High-Performance Induction Motor

Control Via Input-Output Linearization

Marc Bodson, John Chiasson, and Robert Novotnak

T he induction motor is the motor of choice in many industrial

applications due to its reliability, ruggedness, and relatively
low cost. Its mechanical reliability is due to the fact that there is
regulated to a constant value. Therefore, the rotor speed is only
asymptotically decoupled from the rotor flux. Following
Krzeminski [3], Marino et al. [4], [SI developed a (voltage-com-
no mechanical commutation (Le., there are no brushes nor com- mand) input-output decoupling controller which decouples the
mutator to wear out as in a DC motor). Furthermore, it can also control of the magnitude of the rotor flux from the regulation of
be used in volatile environments since no sparks are produced as the rotor speed. In this approach [SI, a nonlinear transformation
is the case in the commutator of a is performed on the state variables
DC motor. However, the induction such that, in the new coordinates,
motor presents a challenging con- the speed and rotor flux magnitude
trol problem. This is due primarily are decoupled by feedback.
to three issues. i) The dynamical Sabanovic and Izosimov [7] used a
system is nonlinear. ii) Two of the similar transformation to achieve a
state variables (rotor fluxes/cur- sliding mode controller for input-
rents) are not usually measurable. output decoupling of rotor flux and
iii) Due to heating, the rotor resis- speed. Kim et al. [8] have reported
tance varies considerably with a on a current-command input-out-
significant impact on the system dy- put linearization controller where
namics. the rotor flux is varied to minimize
The present state of the art in the input power to the stator while
induction motor control is the so- maintaining constant motor speed.
called field-oriented (or, vector In this work, we develop a current-
control) technique introduced by command input-output lineariza-
Blaschke [ 11. This method consists tion controller and present
in rewriting the dynamic equations of the induction motor in a experimental results In which the rotor flux is vaned to achieve
reference frame that rotates with the rotor flux “vector” (i.e., a optimum torque for high-performance motion control applica-
time-varying complex phasor representation of the fluxes in the tions.
rotor windings). In this new coordinate system, one observes that Other approaches to induction motor control have appeared
by holding the magnitude of the rotor flux constant, there is a in the recent literature. In [ 161 dynamic feedback linearization
linear relationship between a control variable and the speed. was shown to be applicable to speed control, but not to position
Further, the reference for the rotor flux can be adjusted in order Control. Ortega and EsPinosa [171[181have designed a nonlinear
to keep the stator voltages from saturating at high speed (this is controller based on the passivity properties of the induction
the so-calledfield-weakening approach to control, which essen- motor while Kanellakopoulos et a1 [I91 have designed a non-
tially decreases the torque output of the motor in exchange for linear controller using back-stePPing techniques.
reducing the back-emf and thus the voltage needed to achieve The article is partly tutorial, providing the necessary back-
higher speeds). ground in induction motor modeling and control which hopefully
A disadvantage of the Blaschke field-oriented controller is makes the paper accessible to control englneers with little or no
that the method that the magnitude of the rotor flux background in motors. We also relate conventional control meth-
ods with modem nonlinear control strategies - proposed
- - recently.
A preliminaly version of this paper was presented at the [,GEE Finally, we develop an algorithm and provide experimental
results that show how the induction motor can be used in demand-
Conference on Decision and Control,December-1993, San Antonio,
ing (high-performance) position control applications.
EX. M.Bodson is with the Department of Electrical Engineering,
The University of Utah, 3262 Merrill Engineering Bldg., Salt Lake Mathematical Model of the Induction Motor
City, UT, 84112. J . Chiasson and R. Novotnakare with the Depart- we now develop the mathematical model for a sinusoidal]y
ment of Electrical Engineering, The University of Pittsburgh, 348 wound, two-phase, single pole-pair induction motor, It is stand-
BmedUm Hall, PitrsbulXh, PA 15261. This work was suPPorred by ard practice, and validated experimentally in [ 131, to also use this
the University of Pittsburgh‘s Center for Motion Control Research model for induction motors with squirrel cage (as opposed to
and its CentralResearch Development Fund as Msell as the PittiCMU sinusoidally wound) rotors. Let isa and iSb denote the currents in
University-1ndustr.y Collegium. phases a and h of the stator, iRa and iRb denote the currents in

