. IEEE Conference on Decision and Control, Hawaii, December 2003.

J. Yuz, G.C. Goodwin, A. Feuer and J. De Doná. “Singular structure convergence for linear quadratic problems”, Proc. European Control Conference, Cambridge, September 2003.

Industry Technical Reports

General Motors USA.

J.S. Welsh, H. Haimovich, G.C. Goodwin, R.H. Middleton and J. De Doná, "Adaptive Powertrain Control: Report 3 - System Identification for Automotive Powertrain Control", Report for General Motors, USA, March 2003.

J.S. Welsh, H. Haimovich, G.C. Goodwin, R.H. Middleton and J. De Doná, "Adaptive Powertrain Control: Interim Report 1”, Report for General Motors, USA, May 2003.

J.S. Welsh, H. Haimovich, G.C. Goodwin, R.H. Middleton and J. De Doná, "Adaptive Powertrain Control: Interim Report 2”, Report for General Motors, USA, July 2003.

J.S. Welsh, H. Haimovich, G.C. Goodwin, R.H. Middleton and J. De Doná, "Adaptive Powertrain Control: Interim Report 3”, Report for General Motors, USA, July 2003.

H. Tidefelt, G.C. Goodwin, J.S. Welsh, R.H. Middleton and H. Haimovich, "Adaptive Powertrain Control: Interim Report 4” - Report for General Motors, USA, October 2003.

A. Rosen, G.C. Goodwin, J.S. Welsh, R.H. Middleton and H. Haimovich, “Adaptive Powertrain Control: Interim Report 5”, Report for General Motors, USA, December 2003.


CCCDDDSSSCCC

BHP Billiton Innovation

J.H. Braslavsky. “A Research Proposal on Control and Optimisation of Copper Heap Bioleaching Processes”. Technical Report EE03029 for BHP-BI, July 2003.

G. Adams, K.Weihe, R.Middleton and W.Heath, “Integrated Mine Planning Summary”, Report for BHP-Billiton Innovations, December 2003.

HiCom International

G. Adams and R.H. Middleton, “Hicom Special Control System: Functional Specification and Proposal” Report for HiCom International Pty. Ltd., December 2003.


CCCDDDSSSCCC

ACTIVITY PLAN FOR 2004 Some of our main plans for 2004 are:

Complete appointments of research academics and postgraduates for a variety of research projects.

Modern Industrial Control Short Course will be held in May 2004.

A workshop was held on 27 February to discuss the BHP-Billiton Hot Briquetted Iron project.

Purchase CPLEX and initiate numerical studies for the integrated mine planning project.

Apply for ARC LIEF funding to provide additional infrastructure for mechatronics control research.

Arrange the 2004 Centre Retreat.

Run postgraduate courses on System Identification and constrained control.

Organize and participate in the IFAC Mechatronics Symposium to be held in Sydney September 2004.

The 2004 Advisory Board Meeting & Official Launch were held on 13 February 2004.

Review Committee, to meet at least 4 times.

Initiate a CDSC Distinguished Lecturer Programme.


CCCDDDSSSCCC

PERFORMANCE INDICATORS REPORT (P.1) Research

Description

2003 Actual

Details of 2003 Outcomes

2003 Targe

t

03-07 Targe

t Refereed International Journal & Conference Publications

62

29 International Journal Publications see page 63. 33 International Conference publications see page 67.

40 200

Patents - - 3 Major Presentations (Keynote etc.)

14

See publications, page 62. 1 5

Science Citation counts for CDSC investigators

301 Search performed on ISI Web of Science 29 March 2003 by Julio Braslavasky, for all Centre Programme Leaders and Deputy Leaders. Numbers are external citations only to this group.

200 1000


CCCDDDSSSCCC

(P.2) Research Training and Professional Education

Description 2003 Actual


2003 Target

03-07 Target

Recruit 25 Postgraduate Students

6 RHD commencements in 2003 (see page 6): Jose Mare; Alejandro Rojas; Jason Tyler; Mark Smith; Ian Searston.

5 25

Completed Research Higher Degrees

4

RHDs awarded in 2003: Dunant Halim, Katrina Lau, Damien Marelli, Adrian Wills: see page 6.

