ISSN 0005-1179, Automation and Remote Control, 2017, Vol. 78, No. 7, pp. 1218–1233. c Pleiades Publishing, Ltd.

, 2017.
c S. Keshtkar, A.S. Poznyak, E. Hernandez, A. Oropeza, 2017, published in Avtomatika i Telemekhanika, 2017, No. 7,
Original Russian Text 
pp. 57–75.

NONLINEAR SYSTEMS

Adaptive Sliding-Mode Controller

Based on the “Super-Twist” State Observer
for Control of the Stewart Platform
S. Keshtkar∗,a , A. S. Poznyak∗,b , E. Hernandez∗∗,c , and A. Oropeza∗∗,d
∗
Center for Research and Advanced Studies of the National Polytechnic Institute (CINVESTAV),
Mexico City, Mexico
∗∗
Higher School of Mechanics and Electrical Engineering (ESIME TICOMAN),
Mexico City, Mexico
e-mail: a [email protected], b [email protected],
c
[email protected], d [email protected]
Received November 18, 2015

Abstract—To provide a highly efficient control of nonlinear systems in the presence of non-
modeled dynamics and external perturbations, a new control law with feedback based on the
sliding modes with an observer of the “Super-Twist” kind was proposed. For acceptable use of
the continuous observer signal in the controller, presented were adaptive laws for adjustment
of the control system parameters. Using the methods of Lyapunov function, system stability
(convergence to a zone) was proved. This technique was proposed as an example of control and
stabilization of the position of a parallel manipulator (Gough–Stewart platform). The presented
mechanism with six degrees of freedom is used to control the secondary mirror of the “Large
Millimeter Telescope Alfonso Serrano” situated in the state of Puebla, Mexico.

Keywords: adaptive controller, sliding modes, state observer, Stewart platform, stability in
zone.
DOI: 10.1134/S0005117917070049

1. INTRODUCTION
The present paper is devoted to the development of control laws providing an acceptable sys-
tem operability under perturbations/uncertainty. In the formulation of any control problem there
always exists a discrepancy between the actual dynamics of the control plant and its mathemat-
ical model. These discrepancies (or inadequacy) mostly are due to the external perturbations,
unknown plant parameters, and unknown dynamics. Additionally, the positioning systems of the
radiotelescope considered in what follows as an example of application operate in a severe moun-
tainous environment such as wind force or radiotelescope structural deformations, and require
use of the so-called robust controllers capable of offsetting these undesirable factors. The robust
PID–controller [1], control of inverse dynamics [2], neuro-fuzzy control [3], and others exemplify
application of the robust algorithms to control the Gough–Stewart platform (GSP). W. Dongsu and
G. Hongbin [4] suggested an adaptive controller based on the method of sliding modes. P.R. Kumar
and B. Bandyopadhyay [5] proposed a technique of GSP control based on the high-order sliding
mode with estimation of the platform kinematics.
The present paper proposes a new controller designed on the basis of sliding mode with a high-
order observer for parallel estimation of the system states. The classical second-order algorithm of
the Super-Twist kind is used to estimate the system speeds [6]. This observer has high precision

1218
 ADAPTIVE SLIDING-MODE CONTROLLER 1219

of system speed estimate and ensures estimation error convergence to a small zone comprising the
origin.
This approach is illustrated by application to the two-mirror antennas which are most popular
is the astronomic radio communication. In the two-mirror radiotelescope the tracking subsystem is
responsible for keeping the secondary mirror at the desired position, which requires six independent
degrees of freedom. This task is executed in the great millimeter telescope by the high-precision
GSP [7].
GSP is a parallel manipulator with octahedral arrangement of the poles with six independent
“legs” (supports) connecting the fixed base with moving platform. The platform can assume
the desired position by varying the pole lengths. These manipulators attract attention of many
researchers owing to their precision, rigidity, and high force to load ratio. They were used in
numerous applications in the machine building, underwater prospecting, aviation rescue operations,
aviation simulators, telescopes, orthopedic surgery, and so on [8–13].
Section 2 presents the mathematical model with regard for GSP kinematics and dynamics and de-
scribes application of these platforms in the radiotelescope tracking mechanism. Section 3 presents
the design of the observer based on the Super-Twist structure and proposes a design of the control
device. The rule for adjustment of the control and observation algorithm is discussed in Sec-
tion 4. The last Section 5 considers with the use of numerical modeling efficiency of the proposed
approach.

