You are on page 1of 5

2019 Sixth Indian Control Conference (ICC)

December 18-20, 2019. IIT Hyderabad, India

Angular rate stabilization using fixed-time


continuous sliding mode control
Salahudden Dipak Kumar Giri Subham Dey Bijoy Krishna Mukherjee
Aerospace Engineering Aerospace Engineering Aerospace Engineering Electrical and Electronics Engineering
IIT-Kanpur IIT-Kanpur IIT-Kanpur BITS Pilani
Kanpur, India Kanpur, India Kanpur, India Pilani, India

Abstract—This paper proposes a novel fault-tolerant con- found to be independent of the initial values of state-space
trol algorithm for angular rate stabilization of satellite variables thereby overcoming the limitations of finite-
system. The control algorithm belongs to the class of time stable algorithms and is therefore termed as fixed-
super-twisting sliding mode control theorem applicable to
relative degree one systems which ensures chattering free time stable control algorithm. Furthermore, the control
control input and also finite-time convergence. Lyapunov algorithm by virtue of being continuous in nature is found
stability theorem is used to prove finite-time stability of to be free from chattering which is one of the significant
the closed loop system under designed control input. The disadvantages of algorithms belonging to the class of
expression of convergence time is calculated and is found to classical sliding mode control algorithms and is also
be independent of initial conditions. Numerical simulations
verify the effectiveness of the proposed control algorithm. termed as super-twisting sliding mode control algorithm
Index Terms—Fixed-time control algorithm, Super- (Nagesh and Edwards (2014)). In realistic scenarios fault
twisting sliding mode control, Fault-tolerant control algo- does occur in the system and therefore it is desirable that
rithm. the designed control algorithms are fault-tolerant in nature
(Blanke et al. (1997)). The fault-tolerant control law is
I. I NTRODUCTION designed along the same lines as given in Ref. (Han et al.
The control algorithm proposed in this paper belongs (2016)). Previously, a fault-tolerant control algorithm was
to the broad class of Sliding Mode Control (SMC) implemented using observers which are designed based
algorithms which is a popular robust control algorithm on principles of fixed-time super-twisting sliding mode
known for its robustness to uncertainties (matched and control for hypersonic vehicles (Yu et al. (2017)).
unmatched) in system dynamics and external disturbances The main contributions of this work are:
(Perruquetti and Barbot (2002), Utkin (1977)). The fun- 1) A fixed-time super-twisting sliding mode control
damental limitations of classical sliding mode control algorithm is implemented for angular rate stabiliza-
theory is the problem of chattering (Utkin (2013b)) tion of satellite system.
which can be overcome by using quasi-sliding mode 2) The control algorithm implemented is fault-tolerant
control techniques. Furthermore, classical sliding mode in nature.
control also suffers from the limitations of finite-time
The paper is organized as follows: In Section II the
convergence which can be overcome by using control
preliminaries related to satellite attitude dynamics is
algorithms belonging to terminal sliding mode category
presented. Section III represents the designed control
(Venkataraman and Gulati (1991)). However, chattering
algorithm as well as the expression of convergence time is
remains a fundamental problem for terminal sliding mode
also calculated in this section. Section IV represents the
control algorithms. Higher-order sliding mode control
numerical simulation results. The paper concludes with
have been proposed in the literature to overcome the
Section V.
limitation of chattering but they require derivatives of
the sliding variables (Shtessel et al. (2014),Fridman and II. P RELIMINARIES
Levant (2002)). The control algorithm implemented in The satellite dynamics considered for simulation pur-
this paper also belongs to the class of finite-time stable poses in this paper is the one given in (Lo and Chen
algorithms for which the expression of upper bound of (1995)).
convergence time is found to be dependent on initial
conditions (Polyakov and Poznyak (2009a), Polyakov and ˙
J ω̃(t) ˜
= −S(ω̃(t))J ω̃(t) + ũ(t) + d(t) (1)
Poznyak (2009b)). The control algorithm further belongs
to the class of second-order sliding mode control for where J ∈ R3×3 denotes the inertia matrix of the satellite
which the expression of upper bound of convergence time system, ω̃(t) ∈ R3×1 is the angular rate of the satellite
is also calculated in Ref. (Basin et al. (2016)) and is system, ũ(t) ∈ R3×1 is the designed control input of the

978-1-7281-3860-2/19/$31.00 ©2019 IEEE 485

Authorized licensed use limited to: BIRLA INSTITUTE OF TECHNOLOGY AND SCIENCE. Downloaded on September 16,2020 at 12:27:20 UTC from IEEE Xplore. Restrictions apply.
˜
satellite system and d(t) ∈ R3×1 denotes the external forces the system trajectories to reach the steady-state in
disturbances of the satellite system under consideration. finite-time, the expression of which is given by
In Eq. (1), S(ω̃(t)) is a skew-symmetric matrix of the
form √ 1
1 2( n) 2 M1
Tf ixed ≤ ( + )(1+ √ )
 