August 1994 0212- 1708/94/$04.0001994IEEE 25

Authorized licensed use limited to: Technische Universitaet Muenchen. Downloaded on February 21,2023 at 08:37:55 UTC from IEEE Xplore. Restrictions apply.
 Similarly, the total flux in each of the rotor phases a and b is

hRb = LRiRh + M(-isasin(B) + isbcos(8))

respectively, where LR is the inductance of each rotor phase and

RR is the resistance of each rotor phase. The rotor electrical
equations are given by

- -d 1 ~
-R~
R ~ =R0~
dt

- +
"sb

Fig. I. Induction motor.

phases a and b of the rotor, 8 be the rotor position, o be the rotor

speed and, usu and US^ be the applied voltages to phases a and h
of the stator. These quantities are indicated in Fig. 1, which is
adapted from [2, fig. 4.1.7.al.
The total flux linkage (Le., due to both stator and rotor
currents) in each of the stator phases a and b is given by

The torque can be determined by considering energy conser-

hsa = Lsisa + M ( i ~ ~ c o s (-0 )iRhsin(8))
vation. To do this, expand the rotor electrical equations to get
hsb = Lsisb + M(iR,sins(B) - iRbcos(8))

respectively,where Ls is the self-inductance of each stator phase,

Rs is the resistance of each stator phase, and M is the coefficient
- wM(-isasin(8) + isbcos(8))
of mutual inductance. These expressions for the fluxes hsa, hsb
(and h ~ hRb~ below)
, can be derived as shown in [l]. However,
by studying Fig. 1, the reader should be able to see that they are
quite plausible. By Faraday's and Ohm's Law, we have

The terms on the right-hand side containing o represent the

so-called speed voltages (corresponding to the back-emf in a DC
machine) and are related to the power converted to mechanical
energy (note, for instance, that if the rotor is not turning, then
these terms are zero and the mechanical power 761is also zero).
Multiplying the rotor speed voltages by their corresponding rotor
or phase currents, the electrical power absorbed in rotor phases a
and b is

-iRawM(-isasin(0) + isbcos(8))
and

-iRbwM( -iSaCOS (8) - issin (8))

26 IEEE Control Systems

Authorized licensed use limited to: Technische Universitaet Muenchen. Downloaded on February 21,2023 at 08:37:55 UTC from IEEE Xplore. Restrictions apply.
 respectively. The sum of these two quantities is the absorbed directly. The resulting flux can be measured through the use of
electrical power and is the negative of (Le., reappears as) the Hall sensors placed in the air-gap, but such sensors are not
mechanical power produced 7 0 . Therefore, the electromagnetic economical and diminish the overall reliability of the control
torque is system.
The particular set of nonlinear differential equations (2) de-
scribing the induction motor is complicated and a control strategy
7 = ~[jR~(-js,sin(0)+ jsbCos(0))+ iRb(-isacoS(e) - jshsin(0))l is by no means self-evident. The solution lies in finding an
equivalent set of equations which are simpler in form for which
the control design becomes apparent. As a first step towards
=~ [ i ~ ~ ( - ~ ~ ~ ~- iSa
i iRhsin(8)) ~ (iRasin(0)
( 0 ) + iRhCoS(0))l
,
simplifying the above equations, we rewrite the dynamic equa-
tions in terms of some equivalent rotor fluxes. This results in an
The torque equation is then equivalent model in which the cos(np 0) and sin(np0) expressions
are eliminated. To proceed, define an equivalent set of rotor
fluxes as [l]

(le)

\ V R ~= ~ i n ( n ~ O +) hcos(np0)h~b
~~
where J is the moment of inertia of the rotor and B is the
coefficient of viscous friction. To model a multiple pole motor, = ~ ~ ( s i n ( n ~ O+)cos(np8)i~b)
i~, + Misb.
the angle 0 in equations (la)-( le) is replaced by np 0 where np is
the number of pole-pairs. Collecting the equations together, the
dynamic model of the induction motor is given by v R,b , isa and
Then, in terms of the state variables 0, 0,V R ~
ish, the dynamic equations describing the induction motor be-
come

or, in state-space form (cf., [4], [5]),we have

i~h(i~~cos(n~0)-i~hsin(n~~))