2. SYSTEM OUTLINE
Figure 1 depicts a mechanism for orientation of the radiotelescope secondary mirror installed
on the “Large Millimeter Telescope Alfonso Serrano” [7]. The system consists of a stationary base,
mirror attached to the moving platform, and six prismatic drives with a screw-nut transmission.
The mirror installed on the telescope can do three linear motions along the axes x, y, z and three

Fig. 1. Secondary mirror of the radiotelescope antenna.

AUTOMATION AND REMOTE CONTROL Vol. 78 No. 7 2017

1220 KESHTKAR et al.

rotations such as pitch, roll and yaw. The drives are connected to the fixed base by the ball-and-
socket hinges and attached to the moving platform by the Hooke joint.
Represent the mathematical model of the system in the vector form (see the Appendix for
description).

ẋ1 (t) = x2 (t) ,

ẋ2 (t) = f + g (x1 , t) u + ξ (x, t) , (1)
y (t) = x1 (t) ,

where
• x1 (t), x2 (t) ∈ IR6 system state at the instant t ∈ IR+ ;
 
• x (t) := x 
1 (t) , x2 (t) ∈ IR12 ;
• y (t) ∈ IR6 is the measurable output at the instant t  0;
• u ∈ IR6 is the control signal;
• ξ (x, t) ∈ IR6 is the immeasurable term including the external and internal perturbations and
uncertainties,
• f = (0, −M g, 0, 0, 0, 0)  ;
• g ∈ IR6×6 = [gk,i ](k, i = 6) is known completely and invertible.
We note that in model (1) only half coordinates (x1 (t)) are measurable in time.
The kinematics and GSP motion equations are studied in depth in [8, 14–16]. The present paper
follows the approach of [17] enabling one to give a more detailed mathematical description of the
considered system in form of (1).

3. FEEDBACK CONTROL
Consider a sliding mode controller with a high-order observer for parallel estimation of the
system states. The system speeds are estimated using the classical second-order algorithm of the
Super-Twist kind [6]. A simplified variant of the adaptive sliding mode for the problem with all

Fig. 2. Block diagram of a controlled plant with nonmodeled dynamics.

AUTOMATION AND REMOTE CONTROL Vol. 78 No. 7 2017

 ADAPTIVE SLIDING-MODE CONTROLLER 1221

model states observable was considered in [18]. Here we present a new adaptive control using the
estimate of the current state. The effect of nonmodeled dynamics and uncertain terms ξ (x, t) is
considered below (see Fig. 2).

3.1. Speed Estimation

To estimate the immeasurable coordinate x2 (t) in (1), we make use of the Super-Twist observer
based on the second-order sliding mode [6]:

x̂˙ 1 = x̂2 + λ y − x̂1 1/2 sgn (y − x̂1 ) ,

(2)
x̂˙ 2 = f + g (y, t) u + α sgn (y − x̂1 ) ,

where in the original publications [6] the positive parameters α and λ are constants. The vector
sgn (z) is given by

sgn (z) := (sgn (z1 ) , . . . , sgn (zn )) ,

⎧
⎪
⎨ 1 if zi > 0 (3)
sgn (zi ) := −1 if zi < 0
⎪
⎩ ∈ [−1, 1] , if z = 0.
i

Define a new variable e := (x − x̂) characterizing the error in estimation of the current state and
obeying the following dependence:

ė1 = e2 − λ e1 1/2 sgn (e1 ) ,

(4)
ė2 = Fe (x1 , x2 , x̂2 , t) − αsgn (e1 ) ,

where

Fe (t, x1 , x2 , x̂2 ) = f + ξ (x, t) (5)

describes the system uncertainty and external perturbations. We notice that the vector

e1 = x1 − x̂1 = y − x̂1 (6)

is measurable.

3.2. Controller Design

Sliding surface. The problem at hand lies in generating a feedback control law u (x̂, t) driving
the system states x (t) to the desired values, that is,

lim x1 = x∗1 , lim x2 = 0.

t→∞ t→∞

This requirement can be expressed in another way using the sliding mode approach. To realize it,
introduce in the system state space a new variable σ (x) defining the desired sliding surface [19]:

σ (x) := ẋ1 + C (x1 − x∗1 ) = x2 + C (x1 − x∗1 ) = 0, (7)

where x∗1 is the desired value of the state and C = diag(c1, c2 , . . . , cn ) is the diagonal matrix with
positive elements. On this surface σ (x (t)) → 0 if t → ∞, which corresponds to the system sliding
to the origin at an exponential speed.

