You are on page 1of 9

By Hyeung-sik Choi, Yong-heon Park,

Yongsung Cho, and Minho Lee

Improved Design for a Brushless DC Motor

S
liding-mode control (SMC) has
long been known as a useful
strategy for controlling uncer-
tain systems. In the early
1960s, SMC originated from
the variable-structure control
system proposed and elaborated in the So-
viet Union by Emelyanov and Ikis [1], [2]. Its
invariance properties and robustness
against uncertain system parameters and

© Powertec Industrial Motors, Inc.


disturbances were studied extensively.
Utkin found that SMC can completely reject
uncertainties and disturbances that satisfy
the matching condition when the state is
constrained to the sliding surface [3]. In [4],
an extensive survey of SMC was undertaken
where fundamental theory, main results,
and practical applications of variable-
structure control were discussed.
To achieve fast path tracking, an improved sliding-mode using an optimal linear regulator technique. To improve the
control employing an optimal sliding surface was proposed tracking behavior of nonlinear second-order dynamical sys-
by Ashchepkov [5]. The optimal sliding surface was deter- tems, a moving sliding surface was proposed by Choi and
mined by minimizing the error performance index for a Park [7]. The surface was initially designed to pass given ini-
given initial condition. Chern and Wu [6] proposed an opti- tial errors and subsequently resulted in a moving sliding
mal variable-structure control with integrated compensa- surface. The robustness of the sliding-mode controller to
tion; an optimal gain matrix for the sliding line was designed uncertainties and disturbances may not be preserved dur-

Choi (hchoi@hanara.kmaritime.ac.kr) and Park are with the Division of Mechanical and Information Engineering, Korea Maritime Uni-
versity, Busam, Korea. Cho is with Dong-A University, Busam, Korea. Lee is with Kyungpook National University, Taegu, Korea.

0272-1708/01/$10.00©2001IEEE
June 2001 IEEE Control Systems Magazine 27
perior performance of the proposed con-
The advantage of the controller is trol scheme is validated by computer
simulation and by experimental compari-
that the arrival time can be expressed sons with other sliding-mode controllers.

in a closed-form equation. Definition of the System


The linear time-varying, second-order sys-
ing the reaching phase. To reduce the reaching phase dura- tem with parameter uncertainties and disturbances is de-
tion, an SMC was proposed in which the components of the fined as
gradient vector of the switching hypersurface were adapted
to the inverse of the tracking error [8]. x&& + a1( t )x& + a2( t )x = b( t )(u + d( t )), (1)
Despite its invariance properties and its robustness,
conventional SMC has important drawbacks limiting its ap- where the upper and lower bounds of uncertain system pa-
plicability: large control input torques [9]. The robustness rameters a1( t ), a2( t ), b( t ), and disturbance d( t ) are speci-
of the conventional SMC is based on the assumption of fied as follows:
ideal large control torques; the robustness properties are
guaranteed only as long as the control torques do not satu- β min ≤ b −1( t ) ≤ β max 
rate. Adhering to input torque limitations is very impor- α 1 min ≤ b ( t )a1( t ) ≤ α 1 max 
−1