The control problem is to choose usa, usb in such a way as to

force 0 and/or 0 to track a given trajectory/reference. Measure-
ments of isa,& and 0 are usually available for feedback control.
However, measurements of &and &are typically not available.
In fact, acommon type of induction motor is the so-called squirrel
cage motor, where rotor currents are distributed on the surface of
the rotor and the total equivalent current can never be measured

August 1994 27

Authorized licensed use limited to: Technische Universitaet Muenchen. Downloaded on February 21,2023 at 08:37:55 UTC from IEEE Xplore. Restrictions apply.
 with or DQ transformation and we will denote the inverse of this
transformation as IDQ. In the new coordinates, the system equa-
tions are

This system is of the form dvdldt = -qvd + qMid

with (XI, X2, X3, x4, X5, X6) = (0, W, v R a , VRa, &a, k b ) and the
obvious definitions forf(x), gl,82 E R6. The field-oriented and dpidt = npw + qMiqlvd. (7)
input-output controllers require full state feedback which intro-
duces a practical problem. Specifically, the usual situation in
practice is that V R ~vRb(or,
, equivalently iRs iRb) are not available Note that the electromagnetic torque Jpv& is now just
for feedback. Further, if the motor has a position sensor, then it proportional to the product of two state variables. The differential
typically does not have a speed sensor. As will be shown, both equations for id and is still contain quite complicated nonlinear&
of these issues are resolved through the use of observers. ties. A common method method of eliminating these nonlinear
terms is to force the system into a current-command mode [14].
Field-Oriented and Input-Output That is, one uses PI current loops of the form
Linearization Control
As pointed out in [4], [SI, the basis of field-oriented control
consists in considering an equivalent model of (3) found using
the following nonlinear state space transformation (the notation
is consistent with [4], [SI):

to force id and iq to track their corresponding references idr and

iqr, respectively. Assuming these PI current loops result in fast
responses for the currents (accomplished using large feedback
gains), idr and iqr can then be considered as the new inputs and
the system equations may be simplified (approximated) as

along with defining a new set of inputs dvdldt = -qvd + kMidr

dpldt = npO + r\Miq/vd (9)
ud = cos(p)us, + sin(p)usb
uq = -sin(p)us, + COS(p)USb. (6) where idr and iqr are the new inputs. Field-oriented control
consists of using idr to force v d to VdO= Mid0 (constant). As a
consequence, there is a linear relationship between the input iqr
Note that vd and p are just the “polar coordinates” for the and the speed. The control of speed/positionis then done through
ordered pair (VIR,, VRb). The termfield-oriented refers to this new the input iqr by choosing, for
rotating coordinate system whose angular position is p so that vq
= 0. That is, this coordinate system is aligned with the field flux
whose position is p and magnitude is vd.The transformation of
the currents/fluxes/voltages is referred to as the direct-quadrature
iqr = to(
0
(8, -

28 IEEE Control Systems

Authorized licensed use limited to: Technische Universitaet Muenchen. Downloaded on February 21,2023 at 08:37:55 UTC from IEEE Xplore. Restrictions apply.
 As v d o is constant, it is standard practice to incorporate pvdo
directly into the controller gains as

The flux dynamics are now decoupled from the speed dynam-
ics. In summary, with U I ,u2 the inputs and o,v the outputs, the
system ( 1 1) is linear from these inputs to the outputs, hence
input-output linearization has been achieved. The overall system
is still nonlinear since the dynamics of p are nonlinear. The
where stability of p is guaranteed since it is an angle and reset to 0 every
2x.
K,’ = K o / p W d o , K1’ = K I / ~ ~ &K2’
I , = K2Ipvdo The new inputs ul and u2 may be chosen as follows: Let vdr(t)
and a,‘ = a,lpyrdo. The roles of vd and i, (idr and i44 could, in be an arbitrary flux reference. To force the motor’s flux to track
principle, be reversed, but this is not done because the input id’s this reference, choose
(id,.’s) ability to rapidly change vd is limited by the rotor time
constant TR = l/r\ = LR/RR,which is not the case for i, (i,,.).