AUTOMATION AND REMOTE CONTROL Vol. 78 No. 7 2017

1222 KESHTKAR et al.

4. ADJUSTMENT OF CONTROLLER AND OBSERVER

Problem. To drive the system to the desired sliding surface

σ (x) = [σ1 (x̂) , . . . , σn (x̂)] ,

sgn (σ (x̂)) = [sgn (σ1 (x̂)) , . . . , sgn (σn (x̂))] , (8)
 
σi (x̂) = x̂2,i + ci x̂1,i − x∗1,i , i = 1, . . . , n,

one has to adjust appropriately the parameters

k (x̂, t) , λ (x̂, t) , α (x̂, t) . (9)

u (x̂, t) is selected as follows:

u (x̂, t) = −k (x̂, t) [g (y, t)]−1 sgn (σ (x̂)) , (10)

where the state estimates x̂ are generated by observer (2) with the variable parameters:

x̂˙ 1 = x̂2 + λ (x̂, t) y − x̂1 1/2 sgn (y − x̂1 ) ,

(11)
x̂˙ 2 = f (x̂, t) + g (y, t) u + α (x̂, t) sgn (y − x̂1 ) .

4.1. Final Form of the Control Algorithm

Theorem. Under the assumptions that f (x, t) and ξ (x, t) satisfy the Lipschitz condition for the
first arguments and are measurable in t, namely, that for all t  0
       
Fe (x1 , x2 , x̂2 , t)− Fe x , x , x̂ , t   Lf x1 − x  + x2 − x  + x̂2 − x̂  ,
1 2 2 1 2 2

the dependent parameters k = k (x̂, e1 , t) (9) in the controller (10) and λ (x̂, t) and α (x̂, t) in the
observer (11) are designed as

p0 p0
k = k (x̂, e1 , t) := (p0 )−1 ρ + κ (x̂, e1 , t) + γ , ρ ,
2 2
(12)
  
 
κ (x̂, e1 , t) := f (x̂, t) + α sgn (e1 ) + C x̂2 + λ e1 1/2 sgn (e1 )  ,

and
⎧ 
p1
⎪
⎪ η 1 (γ) + 
⎪
⎪ 2
⎨ p e 
⎪ if e1   > 0
1 1
λ (x̂, t) := 
⎪
⎪ p1 (13)
⎪
⎪ η1 (γ) + 2 
⎪
⎩ , if e1  < ,
p1
α (x̂, t) := α > 0.

Under such conditions one can guarantee that the “energy function”
p0 p1 p2
V (t) := V (σ (x̂ (t)) , e1 (t) , e2 (t)) = σ (x̂ (t))2 + e1 (t)2 + e2 (t)2 ,
2 2 2 (14)
p0 , p1 , p2 > 0,

AUTOMATION AND REMOTE CONTROL Vol. 78 No. 7 2017

 ADAPTIVE SLIDING-MODE CONTROLLER 1223

converges to the μ-zone in a finite time tf , namely, for all


t  tf = γ −1 2W (0)
the relation
 2
W (t) := V (t) − μ =0 (15)
+
holds where
 
μ := μ0 α, γ, e+ +
2 + μ1 ( ) e2 ,
    10 p1
μ0 α, γ, e+
2 := η0 α, e+
2 + γ , (16)
27 2
  
p2 10  
μ1 ( ) := + p1 + 2Lf p2
2 27
and
 √ √ +   p2
η0 (α) := p2 e+
2 α n + 2ξ + 2Lf e+
2 +γ ,
2
p1   
η1 (γ) := γ + p1 + 2Lf p2 e+ 2, γ > 1,
2
2c
e+
2 = √ 2 , (17)
b + 4ac + b
 10 p1
a := 2Lf p2 , c := (γ − 1) , > 0,
27 2
10     √ √ 
b := p1 + 2Lf p2 + p2 α n + 2ξ + .
27
The initial values of the observer x̂2 (t0 ) need not be “too far” from the actual values x2 (t0 ):
x2 (t0 ) − x̂2 (t0 )  e+
2.

The theorem is proved in the Appendix.

5. NUMERICAL MODELING
Figures 3 and 4 depict behavior of the actual GSP states x and their speeds ẋ at using the
proposed controller (10). The controlled variables reach the desired values virtually in six seconds.
As can be seen from Fig. 5, the nonmodeled dynamics does not exert a substantial effect on the
components of the controlled process. As one would expect, the realized control law rather quickly
drives the system behavior to the desired mode even under a nonmodeled dynamics (ς = 0.05) (see
Fig. 6).