0 −ω3 (t) ω2 (t) p2 (p − 1) p−1 p1 m1 (1 − p2α 1
)
S(ω̃(t)) =  ω3 (t) 0 −ω1 (t) (2) 1
(9)
−ω2 (t) ω1 (t) 0
where  is a positive constant and n denotes the degree .
∀ ω̃(t) ∈ R3×1 . The expression of control algorithm The constants M1 , m1 , p1 and α1 satisfy the following
considering faults in the actuators can be written as conditions.
ũf (t) = A1 ũ(t) + B (3)
where A1 ∈ R 3×3
is 3 × 3 diagonal matrix which M1 = α1 + L
represents the gain faults in the actuators and B ∈ R3×1 m1 = α1 − L
represents bias faults in the actuators. α1 > 4L
Eq. (3) can also be rewritten as √
p1 > 2α1 (10)
ũf (t) = ũ(t) + (A1 − I)ũ(t) + B
As the upper bound of convergence-time depends on ,
= ũ(t) + ũ2 (t) (4) so the minimum value of upper bound occurs when  =
1
where I ∈ R3×3 is an identity matrix and ũ2 (t) is that (n 4 p1 /p2 )1/p+1/2 .
part of control input which can be clubbed with external Proof: The satellite dynamics can be rewritten as
disturbances. ˙
ω̃(t) = −J −1 S(ω̃(t))J ω̃(t) + J −1 ũ(t) + J −1 d˜1 (t)
The expression of attitude dynamics can also be rewritten
as = −J −1 S(ω̃(t))J ω̃(t) + ṽ(t) + J −1 d˜1 (t) (11)
˙
J ω̃(t) ˜
= −S(ω̃(t))J ω̃(t) + ũ(t) + ũ2 (t) + d(t) Performing a change of coordinates with ω̃(t) = z̃(t) ∈
R3×1 , ṽ(t) = ũ1 (t) + J −1 S(ω̃(t))J ω̃(t) and J −1 d˜1 (t) =
˙
J ω̃(t) = −S(ω̃(t))J ω̃(t) + ũ(t) + d˜1 (t) (5)
ζ(t). Therefore Eq. (11) can also be written as
Remarks-1: The external disturbance is assumed to be
bounded in nature and the expression of upper bound is ˙ = −p1 z̃(t) 1 − p2 z̃(t)kz̃(t)kp−1 + z1 (t)
z̃(t)
known in advance. It is also assumed that the external kz̃(t)k 2
disturbance is Lipschitz bounded in nature. z̃(t)
z̃˙1 (t) = −α1 + µ(t) (12)
kz̃(t)k
III. C ONTROL ALGORITHM
In this section, the expression of designed control where z̃(t0 ) = z̃0 = ω̃0 , z̃1 (t0 ) = 0 and µ(t) = ζ̇(t).
algorithm is stated. The expression of sliding surface for µ(t) is also Lipschitz bounded in nature.
fixed-time fault-tolerant sliding mode control algorithm is Assuming kω̃0 k > . The first part of Eq. (12) can also
given by be written as
s̃(t) = ω̃(t) − ω̃d (t). (6) d(kz̃(t)k)
≤ −p2 kz̃(t)kp (13)
The expression of control input in case ω̃d (t) = dt
[0 0 0]T rad/sec is given by Now, considering sign(zi (t)) and sign(z1i (t)) to be
Z t opposite in sign ∀i = 1, 2, ......, n and zi (t) > 0.
s̃(t) p−1 s̃(a) Furthermore, α1 > L. Eq. (13) can be rewritten as
ũ1 (t) = −p1 1 −p 2 s̃(t)ks̃(t)k − α1 da
ks̃(t)k 2 t0 ks̃(a)k
(7)
where p1 , p2 and α1 ∈ R+ . The constant p > 1. d(kz̃(t)k)
= kz̃(t)k−p d(kz̃(t)k) ≤ −p2 dt (14)
Assumption 1: As stated in Remarks-1, that the exter- kz̃(t)kp
nal disturbance is Lipschitz bounded in nature then the Integrating the above equation with z̃(t0 ) = z̃0 ,
following equation holds true.
kz̃(t)k1−p kz̃0 k1−p
− ≤ −p2 (t − t0 )
kJ −1
d˜1 (t)k ≤ L(t − t0 ) (8) 1−p 1−p
kz̃(t)k1−p kz̃0 k1−p
where L is the Lipschitz constant. ≤ −p2 (t − t0 ) +
Proposition 1: For the satellite system with the dy- 1−p 1−p
namics given in Eq. (1) as well as the assumption stated kz̃(t)k1−p
≤ −p2 (t − t0 ) (15)
in assumption-1, the control algorithm given in Eq. (7) 1−p