4 20

Supervise 1st Class Honours students

8 S. Allen: ‘Dual graph constructions for higher-rank graphs” (Pask). M. Freeston: ‘Localisation & World Modelling in Robot Soccer’ (Middleton) P. Gibson: ‘Automatic Book Sorting Machine’, (De Dona) D. McLaren: ‘Fully Interactive Online Filter Design Programme’ (Fu) D. Marelli: ‘Proper actions on graph algebras’ (Raeburn) M. Ozturk: ‘Electromagnetic Levitation using Induction Levitators’, (Braslavsky) S. Webber: ‘Embedded DSP Prototyping System’ (Moheimani) C. Seysener: ‘Vision Processing for Robocup 2003’ (Middleton)

6 30

Professional Courses run

0 - 1 5

Senior undergraduate and postgraduate courses taught by Centre investigators in the area of complex systems

4 ELEC4400 Automatic Control (De Dona) ELEC4410 Control System Design (Braslavsky, Welsh) ELEC4210 Electronics Design (Middleton) STAT3010: Statistical Inference (Mengersen)

4 20

Interactive Learning Laboratories developed

3 New modules completed in 2003: Rolling Mill control Dynamics of Rockets Control of Rockets

2 4


CCCDDDSSSCCC

(P.3) International, national and regional links and networks:



2003 Target

03-07 Target

International Research Visitors

19

see page 11. 10 50

Papers with international co-authors

14

8 journal see page 63. 6 conference see page 66.

20 100

National and International Workshop organisation

4

IPC Memberships: ECC 03 (Goodwin, Middleton), IFAC SysID (Goodwin). Workshops: 4 – see page 13.

5 25

Visits to Overseas Labs

17

Goodwin – Lund, May; Singapore, May; ETH Switzerland, December; University of Western Ontario, May; University of Alberta, May;

Mengersen - Insee, France, (collaborator Christian Robert), June.

Middleton – Hamilton Institute, Maynooth, September;

Moheimani – University of Wisconsin, Madison, February, University of Waterloo, February.

Raeburn - Mathematics Insitute, University of Copenhagen, Denmark

Sims - Department of Analysis, U. of Valencia, Spain.

Szymanski - Department of Applied Mathematics, Korea Maritime University, South Korea; Max Planck Institute, Bonn, Germany.

Pask - University of Washington, Seattle, USA; University of Nevada, Reno, USA.

Ramagge - University of Oslo, Norway; University of Cantabria, Santander, Spain

10 50


CCCDDDSSSCCC

(P.4) End-user links:



2003 Target

03-07 Target

Postgraduate Students involved in industrial projects

S. Dodds – Food Science Aust. H. Haimovich – General Motors Research S. Mitchell – Connell Wagner T. Perez – ADI, D. Quevedo – Lake Technologies A. Syaichu-Rohman – Electricite de France J. Welsh – Matrikon A. Wills – Goodman Fielder, Meadow Lea D. Woodhouse – Energy Australia

5 10

Visits by Centre Researchers to Industry

7 Goodwin – Matrikon (May). Goodwin – IAS. Goodwin, Middleton, Moheimani – Australia

Telescope. Mengersen - Department of Natural Resources,

Queensland; (August/September) Middleton – Matrikon. Middleton, Heath – BHP-BI, Melbourne (April). De Dona, Seron Matrikon (February).

5 25

(P.5) Organisational Support:



2003 Target

03-07 Target

Annual cash contributions from collaborating organisations

$400K BHP-BI: $100K University of Newcastle: $300K

$400K $2250K

In-kind contributions from collaborating Organisations

To be confirmed. $1.5M $7.5M

(P.6) Governance:



2003 Target

03-07 Target

Advisory Board Previous SRC Advisory Board Meeting held 12 February 2003. Positive comments received.

- -


CCCDDDSSSCCC

(P.7) National Benefit:



2003 Target

03-07 Target

Student Placements in other organisations

5 Tristan Perez University of Wales College, Newport, Mechatronics Research Centre.

Simon Dodds University of Nijmegen, The Netherlands.

Dunant Halim University of Adelaide, Mechanical Engineering.

Leif Hanlen NICTA, Canberra.

Damien Marelli France.

10

Case Studies of Benefits

CDSC has agreed to co-fund, with Engineers Australia and other groups a case study to be conducted by Emeritus Professor Mike Brisk

-

Industry Technical Reports

9 See page 70. 5 25


CCCDDDSSSCCC

FINANCIAL STATEMENT Statement of Income & Expenditure for the Year Ended 31 December 2003 (CIDAC SRC Account)

Actual $ $ 2002 Balance Brought Forward 38579 Grant Income 2003` 898,756 Consultancy Fee 3,991 Other Income 3,151

TOTAL REVENUE 2002 944,477 Expenditure Salaries (Academic) 160,126 Salaries (General) 157,628 Salary On-Costs 90,359 Scholarships 32,248

TOTAL SALARY AND RELATED COSTS 440,361 Non-Salary Expenditure Consumables 16,340 Services 17,135 Travel 107,299 Student Support 17,836 Repairs and Maintenance 80 Utilities 4,125 Equipment 35,864 Other Expenditure 312

TOTAL NON-SALARY EXPENDITURE 198,991 TOTAL EXPENDITURE 639,352 BALANCE AS AT 31/12/02 305,125


CDSC - eng.newcastle.edu.au · Chalup. ARC Centre for Complex Dynamic Systems and Control – ANNUAL REPORT 2003 4 . CDSC STAFF DIRECTOR: Professor Richard H. Middleton ... Dr David

Documents