6. CONCLUSIONS
The present paper proposed a new approach to the problem of controlling the nonlinear systems
based on the sliding mode method. At that, used are the estimates of state speeds acquired in
the course of operation of the observer of a “Super-Twist” type. The controller and observer
parameters vary in time depending on the current measurements. The Gough-Stewart platform
exemplifies application of the proposed approach. Its design and basic characteristics are described.
Consideration was given to the characteristics of this platform with six degrees of freedom and its
behavior in the controlled mode with regard for the desired orientation of the secondary antenna
of the radiotelescope.

AUTOMATION AND REMOTE CONTROL Vol. 78 No. 7 2017

1224 KESHTKAR et al.

Fig. 3. Representation of the system states, (1), (2), (3) are the coordinates of the point B (XB , YB , ZB ) and
(4), (5), (6) are the angular coordinates of the platform (ϕ1 , ϕ2 , ϕ3 ).

Fig. 4. Angular speeds and their estimates. The digits denote the states (1)–(6) (dash-dot lines) and their
estimates (solid lines).

Fig. 5. Impact of the nonmodeled dynamics on the system state and the coordinates coordinate XB and ZB
(the dash-dot line represents the state under the action of the nonmodeled dynamics).

AUTOMATION AND REMOTE CONTROL Vol. 78 No. 7 2017

 ADAPTIVE SLIDING-MODE CONTROLLER 1225

Fig. 6. Control signals.

ACKNOWLEDGMENTS
This work was supported by the National Mexican Council for Science and Technology, project
no. 262887.

APPENDIX

A.1. PROOF OF THE THEOREM

1. By calculating the time derivative of the energy function (14) on the trajectories of the
considered dynamics, we establish that
d d
V̇ = p0 σ (x̂) σ̇ (x̂) + p1 e1 ė1 + p2 e2 ė2 = p0 σ (x̂) x̂2 + C x̂1
dt dt
 
+ p1 e1 e2 − λ e1 1/2 sgn (e1 ) + p2 e2 [Fe − αsgn (e1 )]
 −kp0 σ (x̂) Sgn (σ (x̂))
  
+ p0 σ (x̂) f (x̂, t) + αsgn (e1 ) + c1 x̂2 + λ e1 1/2 sgn (e1 )
√   √
+ p1 e1 e2 − λp1 e1 3/2 + 2p2 e2  Lf e + ξ + + αp2 n e2  ,

which leads to
V̇  σ (x̂) [−kp0 + κ (x̂, e1 , t)] − λp1 e1 3/2
 √   √ √ 
+ e2  p1 e1  + 2p2 Lf (e1  + e2 ) + p2 α n + 2ξ + ,

where
  
 
κ (x̂, e1 , t) := f (x̂, t) + αsgn (e1 ) + C x̂2 + λ e1 1/2 sgn (e1 )  .

After adding and subtracting the term

   
γ p0 /2 σ (x̂ (t)) + p1 /2 e1 (t) + p2 /2 e2 (t) ,

AUTOMATION AND REMOTE CONTROL Vol. 78 No. 7 2017

1226 KESHTKAR et al.
√
which is estimated as γ V (γ > 0), we get
√   
p0 p1
V̇  −γ V + σ (x̂) −kp0 + κ (x̂, e1 , t) + γ − e1  λp1 e1  − γ
2 2
   √ √ + 
p2
+ e2  p1 e1  + p2 2Lf (e1  + e2 ) + p2 α n + 2ξ + γ .
2

As the result, for all t  0

e2 (t)  e+
2 (A.1)

and, consequently,
√  
p0
V̇  −γ V − σ (x̂) −kp0 + κ (x̂, e1 , t) + γ
2
      (A.2)
− e1  λp1 e1  − η1 γ, e+
2 + η0 α, e+
2 ,

where
    √ √ + 
η0 α, e+
2 := e+
2 2Lf p2 e+
2 + p2 α n + 2ξ ,
     (A.3)
p1
η1 γ, e+
2 := p1 + 2Lf p2 e+
2 −γ .
2
For k = k (x̂, e1 , t) (see (12)), we establish (A.2) that
√      
+
V̇  −γ V − ρ σ (x̂) − e1  λp1 e1  − η1 γ, e2 + η0 α, e+
2 , (A.4)

where η0 (α) and η1 (γ) are defined in (A.3). Introduce the Lyapunov function
√ 2
W := V −μ , μ > 0, (A.5)
+

for which obtain with regard for (A.4)) that

dW √  1 √ 
Ẇ = V̇ = −γ V − μ + √ V −μ
dV + V +
      (A.6)
× −ρ σ (x̂) − e1  λp1 e1  + e1  η1 γ, e+
2 + η0 α, e+
2 .