486

Authorized licensed use limited to: BIRLA INSTITUTE OF TECHNOLOGY AND SCIENCE. Downloaded on September 16,2020 at 12:27:20 UTC from IEEE Xplore. Restrictions apply.
Multiplying both sides of Eq. (15) by p − 1. under the influence of control law
Z t
−kz̃(t)k1−p ≤ −p2 (t − t0 )(p − 1) (16) z̃(t) z̃(a)
ũ1 (t) = −p1 1 − α1 da (25)
kz̃(t)k 2 t0 kz̃(a)k
Applying rules of mathematical inequality and consider-
ing the value of constant p. is finite-time stable. The initial conditions are z̃(t0 ) 6= 0
1 and z̃1 (t) = 0.
kz̃(t)kp−1 ≤ (17) Proof: Considering a Lyapunov function candidate as
p2 (p − 1)(t − t0 )
given below
Hence the time taken to reach kz̃(t)k =  is given by
1
1 V (z̃(t), z̃1 (t)) = 2α1 kz̃(t)k + (kz̃1 (t)k2 +
T1 ≤ (18) 2
p2 (p − 1)()p−1 z̃(t)
T1 forms the first component of Tf ixed and is valid only kp1 − z̃1 (t)k2 ) (26)
kz̃(t)k
if kω̃0 k > . When kω̃0 k <  and t > T1 . Furthermore as
kz̃(t)k decreases when t > T1 ,then the first part of Eq. The above Lyapunov function is similar to that given in
(12) can be written as the proof of Proposition 1 of Ref. (Nagesh and Edwards
(2014)) if κ2 and κ4 is considered zero. On differentiating
d(kz̃(t)k) 1
the Lyapunov function given in Eq. (26) (Nagesh and
= −p1 kz̃(t)k 2 (19)
dt Edwards (2014)).
On integrating Eq. (19),
˙
p21 z̃(t)T z̃(t)
1
2kz̃(t)k ≤ −p1 (t − T1 ) + 2kz̃(T1 )k
2
1
2 V̇ (z̃(t), z̃1 (t)) = (2(α1 ) + ) + 2z̃1 (t)T z̃˙1 (t)
2 kz̃(t)k
1
≤ −p1 (t − T1 ) + 2 2 (20) ˙
1 (z̃(t)T z̃(t))(z̃ T
1 (t) z̃(t))
− p1 (− 5 +
Therefore, the time taken by each component of z̃(t) to 2 kz̃(t)k 2
reach zero is given by (z̃˙1 (t)T z̃(t)) + (z̃1 (t)T z̃(t))
˙
) (27)
√ 1 kz̃(t)k 2
1

2( n) 2
T2 ≤ (21)
p1 On replacing the system dynamics given in Ref. (24) in
T2 forms the second component of the time Tf ixed . At Eq. (27)
the end of T2 , the components of z̃(t) reach zero. From p31 kz̃(t)k2 T
2 z̃(t) z̃1 (t)
the second part of Eq. (12), kz̃1 (T2 )k is given by V̇ (z̃(t), z̃1 (t)) = −(p1 α1 + ) 3 + p1
2 kz̃(t)k 2 kz̃(t)k
kz̃1 (T2 )k < M1 (T2 − t0 ) p1 (z̃(t)T z̃1 (t))(z̃1 (t)T z̃(t)) z̃1 (t)T z̃1 (t)
√ 1 + 5 − p1 1
1 2( n) 2 2 kz̃(t) 2 k kz̃(t)k 2
< M1 ( p−1
+ ) (22)
p2 (p − 1) p1 µ(t)T z̃(t)
+ 2z̃1 (t)µ(t) − p1 1 (28)
Considering the proof of Theorem 4.5 as given in kz̃(t)k 2
Ref. (Shtessel et al. (2014)) and Lemma 1 of Ref.
(Basin et al. (2016)), the 1time taken to reach from
√ p31 1 |z̃(t)T z̃1 (t)T |
1
(0, M1 ( p2 (p−1) + 2( n) 2
) to origin is given by V̇ (z̃(t), z̃1 (t)) ≤ −(p1 α1 + )kz̃(t)k 2 + p21
p−1 p1 2 kz̃(t)k
√ 1 T
p1 |z̃(t) z̃1 (t)|2
|µ(t)T z̃(t)|
1 2( n) 2 1 + + 2z̃1T µ(t) + p1
T ≤ M1 ( + )( √ ) 2 kz̃(t)k 2 5
kz̃(t)k 2
1
p2 (p − 1)p−1 p1 m1 (1 − p2α 1
)
1 (29)
(23)
Using the expression of T1 , T2 and T , the final expression Applying Cauchy-Schwartz inequality on Eq. (29).
of Tf ixed can be easily obtained. Optimum value of  can
be obtained by minimizing the first two values of Tf ixed p31 1
V̇ (z̃(t), z̃1 (t)) ≤ −(p1 α1 + )kz̃(t)k 2 + p21 kz̃1 (t)k
with respect to . 2
Lemma 1: A closed loop multi-variable system is given p1 kz̃1 (t)k2 1
1 + 2Lkz̃1 (t)k + p1 Lkz̃(t)k
2
by 2 kz̃(t)k 2
(30)
˙ = −p1 z̃(t) 1 + z1 (t)
z̃(t)
kz̃(t)k 2 Eq. (30) can also be reframed as
z̃(t)
z̃˙1 (t) = −α1 + µ(t) (24) aT Σa
kz̃(t)k V̇ (z̃(t), z̃1 (t)) ≤ − (31)
kz̃(t)k1/2