Field Weakening
The torque is controlled by the quadrature current i,. Inspec-
tion of the fourth equation in (7) reveals that in order to build up
(increase) i,, the voltage uq must be chosen such that diq/dt 0,
asin [1].Let8,(t),or(t)=dOr/dranda,=dor/~tbethemechanical
that is,
reference trajectory the motor is required to track. Choosing

As the quantity OLS is quite small, the dominant term on the

right-hand side of this inequality is MnpWyrd/LR.Thus, to achieve
high speed requires a large input voltage. To get around this
requirement, the flux v d is decreased as the speed o increases.
In particular, idris often chosen to force vd to vary as l/wat higher will result in 8 -+8, and o -+ w, as t - w .
speeds to keep MnpW\Vd/LR from becoming too large and thus Remark. This control algorithm depends on the inner PI
reducing the amount of input voltage required to overcome it. current control loops working satisfactorily. That is, they must
This method of reducing yfd at higher speeds is referred to asfield be given sufficient time, as well as sufficient source voltage, to
weakening . track the current references. As pointed out in the section on field
weakening, the dominant term in the fourth equation of (7) (Le.,
Input-Output Linearization for diq/dt) is the “back-emf’ MnpoWd/LR=n,wyrd. Thus for n,
In the above field-oriented control scheme, the speed 0 and large, a small change in o results in a correspondingly large
flux vd are only asymptotically decoupled [4], [SI.That is, the change in the “back-emf’, making it difficult (Le., takes time) for
the PI current controller to overcome this disturbance in order to
speed is linearly related to iqronly after vd is constant. For high
track the reference 4,.
speeds, field weakening (Le., decreasing vd)is necessary so as
In the experiments reported below on a six-pole motor, these
not to saturate the stator voltages. Since field-weakening depends
PI current loops perform satisfactorily. Simulations have shown
on speed, the dynamics of vd may interfere with the dynamics
that voltage control is only advantageous for machines with a
of o,especially when the speed varies rapidly. This coupling of large number of pole pairs. In that case, for the same trajectory,
the flux and speed dynamics can be eliminated by considering less source voltage may be needed compared to the current-com-
an input-output linearizing controller [4]-[9], [ 111. Specifically, mand input-output controller. (See [ 121 where voltage command
let u1 = idr and define u2 = pv&, Le., Ju2 = J p ~ d i ,is the was essential in achieving high-performance motion control of a
electromagnetic torque. Once u2 is specified in order to accom- SO pole-pair PM stepper motor.) In many cases however, the full
plish the control task, the current reference i,?- is then specified (voltage-command) input-output linearization controller will not
by i,. = u2 / ( p ~ d )The
. system (9) becomes be needed.

Flux Reference
dOidt = w Aprocedure for selecting the flux reference at any given speed
to achieve optimal torque (maximum for acceleration and mini-
doldt = u2 - (Bl.l)w mum for deceleration) is given by the authors in [ 151. As shown
there, if the motor speed of our test motor is below approximately
3000 rev/min, then the optimal flux reference is constant with

August 1994 29

Authorized licensed use limited to: Technische Universitaet Muenchen. Downloaded on February 21,2023 at 08:37:55 UTC from IEEE Xplore. Restrictions apply.
 Speed Observer
In a position (servo) control sys-
A tem, typically one has a position
\' Flux 0 measurement, e.g., from an optical
P
encoder, but a speed measurement is
i
u --u 5. not available. It is standard practice in
sa
- 2 IDQ
- Inductionj 'sb.: industry to compute the speed by a
Motor
U
discrete differentiationof the position
q- 'Sb
output from the optical encoder as
A A A
o(k)=(O(k)-O(k-l))/Twhere Tis the
sample period. Using a typical 2000
I I pulse/rev encoder, the error in this
estimate is tightly bounded by
(27~/2000)/T.This noise is particu-
- u d = K P I i d r - l d )+ KIJ:(idr- I d ) d t larly significant at high sample rates
and moderate to low speeds (since
I I
less encoder counts are detected per
*d t sample period than at higher speeds).
For example, in [20] (using the same
motor as in this work), the authors
give experimental results of tuming
the motor 180" in 80 ms where the top
speed of the motor was less than 65

I
I

er U
I

= der/dt
I

u = dwp/dt t
I
rad/s and a sample rate of 8 kHz was
required. In this circumstance, the er-
ror bound on the speed computed by
differentiation is (2~c/2000)8000= 25
radls and such an estimation scheme
is inadequate. For the DC motor, an
Fig.2. Current-command input-output servo (position) controllei:
accurate speed measurement is easily
found using an observer. This moti-
vates one to consider a speed observer
Flux Observer for the induction motor. In the field-oriented coordinates, the
An observer for p and yfd based on the model (7) is now equation governing the speed in an induction motor is given by
constructed as shown in [ 11. The estimates for p and yfd are doidt = pvdi,-(B/J)o. The quantities yfd and i, are not known,
defined as the real-time solutions to but can be estimated as shown above. Thus, consider an observer
defined by