By selecting the parameters λ, α, ρ, and μ so that

√       
Φ := V −μ −ρ σ (x̂) − e1  λp1 e1  + e1  η1 γ, e+ +
2 + η0 α, e2  0, (A.7)
+

provided that ρ  p0 /2, we obtain
  √
ρ σ (x̂) + p1 /2 e1 (t) + p2 /2 e2 (t)  V

and
√     
Φ  − V − [μ]+ − η0 + p2 /2e+
2 + e1  η1 + p 2 /2 − λp e1  . (A.8)
1

AUTOMATION AND REMOTE CONTROL Vol. 78 No. 7 2017

 ADAPTIVE SLIDING-MODE CONTROLLER 1227

By taking λ as
⎧   
⎪
⎪ η1 γ, e+ + p1 /2 
⎪
⎪ 2
⎪
⎪  , if e1   > 0
⎨ p1 e1 
λ=
⎪    (A.9)
⎪
⎪ +
⎪
⎪
η1 γ, e2 + p1 /2 
⎪
⎩ , if e1  < ,
p1
one can readily demonstrate that
   
e1  η1 γ, e+
2 + p1 /2 − λp1 e1 
⎧ 
⎨ 0, if e1  
=        
⎩ e1  η1 γ, e+
2 + p1 /2 1− e1 / , if e1  < .

Taking into consideration that x − x3/2  10/27 ∀x  0, we get for x = e1  that
      
e1  η1 γ, e+
2 + p1 /2 1− e1 /  10/27 η1 γ, e+
2 + p1 /2 .

With regard for (5), the latter relation is representable as

√  √       
Φ− V −μ V − η0 α, e+
2 + p1 /2e+ +
2 + 10/27 η1 γ, e2 + p1 /2 .
+

By defining μ as in (A.5), the latter inequality is rearranged in

√  √  √ 2
Φ− V −μ V −μ = − V −μ 0
+ +

and satisfies (A.7). With regard for (A.6), we get that

√  √
Ẇ  −γ V − μ = −γ W (A.10)
+

under the assumption that Wt = 0 ∀ t  tf = 2γ −1 W (0).
2. To complete the proof, it remains to show that if (A.1) is satisfied for the instant t0 , then
the theorem is valid for any other instant greater than t0 . With regard for (A.10), we get
 
W (t0 + Δt)  W (t0 ) − γΔt/2

which is equivalent to
     
V (t0 + Δt) − μ  V (t0 )−μ −γ/2Δt < V (t0 ) − μ :
+ + +
 
(a) if V (t0 )  μ, then V (t0 + Δt)  μ for any small Δt > 0 and
      
e2 (t0 + Δt)  2/p2 V (t0 + Δt)  2/p2 μ0 α, γ, e+ +
2 + μ1 ( ) e2 .

To satisfy the relation

    
2/p2 μ0 α, γ, e+ + +
2 + μ1 ( ) e2 := e2 ,

AUTOMATION AND REMOTE CONTROL Vol. 78 No. 7 2017

1228 KESHTKAR et al.

it suffices to realize the equality

  2   
10    √ √  10 p1
2Lf p2 e+
2 + p1 + 2Lf p2 + p2 α n + 2ξ + e+
2 = (γ − 1) ,
27 27 2

by selecting e+
2 as in (17), which means that for t0 = 0 the initial values of the observer must satisfy
the relation x2 (t0 ) − x̂2 (t0 )  e+
2;

(b) if V (t) > μ, then

p2   γ 
e2 (t + Δt)  V (t + Δt)  V (t)− Δt < V (t)
2 2
and, consequently, one can take
 
2 p0 p1  2
e+
2 := V (0)  σ (x̂ (0))2 + e1 (0)2 + e+
2 ,
p2 p2 p2

which means that the choice of e+

2 as (17) encompasses both cases (a) and (b) and proves the
theorem.

A.2. MATHEMATICAL MODELING OF THE SYSTEM

A simplified kinematic model of the proposed GSP is depicted in Fig. 7. The skid feet (drives)
Ai Bi , i = 1, 6, consist of the upper and lower parts connected with the progressive kinematic pair
and attached to the moving platform at the points Bi through the universal Hooke joints. They are
also connected to the base at the points Ai by spherical joints. Each skid feet Ai Bi has length li
and inclined at an angle of γi .
Associate with the fixed base of the platform the coordinate system AXY Z so that the origin A
coincides with the center of symmetry of the base, axis AX passes through the joint A3 , axis AY
is directed along the normal to the base, and the axis AZ makes up the right triple with the two
above axes. Associate in a similar manner the coordinate system Bxyz with the moving platform.