α 2 min ≤ b −1( t )a2( t ) ≤ α 2 max 


tant in realistic and practical applications and should not .
be neglected. To solve the robustness problem of SMC un- 
max d( t ) < D 
der input torque limits, a scheme to estimate the region of t  (2)
the asymptotic stability with bounds on the control inputs
was proposed in [10], but this scheme results in excessive The input torque is bounded, since all physical systems
chatter. Lu and Chen devised a global sliding-mode control have input torque limits, and is specified as
(GSMC) scheme that ensures sliding behavior throughout
an entire response [11]. With given uncertainty bounds, U min ≤ u ≤ U max . (3)
the maximum and minimum values of the control action
are estimated according to allowable reference inputs by The improved GSMC algorithm proposed to drive the states
the control scheme. of the uncertain system with bounded input torques along
In addition to extensive research on the control algo- the minimum trajectory is described in the next section.
rithm, much attention has been paid to applications of SMC.
One of the application areas is an ac or brushless dc (BLDC) Design of the Improved GSMC for a
motor. Hashimoto et al. applied the conventional vari- Time-Varying, Second-Order System
able-structure control to a BLDC motor system containing
system uncertainties [12]. Park et al. applied a sliding-mode Description of the GSMC
control with observer to reduce steady-state error and to es- For the uncertain system defined in (1), a GSMC with esti-
timate the acceleration of the ac servomotor [13]. mates of the uncertain parameters is described as
In this article, we propose an improved GSMC design for
controlling an uncertain second-order, time-varying sys- ( )
u = β$ f& − cx& + a$ 1 x& + a$ 2 x
tem with torque limits. By applying the control scheme to
the second-order system, tracking along the minimum time {
− ∆β f& − cx& + ∆α 1 x& + ∆α 2 x + D sgn( s ) } (4)
trajectory can be achieved and the minimum arrival time
to reference inputs can be estimated if ranges of paramet- where
ric uncertainties, torque limits, and reference inputs are
β + β min β − β min
specified. The eminent advantage of the proposed control- β$ = max , ∆β = max
ler is that the arrival time can be expressed in a 2 2
closed-form equation. The proposed controller was imple- α + α 1 min α − α 1 min
α$ 1 = 1 max , ∆α 1 = 1 max
mented on an uncertain BLDC motor with unknown but 2 2
bounded loads. When the proposed control scheme is ap- α + α 2 min α − α 2 min
α$ 2 = 2 max , ∆α 2 = 2 max .
plied to the BLDC motor system, the arrival time at the ref- 2 2
erence revolution angle and the maximum allowable
acceleration are expressed in a closed-form equation. The The global sliding mode is defined in conjunction with f ( t )
closed-form equation can be utilized in designing or select- as
ing the capacity of motor systems if bounds on uncertain
parameters, loads, and target angle are specified. The su- s = e& + ce − f ( t ), (5)

28 IEEE Control Systems Magazine June 2001


where the error state is e = x − r with constant reference in- a 2 a 
x =r− t f + at f t − t 2
put r. If the initial condition is satisfied, the forcing function 2 2 

drives the system states in any state space directly to the x& = a( t f − t )  for t b ≤ t ≤ t f ,
switching plane without a reaching phase. The forcing func- &&
x = −a 

tion is a desired trajectory function for the system. We pro-  (8b)
pose a forcing function to drive the system states along the
minimum time trajectory, which is designed within the in- where v is the velocity, a is the acceleration, and t b is the me-
put torque limit. By applying the proposed controller, we dian travel time. The profiles of the position, velocity, and
can calculate the minimum arrival time at the reference in- acceleration of the trajectory are shown in Fig. 1.
put in closed form. The forcing function driving the sliding mode along the
The response of the conventional SMC is sensitive to sys- minimum time trajectory is described as
tem perturbations during the reaching phase. One of the ad-
vantages of the GSMC is that it can have sliding-mode a 
f ( t ) = x& + c( x − r ) = at + c  t 2 − r  for 0 ≤ t < t b
characteristics over the entire range without a reaching 2  (9a)
phase. For this, the conditions of the forcing function f ( t )
should be satisfied; that is
and

f (0 ) = e&0 + ce0 (6a)   1 1 


f ( t ) = a ( t f − t ) + c  t f t − t f2 − t 2  for t b ≤ t ≤ t f .
  2 2  (9b)