= -q& + qM(isacos(b) + ig,sin(b)). (14)

Aerotech
PC Motorola Data
Induction
The convergence of the error dynamics of this observer is
limited by the rotor time constant TR =LR/RR= l/q [ 101.Verghese
and Sanders [lo] have proposed clever ways to allow one to
specify an arbitrary rate of convergence. However, the motor
used in our experiments has a small rotor time constant (less than
4 ms) so that the above observer suffices. The rotor resistance Fig.3. Experimental setup.

30 IEEE Control Systems

Authorized licensed use limited to: Technische Universitaet Muenchen. Downloaded on February 21,2023 at 08:37:55 UTC from IEEE Xplore. Restrictions apply.
 Althou h one cannot analytically guarantee the convergence of
f . A A &
o + 0,since y d , p and thus, = -isasin($) + isbcos($) are
estimates, it tums out in practice that this estimate of o is much
350 -
better than plain differentiationof the position [21]. This estimate
30)- is then used in the feedback (13) and flux observer (14). A block
diagram of the Current-Command Input-Output Linearization
250 -
Controller is given in Fig. 2.
200 -
Experimental Results
150 - A 6-pole 1/12 horsepower two-phase induction motor with a
squirrel cage rotor was used for experimentation. It is rated for
100 -
2.4 A (continuous) and 60 V. The hardware setup (Fig. 3) is
50 - identical to that used by the authors in [12], consisting of a
Motorola DSP56001 digital signal processing system, two P W
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 amplifiers (+SO V and k6 A), A/Ds and D/As with associated
electronic interface boards and an induction motor equipped with
a 2000 pulse/rev encoder (resolution of 360'/2000 = 0.18"). The
Fig. 4 . Measured (solid) and reference (dashed)position in radians parameters were identified using the technique given in Stephan
versus time in seconds.
et al. [13], and were found to be M = 0.011 H, RR = 3.9 R, Rs =
1.7 R, LR = 0.014 H, Ls = 0.014 H, B = 0.00014 N-m/rad/s, and
J = 0.00011 kg-m2. The observer gains were set as 11 = 1 . 8 ~ 1 0 ~
and 12 = 8 ~ 1 so0 that
~ the poles of the observer error system are
-800 and -1000. This choice of gains was determined experimen-
tally and found to be a good compromise between the desire for
a fast rate of decay of the estimation error and avoiding the
amplification of high-frequency noise.
An experiment requiring a point-to-point position move was
camed out in which the motor was brought up to a speed of 8000
rev/min in 0.38 s and brought down from 8000 rev/min to 0
rev/min in 0.265 seconds. The PI current gains in (8) were set at
K d l = 9000, Kdp = 15, K4[ = 9000, K4p = 15. The PID gains for
the tracking error in (10) were chosen as KO = 3.Ox1O5, K1=
5.5x104, K2 = 125. The sample rate was 4 kHz. Fig. 4 below
shows tracking of the measured position to its reference. More
-50 1 1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 detail on the tracking is found in Fig. 5, which is a plot of the
position error in encoder counts.
Fig. 5. Position error in encoder counts versus time in seconds. Note in Fig. 5 that the final position error is zero at the end of
the run which is required for a point-to-point move. Comparing

800-
700 -
600-
500-

40)-

7/
200
-0.21
I

0'
loo > 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 -0.6I
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7

Fig. 6 . Estimated (solid) and reference (dashed) speeds in radians Fig. 7. Optimum ( s o l i d ) and reference (dashed) torque in
per second versus time in seconds. Newton-meters versus time in seconds.