Fig. 7. Geometry of the platform with degrees of freedom.

AUTOMATION AND REMOTE CONTROL Vol. 78 No. 7 2017

 ADAPTIVE SLIDING-MODE CONTROLLER 1229

Fig. 8. Representation of the Eulerian angles

Positions of the joints A1 , A6 and B1 , B6 in the coordinate system AXY Z are defined by the
vectors

Āi = (Ai,1 , Ai,2 , Ai3 ) , B̄i = (Bi,1 , Bi,2 , Bi3 ) , i = 1, 6. (A.11)

Position of the moving platform with respect to the base is defined by the Eulerian angles ϕ1 ,
ϕ2 and ϕ3 (Fig. 8) and vector B̄ = (XB , YB , ZB ).
The geometrical relations between the coordinate systems AXY Z and Bxyz are defined as the
matrix of homogeneous transformations TH :1
⎡ ⎤
a1,1 a1,2 a1,3 XB
⎢ ⎥
⎢ a2,1 a2,2 a2,3 YB ⎥
TH = ⎢ ⎥.
⎣ a3,1 a3,2 a3,3 ZB ⎦
0 0 0 1

Correspondingly, the position of the combined Bi in the coordinate system XY Z is defined as

follows:

B̄i = TH [Bi,1 , Bi,2 , Bi,3 , 1] . (A.12)

It follows from (A.11) and (A.12) that the generalized coordinate li as a function of XB , YB ,
ZB , ϕ1 , ϕ2 , and ϕ3 obeys the expression

2
li = Ai,j − B̄i , j = 1, 2, 3.
j

A.3. DYNAMICS OF THE MECHANISM

The second-order Lagrange equation
d ∂T ∂T
− = Qi , i = 1, . . . , 6,
dt ∂ q̇i ∂qi
1
Values of ai,i are given in the Appendix (Section A.3).

AUTOMATION AND REMOTE CONTROL Vol. 78 No. 7 2017

1230 KESHTKAR et al.

Fig. 9. External forces acting on the platform.

is used to formulate the platform motion equation, where T is the kinetic energy of the system, Qi
is the generalized force corresponding to the generalized coordinate qi :

q1 = XB , q2 = YB , q3 = ZB , q4 = ϕ1 , q5 = ϕ2 , q6 = ϕ3 .

With neglect of the mass of skid feet, the system kinetic energy is representable as
1  2  
T = M ẊB + ẎB2 + ŻB
2
+ Jx ϕ̇21 + Jy ϕ̇22 + Jz ϕ̇23 , (A.13)
2
where M is the platform mass and Jx,y,z is the inertia moment relative to the corresponding axis.
The generalized forces acting on the system are shown in Fig. 9 and given by
6
 6

Q1 = Fi cos γi,1 , Q2 = −P + Fi cos γi,2 ,
i=1 i=1
6
 6
 3

Q3 = Fi cos γi,3 , Q4 = Fi (ai,j,1 cos γi,j ) , (A.14)
i=1 i=1 j=1
6
 3
 6
 3

Q5 = Fi (ai,j,2 cos γi,j ) , Q6 = Fi (ai,j,3 cos γi,j ) ,
i=1 j=1 i=1 j=1

where Fi represents the forces acting on the linear drives, and P = M g. Rearrange the system in
terms of the new variables. The coordinates and speeds of the point B are given by

XB = x1 , Y B = x2 , ZB = x3 ,
ϕ1 = x4 , ϕ2 = x5 , ϕ3 = x6

and the angles and angular speeds of the point B by

ẊB = x7 , ẎB = x8 , ŻB = x9 ,

ϕ̇1 = x10 , ϕ̇2 = x11 , ϕ̇3 = x12

and the control signals of force drive by

ui = Fi , i = 1, 6.