f ( t ) → 0 as t → ∞ (6b) The states of the system follow the minimum time trajec-
tory when the controller reflecting the proposed forcing func-
tion is applied to the system. Therefore, we can estimate the
f&( t ) exits and is bounded, (6c) minimum arrival time at the reference input and the maxi-
mum acceleration within the input torque limit. The estima-
where e0 = e( t = 0 ) and c > 0. In (6), condition (6a) repre- tion process is described in the following section.
sents the initial location of the states on the sliding surface,
(6b) represents asymptotic stability, and (6c) represents
Estimation of the
the existence of the sliding mode. The stability of the pro-
posed GSMC satisfying the conditions in (6) can be proved
Minimum Arrival Time
using the Lyapunov function V = (1 / 2)s 2 > 0. The negative
Estimation of the Maximum and
definiteness of the time derivative of V except for s = 0 en-
Minimum Input Torques for the
sures that the proposed control scheme guarantees asymp-
totic stability. The proof of the stability of the closed system
Second-Order System
The input torques of the system with unknown parameters
is simple and has been shown in [11].
and uncertain disturbances defined in (2) are bounded as
specified in (3). Within the bounded range, we propose a
scheme to estimate the minimum arrival time for the sec-
Design of the Forcing Function
ond-order, time-varying system if a constant reference input
We propose a function representing the minimum time tra-
is given. To do this, we divide the controller into two parts
jectory that satisfies the conditions in (6). The initial and fi-
according to its magnitudes. The magnitudes are dependent
nal conditions of the function are specified as
on the sign of s as follows:

for t = 0: x (0 ) = 0 , x&(0 ) = 0 

for t ≥ t f : x ( t ) = r , x&( t ) = 0 
( )
u h = β$ f& − cx& + a$ 1 x& + a$ 2 x
(7)
{
+ ∆β f& − cx& + ∆α 1 x& + ∆α 2 x + D } for s < 0

where t f is the final arrival time. According to the boundary (10a)


conditions, the minimum time trajectories are described as
( )
u l = β$ f& − cx& + a$ 1 x& + a$ 2 x
a
x = t2
2

 {
− ∆β f& − cx& + ∆α 1 x& + ∆α 2 x + D } for s > 0.
 (10b)
x& = at  for 0 ≤ t ≤ t b and
When the controller performs appropriately such that sys-
v
x&& = a =  tem states track the trajectory without errors, s = s& = 0 is
tb 
(8a) satisfied for t ≥ 0. By exploiting (5), we have

June 2001 IEEE Control Systems Magazine 29


x&& = f& − cx&. (11) u h = aW + D for 0 ≤ t < t b 
,
u h = aX + D for t b ≤ t ≤ t f 
Rewriting (10) using (11) yields (14)

where
u h = β$ x&& + a$ 1 x& + a$ 2 x + {∆β x&& + ∆α 1 x& + ∆α 2 x + D} (12a)
1
u l = β$ x&& + a$ 1 x& + a$ 2 x − {∆β x&& + ∆α 1 x& + ∆α 2 x + D}. (12b) W = β max + α 1 max t + α 2 max t 2
2
1 1 
X = −β min + α 1 max ( t f − t ) − α 2 max  t f2 − t f t + t 2 .
Equation (12) can be rearranged according to the input 4 2 
profiles and trajectory tracking time. Equation (12a) is ex-
pressed as
Using the same approach, substituting the functions of the
& and x&& of (13b) and rear-
minimum time trajectory into x, x,
u h = β max x&& + α 1 max x& + α 2 max x + D for 0 ≤ t < t b 
, ranging yields
u h = β min x&& + α 1 max x& + α 2 max x + D for t b ≤ t ≤ t f  (13a)
u l = aY − D for 0 ≤ t < t b 
where the input torques are obtained by exploiting the input ,
u l = aZ − D for t b ≤ t ≤ t f 
profiles shown in Fig. 1. For the time duration 0 ≤ t < t b , we (15)
have x&& = x&& = a, and for t b ≤ t ≤ t f , we have x&& = − x&& = − a.
Using the same procedure, (13b) is expressed as where