August 1994 31

Authorized licensed use limited to: Technische Universitaet Muenchen. Downloaded on February 21,2023 at 08:37:55 UTC from IEEE Xplore. Restrictions apply.
 We also attempted these high-speed experiments using a
field-oriented controller (including field weakening) and were
0.05 unable to achieve tracking as the phase voltages saturated. We
0.045 then performed simulations and the same result occurred. Con-
sequently, the input-output linearization controller is essential for
0.04
obtaining high-performance.
0.035
0.03 High-Performance Motion Control
0.025
We have shown that a current-command input-output lineari-
zation controller can achieve high-performance motion control,
0.02 that is, the precise tracking of a fast point-to-point position
0.01 5 reference. Specifically, this controller was shown to provide the
0.01 means of decoupling the speed and flux dynamics in an induction
I’ motor. This decoupling of speed and flux was exploited to
0.005
simultaneously track the position/speed reference and an optimal
Oi 0:l 0:2 0:3 ’ 0:4 0:5 0:s 0:7 flux reference. This flux reference was used to obtain the optimal
(max/min) motor torque at any given speed without violating
voltage and current limits [15]. Experimental results were pre-
Fig. 8. Estimated yfd (solid)and reference (dashed) y~d,-in uvbers sented to demonstrate the effectiveness of this scheme.
versus time in seconds
Acknowledgment
The authors thank Aerotech, Inc., for their generosity in
supplying the hardware used to carry out this research. They also
thank Jerry Frank of Cleveland Machine Controls and the Center
for Motion Control Research under the directorship of Prof. E.W.
Kamen for supporting the early stages of this work. The
Pitt/CMU University-Industry Collegium is gratefully acknow-
ledged for their on-going support of this work. J. Chiasson also
thanks Prof. Wemer Leonhard of the Technical University of
Braunschweig and his Ph.D. students for their helpful discus-
sions on induction motor control during his visit to their Institute.
Finally, the Motorola Corporation is gratefully acknowledged for
donating the DSP56001 Development System used in this re-
search.

References
[ 11 W. Leonhard, Control of Elecrricul Drives. New York: Springer-Verlag,
1985.
[21 H.H. Woodson and J.R. Melcher, Electromechanical Dynamics, Pt. 1.
Fig. 9. Estimated i, and iqvin amperes i’ersus rime in seconds New York: Wiley, 1968.
[3] Z. Krzeminski, “Nonlinear control of the induction motor,” in Proc. loth
with Fig. 7, it is seen that the maximum errors occur when the IFAC World Congress, Munich, Germany, 1987, pp. 349-354.
acceleration or deceleration are at a maximum. This is especially [4] R. Marino, S. Peresada, and P. Valigi, “Adaptive partial feedback lineari-
true at about 0.1 s where, in addition, the jerk (Fig. 7) is discon- zation of induction motors,” in Proc. 29th Cot$ Decision and Control,
tinuous. Honolulu, HI, Dec. 1990.
Fig. 6 is a plot of the estimated speed and the reference speed. [SI R. Marino, S. Peresada, and P. Valigi, “Adaptive input-output linearizing
Note the excellent speed tracking despite the time-varying flux control of induction motors,”IEEE Trans. Auto. Control, vol. 38, Feb. 1993.
(Fig. 8). [6] R. Marino and P. Valigi, “Nonlinear control of induction motors: A
The dashed line in Fig. 7 is a plot of the reference torque simulation study,” in Proc. 1991 Euro. Conrrol Conf., Grenoble, France,
corresponding to the position and speed references given in Figs. 1991.
4 and 6. The solid line is a plot of the optimum torque achievable [7] A. Sabanovic and D.B. Izosimov, “Application of sliding modes to
based on the results in [15]. Note the closeness of the reference induction motor control,” IEEE Trans. Ind. App., Jan./Feb. 1981.
to the optimum.
[8] G.S. Kim, I.J. Ha, and M.S. KO,“Control of induction motors for both
Fig. 9 is a plot of the estimated i, along with its reference zqr. high dynamic performance and high-power efficiency,” IEEE Trans. Ind.
Figs. 8 and 9 indicate that vd and i, vary significantly, yet the Elec.. VOI. 39, pp. 323-333, Aug. 1992.
input-output controller forces the resultant torque JFyfni, to [9] J. Chiasson, A. Chaudhari, and M. Bodson, “Nonlinear controllers for the
provide smooth tracking of the position and speed as shown in induction motor,” in Proc. IFAC Nonlinear Conrrol System Design Symp.,
Figs. 4 and 6, respectively. Bordeaux France, June 1992, pp, 150-155.