AUTOMATION AND REMOTE CONTROL Vol. 78 No. 7 2017

 ADAPTIVE SLIDING-MODE CONTROLLER 1231

Remark. The drive system is described in detail in [20]. The drives for the platforms under
consideration consist of six engine-ball drives of the screw-nut type. The electrical power controller
is a d.c. engine whose dynamics obeys the differential equations

F = l0−1 τ, τ = Ki (l0 > 0) ,

di 2πn
L + R + Kb l̇ = v,
dt p

where τ , K, L, R, Kb , p, and v are the parameters of electrical engines and u is a six-dimensional

current vector in six engines. The six-dimensional vector of the rates of variation of rod lengths is
denoted by l̇. If by design the values of L and the terms Kb 2πn p l̇ are small, then we deal with a
low-frequency filter providing the property

i  R−1 v,

assuming that

K
F  G0 v, G0 := ,
R
which means that the drives can be regarded as proportional-control devices.
Values of ai,k .

a1,1 = cos ϕ1 cos ϕ3 − sin ϕ1 cos ϕ2 cos ϕ3 ,

a1,2 = sin ϕ1 cos ϕ3 + cos ϕ1 cos ϕ2 sin ϕ3 ,
a1,3 = sin ϕ2 sin ϕ3 ,
a2,1 = − cos ϕ1 sin ϕ3 − sin ϕ1 cos ϕ2 cos ϕ3 ,
a2,2 = − sin ϕ1 sin ϕ3 + cos ϕ1 cos ϕ2 cos ϕ3 ,
a2,3 = sin ϕ2 cos ϕ3 ,
a3,1 = sin ϕ1 sin ϕ2 ,
a3,2 = − cos ϕ1 sin ϕ2 ,
a3,3 = cos ϕ2 .

Values of gk,i and ak,i,j .

g1i = M −1 cos γi,1 ;

g2i = M −1 cos γi,2 ;
g3i = M −1 cos γi,3 ,
g4i = Jx−1 (ai,1,1 cos γi,1 + ai,1,2 cos γi,2 + ai,1,3 cos γi,3 ) ,
g5i = Jy−1 (ai,2,1 cos γi,1 + ai,2,2 cos γi,2 + ai,2,3 cos γi,3 ) ,
g6i = Jz−1 (ai,2,1 cos γi,1 + ai,3,2 cos γi,2 + ai,3,3 cos γi,3 ) ,

where γi,k is the angle between Ai Bi and the axes AX, AY , and AZ:

B̄i − Ai,j
cos γi,j = cos γi,j (XB , YB , ZB , ϕ1,2,3 ) =  2
Ai,j − B̄i
j

AUTOMATION AND REMOTE CONTROL Vol. 78 No. 7 2017

1232 KESHTKAR et al.

and

ai,1,1 = Bi,1 (− sin ϕ1 cos ϕ3 − cos ϕ1 cos ϕ2 cos ϕ3 ) + Bi,2 (cos ϕ1 cos ϕ3 − sin ϕ1 cos ϕ2 sin ϕ3 ) ,
ai,1,2 = Bi,1 sin ϕ1 cos ϕ2 cos ϕ3 − Bi,2 cos ϕ1 sin ϕ2 sin ϕ3 + Bi,3 cos ϕ2 sin ϕ3 ,
ai,1,3 = Bi,1 (sin ϕ1 cos ϕ2 sin ϕ3 − cos ϕ1 sin ϕ3 )
+ Bi,2 (cos ϕ1 cos ϕ2 cos ϕ3 − sin ϕ1 sin ϕ3 ) + Bi,3 sin ϕ2 cos ϕ3 ,
ai,2,1 = Bi,1 (sin ϕ1 sin ϕ3 − cos ϕ1 cos ϕ2 cos ϕ3 ) − Bi,2 (cos ϕ1 sin ϕ3 + sin ϕ1 cos ϕ2 cos ϕ3 ) ,
ai,2,2 = Bi,1 sin ϕ1 sin ϕ2 cos ϕ3 − Bi,2 cos ϕ1 sin ϕ2 cos ϕ3 + Bi,3 cos ϕ2 cos ϕ3 ,
ai,2,3 = Bi,1 (sin ϕ1 cos ϕ2 sin ϕ3 − cos ϕ1 cos ϕ3 )
− Bi,2 (sin ϕ1 cos ϕ3 + cos ϕ1 cos ϕ2 sin ϕ3 ) − Bi,3 sin ϕ2 sin ϕ3 ,
ai,3,1 = − Bi,2 sin ϕ1 sin ϕ2 ,
ai,3,2 = cos ϕ2 (Bi,1 cos ϕ3 − Bi,2 cos ϕ1 ) − Bi,3 sin ϕ2 ,
ai,3,2 = − Bi,1 sin ϕ2 sin ϕ3 .