u l = β min x&& + α 1 min x& + α 2 min x − D for 0 ≤ t < t b   1 


. Y = β min + α 1 min t + α 2 min t 2
u l = β max x&& + α 1 min x& + α 2 min x − D for t b ≤ t ≤ t f  (13b)  2 
1 1 
Z = −β max + α 1 min ( t f − t ) − α 2 min  t f2 − t f t + t 2 .
To estimate the maximum of the input torques, we apply (8) 4 2 
& and x&& of (13a), which yields
with r = 1 4 at f2 to x, x,

r The maximum of u h in (14) and the minimum of u l in (15) ex-


ist at certain times in the tracking time range. Using the max-
imum and minimum values of the input torques obtained,
we can estimate the maximum acceleration of the uncertain
system with bounded input torques. We express the maxi-
mum and minimum values of (14) and (15) as follows:

0 tb tf t max u h ( t ) = aTmax + D, (16)


t≥0
(a)

v
min u l ( t ) = aTmin − D. (17)
t≥0

For realistic and practical application, the input torques


should be designed within the physical bounds, since physi-
cal bounds always exist on all systems. Therefore, the maxi-
mum and minimum input torques are bounded as
0 tb tf t
(b)
U min ≤ min u l ( t ) ≤ max u h ( t ) ≤ U max . (18)
a To estimate the minimum arrival time, we need to obtain
the maximum value ofu h in (16) and the minimum value ofu l
0 in (17). For the continuous control input region, we can dif-
ferentiate u h and u l with respect to time. At the initial, me-
dian, and final times, we can only obtain the limit values.
−a Following this approach, we obtain several maximum and
tb tf t
minimum candidates for input torques. The maximum or
(c)
minimum values are obtained from these candidates. To get
Figure 1. The minimum time trajectory: (a) position, (b) velocity, the maximum value within the differentiable range, we dif-
and (c) acceleration. ferentiate (14) and (15) with respect to time as

30 IEEE Control Systems Magazine June 2001


du h   1 
= α 1 max a + α 2 max at for 0 ≤ t < t b ul = a −β max + α 1 min t b + α 2 min t b2  − D
dt  t = tb + 0  2 
du h . (24c)
= −α 1 max a + α 2 max a( t f − t ) for t b ≤ t ≤ t f 
dt  (19)

In (19), for each time duration, we obtain the solution of


du h dt = 0 for t ≥ 0 as follows:  1 
ul = a −β max + α 2 min t f2 − D.
t=t f  4 
(24d)
α
t hd = t f − 1 max for t b ≤ t ≤ t f .
α 2 max (20) Since the value of (24b) is larger than that of (24c), the mini-
mum value of the control torques is selected among the can-
Substituting the solution in (20) into (16) yields didates as follows:

uh = aTmax + D for t b ≤ t ≤ t f . (21a)


{ }
t = t hd t = t hd
min u ( t ) = min u l t = t , u l t = tb + 0 , ul t=t f for t b ≤ t ≤ t f .
ld
At the initial, final, and midpoint of tracking time, the values
of the control torques are expressed as (25)

The goals of the proposed control scheme are to achieve


uh t=0
= aβ max + D (21b)
tracking of the trajectory in minimum time and to esti-
mate the minimum arrival time at the reference input in
the second-order, time-varying system with unknown but
 1 
uh = aβ max + α 1 max t b + α 2 max t b2  + D bounded parameters and disturbances. An advantage of
t = tb − 0  2  (21c) the proposed control scheme is that we can obtain the
closed-form solution of the minimum trajectory arrival
time t f , since all the candidate equations are expressed in
 1  second-order polynomials. Using the closed-form solution,
uh = a −β min + α 1 max t b + α 2 max t b2  + D
t = tb + 0  2  (21d) we can estimate the minimum arrival time easily without
numerical approaches.