32 IEEE Control Systems

Authorized licensed use limited to: Technische Universitaet Muenchen. Downloaded on February 21,2023 at 08:37:55 UTC from IEEE Xplore. Restrictions apply.
 [lo] G.C. Verghese and S.R. Sanders, “Observers for flux estimation in degree in electrical engineering and computer sci-
induction machines,” IEEE Trans. Ind. Elec., vol. 35, pp. 85-94, Feb. 1988. ence, and the M.S. degree in aeronautics and astro-
nautics from The Massachusetts Institute of
[ l l ] A. Chaudhari, “Analysis of state-space controllers for an induction
Technology, Cambridge, MA, in 1982, and the elec-
motor,” M.S. thesis, Univ. Pittsburgh, 1992.
trical and mechanical engineering degree from the
[I21 M. Bodson, J. Chiasson, R. Novotnak, andR. Rekowski, “High perform- Universite de Libre de Bruxelles, Belgium, in 1980.
ance nonlinear control of a permanent magnet stepper motor,” IEEE Trans. Currently, he is an Associate Professor in the De-
Control Syst. Technol. Mar. 1993. partment of Electrical Engineering, The University
of Utah, Salt Lake City, UT. He is coauthor with S.
[13] J. Stephan, M. Bodson, and J. Chiasson, “Real-time estimation of
Sastry of the book, Adaptive Control: Stability, Con-
induction motor parameters,” in Rec. 27th Ann. Meet. Ind. App. Soc., Hous-
t,ergence,and Robustness (Prentice-Hall, Englewood Cligs, NJ, 1989.) He
ton, TX, Oct. 1992; also IEEE Trans. Ind. App., vol. 30, May/June 1994.
was a Belgian American Educational Foundation Fellow in 1980, a Visiting
[ 141 Prof. Werner Leonhard, private communication. Lecturer at the University of California, Berkeley, in 1987, and a Lady Davis
[15] M. Bodson and J. Chiasson, “Asystematic approach to selecting optimal Fellow at the Technion-Israel Institute of Technology, Haifa, Israel, in 1990.
flux references in induction motors,” in Rec. 27th Ann. Meet. Ind. App. Soc.,
Houston, TX, Oct. 1992. John Chiasson received the B.S. degree in mathe-
[ 161 J. Chiasson, “Dynamic feedback linearization of the induction motor,”
matics from the University of Arizona, the M.S.
IEEE Trans. Auto. Control, vol. 38, Oct. 1993. degree in electrical engineering from Washington
State University, and the Ph.D. degree in control
[17] R. Ortega and G. Espinosa, “An exponentially convergent controller for sciences from the University of Minnesota. Since
induction motors,” in Proc. 1991 Euro. Control Conf.. Grenoble, France.
1988 he has been an Assistant Professor in the
[I81 R. Ortega and G. Espinosa, “Passivity properties of induction motors,” Department of Electrical Engineering at the Univer-
in Proc. 1991 Con5 Ind. App., Dearborn, MI. sity of Pittsburgh. His current area of research inter-
[I91 I. Kanellakopoulos, P.T. Krein, and F. Disilvestro, “Nonlinear flux-ob- est is in the applications of geometric nonlinear
server-based control of induction motors,”inProc. 1992Amer. CnntroIConf , control theory to electromechanical systems.
June 26-28, Chicago IL.
Robert T. Novotnak received the B.S. dnd M.S.
[20] M. Bodson and J. Chiasson, in notes for “Tutorial on induction motor
degrees in electncal engineering from the Univer-
control,” presented at Pitt/CMU University-Industry Collegium, Univ. of
sity of Pittsburgh, Pittsburgh, PA, in 1989 and 1991,
Pittsburgh, July 20, 1994.
respectively. He IS currently enrolled in the Ph.D.
[21] M. Bodson, .I.
Chiasson and R. Notvotnak, “Nonlinear speed observer program at the University of Pittsburgh. His inter-
for high-performance induction motor control,” submitted for publication ests included the practical issues involved in imple-
Marc Bodson received the Ph.D degree in electncal engineering and com- menting nonlinear control algorithms for electrical
puter science from the University of California, Berkeley, in 1986, the M S. drives

August 1994 33

__~
I

Authorized licensed use limited to: Technische Universitaet Muenchen. Downloaded on February 21,2023 at 08:37:55 UTC from IEEE Xplore. Restrictions apply.