REFERENCES
1. Jan, R.M., Tseng, C.S., and Liu, R.J., Robust PID Control Design for Permanent Magnet Synchronous
Motor: A Genetic Approach, Electr. Power Syst. Res., 2007, no. 7, pp. 1161–1168.
2. Lee, S.H., Song, J.B., Choi, W.Ch., et al., Position Control of a Stewart Platform Using Inverse Dynamics
Control with Approximate Dynamic, Mechatronics, 2003, no. 13, pp. 605–619.
3. Elmas, C., Ustun, O., and Sayan, H.H., A Neuro-Fuzzy Controller for Speed Control of a Permanent
Magnet Synchronous Motor Drive, Exp. Syst. Appl., 2008, no. 1, pp. 657–664.
4. Dongsu, W. and Hongbin, G., Adaptive Sliding Control of Six-dof Flight Simulator Motion Platform,
Chinese J. Aeronaut., 2007, pp. 425–433.
5. Kumar, P.R. and Bandyopadhyay, B., Stabilization of Stewart Platform Using Higher Order Sliding
Mode Control, in 7th Int. Conf. Electrical Computer Engineering (ICECE), 2012, no. 13, pp. 945–948.
6. Shtessel, Y., Edwards, Ch., Fridman, L., et al., Sliding Mode Control and Observation, Exeter:
Birkhauser, 2014.
7. Schloerb, F.P., Stepp, L.M., and Gilmozzi R., The Large Millimeter Telescope, in Ground-Based and
Airborne Telescopes, in Proc. SPIE, 2008, vol. 7012, pp. 1117–1129.
8. Merlet, J.P., Parallel Robots, France: Springer, 2006, 2nd ed.
9. Su, X.S., et al., Singularity Analysis of Fine-Tuning Stewart Platform for Large Radio Telescope Using
Genetic Algorithm, Mechatronics, 2003, vol. 5, pp. 413–425.
10. Leroy, N., et al., Dynamic Modeling of a Parallel Robot. Application to a Surgical Simulator, in IEEE
Int. Conf. on Robotics and Automation, 198, Flight Simulation, 1975, pp. 4330–4335.
11. Koevermans, W.P., et al., Design and Performance of the Four D.O.F. Motion System of the NLR
Research Flight Simulator, in AGARD Conf. Proc., 1975, no. 198, pp. 17(1)–17(11).
12. Dubowsky, A., et al., The Design and Implementation of a Laboratory Test Bed for Space Robotics, in
ASME Design Automat. Conf., 1994, pp. 99–108.
13. Shang, C., Guoliang, T., and Meng, D., Adaptive Robust Trajectory Tracking Control of a Parallel
Manipulator Driven by Pneumatic Cylinders, Adv. Mechan. Eng., 2016, vol. 8, no. 4, pp. 1–15.
14. Leonov, G.A., Zegzhda, S.A., Kuznetsov, N.V., et al., Motion of Solid Body Controlled by Six Variable-
length rods, Dokl. Ross. Akad. Nauk, 2014, vol. 455, pp. 153–157.

AUTOMATION AND REMOTE CONTROL Vol. 78 No. 7 2017

 ADAPTIVE SLIDING-MODE CONTROLLER 1233

15. Leonov, G.A., Zegzhda, S.A., Zuev, S.M., et al., Dynamics of the Stewart platform and Control of Its
Motion, Dokl. Ross. Akad. Nauk, 2014, vol. 458, no. 1, pp. 36–41.
16. Alyushin, Y.A. and Elenev, S.A., Mathematical Model of Stewart Platform Motion, J. Mach. Manuf.
Reliab., 2010, vol. 39, no. 4, pp. 305–312.
17. Kaganov, Yu.T. and Karpenko, A.P., Mathematical Modeling of Kinematics and Dynamics of the Ma-
nipulating Robot of the “Trunk” Type, Nauka Obrazovan., 2009, no. 77, pp. 1–14.
18. Keshtkar, S. and Poznyak, A., Tethered Space Orientation via Adaptive Sliding Mode, Int. J. Robust.
Nonlin. Control, 2016, vol. 26, no. 8, pp. 1632–1646.
19. Utkin, V.I., Skol’zyashchie rezhimy v zadachakh optimizatsii i upravleniya, Moscow: Nauka, 1981. Trans-
lated into English under the title Sliding Modes in Control Optimization, Heidelberg: Springer, 1992.
20. Harib, K. and Srinivasan, K., Kinematic and Dynamic Analysis of Stewart Platform-Based Machine
Tool Structures, Robotica, 2003, no. 21, pp. 541–554.

This paper was recommended for publication by A.P. Kurdyukov, a member of the Editorial
Board

AUTOMATION AND REMOTE CONTROL Vol. 78 No. 7 2017