u h t = t = a( −β min + α 2 max t b2 ) + D. (21e)


f
Estimation of the Minimum Arrival Time
In (21), since the values of (21b) and (21d) are obviously of the Motor System
smaller than that of (21c), the maximum exists among (21a), In electric motor systems, the stiffness coefficient is not
(21c), and (21e), which is expressed as considered such that a2 = 0. To derive the closed-form solu-
tion of the equation, we analyze (20)-(25). Since α 2 is zero, a
max u ( t ) = max u h { t = t hd
, uh t = tb − 0
, uh t=t f } for 0 ≤ t < t b
solution does not exist in (20). By evaluating the values of
(21c) and (21d), we obtain the maximum value as
(22)
In the same way as shown for obtaining the maximum value,
the minimum value candidates of u l can be obtained. By re- max u ( t ) = u h t = tb − 0
for t ≥ 0. (26)
ferring to the trajectory profiles, we can get a minimum can-
didate t ld by differentiating u l as
In the minimum value candidates, since α 2 = 0, a solution
does not exist in (23). By evaluating the last two equations,
α 1 min
t ld = t f − for t b ≤ t ≤ t f . (24c) and (24d), we can tell thatu l|t = t f is the minimum, which
α 2 min (23) is expressed as

Substituting t ld in (23) into (17) yields


min u ( t ) = u l t=t f
for t ≥ 0. (27)
ul t = t ld
= aTmin t = t ld
−D for t b ≤ t ≤ t f . (24a)
From the minimum and maximum values, we can deter-
Other minimum candidates at the median and final mine the minimum arrival time. Using the trajectory function
points of the tracking time can be obtained as r = (1 / 4 )at f2 , we can obtain a = 4 r /t f2 . By substituting this
equation into (21c) and (24d) and arranging them within the
ul t = tb − 0
= a(β min + α 1 min t b + α 2 min t b2 ) − D (24b) bounded region in (3), we obtain the following equations:

June 2001 IEEE Control Systems Magazine 31


4r As shown in (32) and (33), the minimum arrival time is ex-
(β max + α 1 max t b ) + D ≤ U max
t f2 (28) pressed in closed form. Therefore, we can easily calculate
the minimum arrival time at the reference input (revolution
angle). This means that we can easily design or select appro-
4r priate motors to drive the mechanical system according to
U min ≤ − β max − D.
t f2 (29) the control specifications.

Rearranging (28) yields a second-order inequality: Computer Simulation and


Implementation
(U max − D )t f2 − 2rα 1 max t f − 4 rβ max ≥ 0. (30)
Model of a BLDC Motor System
In the computer simulation and implementation, we apply
By solving (30), the minimum time candidate to arrive at ref-
the proposed controller to the permanent magnet BLDC mo-
erence input r can be obtained in closed form as
tor with unknown but bounded parameters, loads, and dis-
turbances and show that the proposed control scheme is
r α 1 max + r 2α 21 max + 4 rβ max (U max − D )
t h min = . valid and superior to other conventional SMCs. We also
U max − D (31) show that the minimum arrival time estimated from the de-
rived closed-form equation is similar to the simulation and
In the same way, another minimum time candidate can be experimental results. For this, the dynamics of the BLDC
obtained by solving the inequality (29) as follows: motor system are described as

4 rβ max
t l min = . θ&& + a1θ& = b(u + d ), (34)
−U min − D (32)
where θ is the position angle, θ& = dθ dt, a1 = B / J is com-
The minimum arrival time should be either of t h min and t l min
posed of the viscous-friction coefficient B and an unknown
satisfying both torque bounds in (28) and (29), which is:
but bounded J consisting of the rotor inertia and load, and
b = K t K c /J factors in the motor torque coefficient K t and the
t min = max{t h min , t l min }. (33)
PWM inverter current coefficient K c . The control input u is
the voltage input.
Usually, the motor employs a
third-order model with the voltage in-
AC
put. In electrical dynamics, however,
Load
the inductance L is much smaller than
Current the resistance R ( L R ≈ 0.001) such
Command c
ACL that its mode decays quickly in the
Pentium 726 PWM BLDC
Motor transient state. The third-order dy-
586 Inverter
Encoder namics of the motor become second-
order dynamics in a short time. To
Proposed compensate for the dynamic effect, we
GSMC employed a conversion coefficient K t ,
Controller PCL which cannot be exactly measured but
833 is bounded. Using this, we modeled the
BLDC motor as second order with an un-
Figure 2. Implementation structure of the proposed GSMC for a BLDC motor. certain conversion coefficient reflecting
electric dynamics.
Table 1. Experimental and simulation values of control schemes. The parameter value of
2 K t is given in the catalogue
Method Controller t f [s] c a[rad/s ] Maximum Input
but varies according to the
Calculation GOSMC 0.635 5.0 supply voltage. J and K c are
roughly measured and B is
Simulation GOSMC 0.635 7.878 623.897 5.017
estimated. The parameters
Experiment GOSMC 0.647 7.878 623.897 5.021 are difficult to measure ac-
Experiment GSMC 0.700 7.878 623.897 3.972 curately, and they are speci-
fied with about 15% error
Experiment SMC 0.760 7.878 5.640
ranges in the simulation and

32 IEEE Control Systems Magazine June 2001


experiments. The bounds on the uncertain parameters and r = 20π (rad). In the simulation and experiments, the same
disturbances are specified as control gains, parameters, and uncertainty bounds were
used. For the simulation, we used the Runge-Kutta
4140
. × 10 −5 ≤J ≤ 6.210 × 10 −5 ( Kg − m 2 ) fourth-order numerical method in the C programming lan-
1314
. × 10 −1 ≤ Kc ≤ 1971
. × 10 −1 ( A/V ) guage. We derived the closed-form equation for the arrival
.
2902 × 10 −1 ≤ Kt ≤ 4.354 × 10 −1 ( N − m/A ) time in (33). Using the closed-form equation, we estimated
the minimum arrival time at the given reference input, the
3.814 × 10 −2 ≤ K c K t ≤ 8.581 × 10 −2 ( N − m/V )
maximum acceleration, and the maximum control torques
5.078 × 10 −4 ≤ B≤ 7617
. × 10 −4 ( Kg − m 2 /s)
according to the specified parameters and disturbances
D = 0.03 (N − m ) (see Table 1).
We compared the performance of the proposed control-
Ranges of uncertain parameters for the control inputs are ler with the GSMC and the SMC and simulation results using
specified as

4.824 × 10 −4 ≤ β ≤ 1628
. × 10 −3 , Conventional SMC has important
5.917 × 10 −3 ≤ α 1 ≤ 1997
. × 10 −2 ,
β$ = 1055
. × 10 −3 , ∆β = 5.729 × 10 −4 ,
drawbacks limiting its applicability:
α$ 1 = 1294
. × 10 −2 , ∆α = 7027
. × 10 −3. large control input torques.
To implement the proposed controller, we
the same control gains, parameters, and uncertainty
used an IBM 586 personal computer as a processor, a
bounds. We evaluated two aspects of the performance of the
PCL-722 data acquisition board to count the encoder signal,
proposed control, the reference input tracking capability
and a PCL-833 digital-to-analog conversion board. The limit
and the obedience to torque limits, and also compared the
of the control input to the motor driver is ±5 V. The applied
input voltage is again transformed into the control torque
through K t and K c . The hardware system loop is shown in
Fig. 2. The control gain is set as c = 73787
. for the experimen-
5
tal case of SMC, GSMC, the proposed control, and simula-
tion. The conventional SMC is constructed as

{ }
τ = −β$ cθ& + a$ θ& − ∆β cθ& + ∆α θ& + D sgn( s ) − K d s,
Input [v]

(35)

where s = e& + ce and kd = 0.05. The sampling time is set to


0
0.905 ms.

Results of the Computer Simulation


and Implementation 0.0 0.5 1.0
Time [s]
We performed one simulation and three experiments for the (a)
proposed control, GSMC, and SMC with reference input

62.8319
Input [v]
Trajectory [rad]

Simulation
Experiment 0.0 0.5 1.0
0.0000 Time [s]
0.0 0.5 1.0 (b)
Time [s]
Figure 4. Control inputs with proposed GSMC. (a) Experiment
Figure 3. Trajectory tracking with proposed GSMC. and (b) simulation.

June 2001 IEEE Control Systems Magazine 33


estimation result of the derived closed-form equation for
the arrival time with that of the proposed controller, those
62.83185
of other SMCs, and simulation results.
All the estimated values, simulation and experimental re-
sults are shown in Figs. 3-7and Table 1. In Figs. 3 and 5, trajec-
Trajectory [rad]

tory tracking results are shown, and in Figs. 4, 6, and 7,


applied control inputs are shown. According to the simula-
tion and experimental results shown in Figs. 3 and 5 and Ta-
ble 1, the arrival time of the proposed controller is the
shortest of the three controllers. The reason is that the pro-
GSMC
SMC posed controller utilizes the control torques as fully as pos-
0.00000 sible within the torque limits, unlike the other controllers,
0.0 0.5 1.0 as shown in Figs. 5 and 7. The maximum of the applied
Time [s] torque inputs with the proposed controller rarely goes
above the input torque limit, as shown in Fig. 4 and Table 1.
Figure 5. Trajectory tracking with GSMC and SMC.
The SMC has a faster rise to the reference input, but, near
the reference input, convergence time is much longer than
other controllers despite the application of over-limit con-
trol torques, as shown in Figs. 5 and 7 and Table 1. Although
the GSMC does not exceed the control torque limits, it does
Experiment not fully exploit the control torques and the arrival time in-
5
creases. By comparing the simulation and experimental re-
sults of the controllers, we can tell that the performance of
the proposed controller is superior to other SMCs.
The estimated minimum arrival time t f and the maximum
Input [v]

input torque using the closed-form equation (33) are similar


to the simulation results, as shown in Table 1. Also, in the ex-
perimental results for the proposed controller, the arrival
time at the reference input is quite similar to those of the sim-
0
ulation results, as shown in Fig. 3 and Table 1. The accor-
dance of the estimated minimum arrival time using the
closed-form equation to the simulation and experimental re-
0.0 0.5 1.0 sults implies that we can easily design motor-based actuators
Time [s] for mechanical systems without simulation and experiments.

Figure 6. Control inputs with GSMC.


Conclusion
An improved GSMC was proposed for controlling second-or-
der, time-varying systems with bounded uncertain parame-
ters and disturbances. The proposed controller drives the
system states along the minimum time trajectory within the
Experiment input torque limit. If the reference input and the bounds of
5
the uncertain parameters and disturbances are specified,
the minimum arrival time and the acceleration are ex-
pressed in closed-form equations. The proposed controller
was applied to the BLDC motor with uncertain loads. Experi-
Input [v]

mental results of the proposed controller are quite similar


to simulation and closed-form equation results and showed
the best performance compared with other SMCs. The
0 closed-form equation can be used for designing mo-
tor-based actuators for mechanical systems without simula-
tion and experiments.

0.0 0.5 1.0


Time [s] References
[1] S. V. Emelyanov, Variable Structure Control Systems. Moscow, USSR: Nauka,
Figure 7. Control inputs with SMC. 1967.

34 IEEE Control Systems Magazine June 2001


[2] U. Ikis, Control Systems of Variable Structure. New York: Wiley, 1976. North Carolina State University in 1993. Since 1993, he has
[3] V. I. Utkin, Sliding Mode and Their Applications. Moscow, USSR: Mir, 1978. been with the Division of Mechanical and Information Engi-
neering at Korea Maritime University, where he is currently
[4] J.Y. Hung, W. Gao, and J.C. Hung, “Variable structure control: A survey,”
IEEE Trans. Ind. Electron., vol. 40, no. 1, pp. 2-225, 1993. an Associate Professor. His main interests are in dynamics
and control of multiple cooperating robots, underwater ro-
[5] L.T. Ashchepkov, “Optimization of sliding motion in a discontinuous con-
trol,” Autom. Remote Cont., vol. 44, no. 11, pp. 30-37, 1983. botics, dynamics and control of the biped walking robot,
and robust control.
[6] T.L. Chern and Y.C. Wu, “An optimal variable structure control with inte-
grated compensation for electrohydraulic position servo control system,”
IEEE Trans. Ind. Electron., vol. 39, no. 5, pp. 460-464, 1992. Yong-heon Park received the B.S. degree from Korea Maritime
University in 1988. He is currently working toward an M.S. de-
[7] S.B. Choi and D.W. Park, “Moving sliding surfaces for fast tracking control
of second-order dynamical systems,” J. Dyn. Syst., Measure., Contr., vol. 116,
gree at the Division of Mechanical and Information Engi-
pp. 154-158, 1994. neering at Korea Maritime University. His main interests are in
control of motor systems, factory automation, dynamics and
[8] C.K. Lee and N.M. Kwok, “A variable structure controller with adaptive
switching surfaces,” in Proc. Amer. Control Conf., Seattle, WA, 1995, pp. control of the biped walking robot, and robust control.
1033-1034.
Yonsung Cho received the B.S. degree from the Pookyung
[9] J.J.E. Slotine and M.W. Spong, “Robust robot control with bounded input
torques,” J. Robot. Syst., vol. 2, no. 4, pp. 329-351, 1985. University in 1988 and the M.S. degree in electrical engineer-
ing from Dong-A University in 1990. He is currently working
[10] S.M. Madani-Esfahani, M. Hached, and S.H. Zak, “Estimation of slid-
ing-mode domains of uncertain variable structure systems with bounded toward the Ph.D. degree in electrical engineering at Dong-A
controller,” IEEE Trans. Automat. Contr., vol. 35, pp. 446-449, 1990. University. He worked as a Senior Engineer at Dongjin
[11] Y.S. Lu and J.S. Chen, “Design of a global sliding-mode controller for a mo-
Precision Inc. from 1990 to 1996. His research interests are in
tor drive with bounded control,” Int. J. Cont., vol. 62, no. 5, pp. 1001-1019, 1995. the areas of power electronics and control, which include
the drive system, and robust control.
[12] H. Hashimoto, H. Yamamoto, S. Yanagisawa, and F. Harashima,
“Brushless servo motor control using variable structure approach,” IEEE
Trans. Ind. Applicat., vol. 24, no. 1, pp. 160-170, 1988. Minho Lee received the B.S. degree from Kyungpook Na-
tional University in 1988 and the M.S. and Ph.D. degrees in
[13] S. Park, M. Kang, J. Lee, U. Huh, and J. Rhee, “A speed control of AC servo
motor using sliding mode controller,” in Proc. Korean Automatic Control Conf., electrical engineering from KAIST, Korea, in 1992 and 1995,
Taejon, Korea, 1994, pp. 1406-1410. respectively. From 1997 to 1999, he was with the Depart-
ment of Electrical Engineering at Korea Maritime University.
Hyeung-sik Choi received the B.S. degree from Korea Uni- He is currently with the Department of Electric and Electri-
versity in 1983, the M.S. degree in mechanical engineering cal Engineering at Kyungpook National University. His main
from the University of South Carolina in 1989, and the Ph.D. interests are in sensor fusion, intelligent control based on
degree in mechanical and aeronautics engineering from neural networks, and robotics.

June 2001 IEEE Control Systems Magazine 35

You might also like