487

Authorized licensed use limited to: BIRLA INSTITUTE OF TECHNOLOGY AND SCIENCE. Downloaded on September 16,2020 at 12:27:20 UTC from IEEE Xplore. Restrictions apply.
where a = [kz̃(t)k1/2 , kz̃1 (t)k]T . The matrix Σ can also
be written as
 3
p /2 + p1 (α1 − L) −p21 /2 − L

Σ= 1 (32)
−p21 /2 − L −p1 /2
Positive definitiveness of matrix Σ can be ensured by
p21 > 2L
α1 ≥ 3L + 2L2 /p21 (33)
From the second part of Eq. (33), the first inequality of
Eq. (10) can be proven. As, α1 > L (Utkin (2013a)). The
first part of Eq. (33) can be written as p21 > 2α1 which
Fig. 1. Plots of angular rates for fixed-time fault-tolerant super-twisting
proves the second inequality given in Eq. (10). sliding mode control for Case-a.
Equation (31) can be rewritten as
kak2
V (z̃(t),˙ z̃ 1 (t))≤ −λmin (Σ) (34) M1 and m1 is calculated as 5.5 and 3.5. The expression
kz̃(t)k1/2
of ũ2 (t) is considered as
λmin denotes the minimum eigen value of the matrix Σ. (
Making a transformation according to the relation [0 0 0 ]0 ∀t < 2sec
ũ2 (t) =
z̃(t) 10−4 [sin(t) cos(1.5t) (sin(2t) + 1)]0 ∀t ≥ 2sec
a1 = [ , kz̃1 (t)k]T (35)
kz̃(t)k1/2
The value of  as calculated using the formulae is given
equation (34) can also be written as by
ka1 k2 1 5 1
V̇ ≤ −λmin (Σ) (36)  = (3 4 ) 2
kz̃(t)k1/2 5
The Lyapunov function considered in Eq. (26) can be = 1.1472 (41)
written as
V (z̃(t), z̃1 (t)) = aT1 Aa1 (37) Using the value of  as given in Eq. (41), the expression
of Tf ixed can be obtained as
where A matrix is considered as √ 1

(2α1 + p21 /2)In −(p1 /2)In
 1 2( 3(1.1472)) 2
A= (38) Tf ixed ≤ ( + )
−(p1 /2)In In 5(1.5 − 1)1.14721.5−1 5
5, 5
With the help of Rayleighs inequalities, (1 + √ ) (42)
2(4.5)
3.5(1 − 5 )
V ≤ λmax (A)ka1 k2
1 p ≤ 4.62sec (43)
V ≥ λmin (A)ka1 k2
2 (39)
The initial values of angular-rates for simulations is
Equation (36) can be written as
considered as ω̃0 = [0.06 0.04 0.05]rad/sec for Case-a
V̇ ≤ −kV 1/2 (40) and ω̃0 = [0.6 0.4 0.5]rad/sec for Case-b respectively.
√ The steady state value of the angular-rates is considered
λ (Σ) λmin (A)
where k = min λmax (A) . From Eq. (40) it can be as ω̃t = [0 0 0]rad/sec. The plots of angular rates and
easily stated that the system is finite-time stable. control-inputs for Case-a is given in Fig-(1-2). From the
plots of angular rates as given in Fig-(1), it can be said
IV. R ESULTS that the control algorithm successfully stabilizes the rates
In this section numerical simulations are performed to to their steady state values within the value of Tf ixed
illustrate the efficacy of the designed control algorithm. calculated before in this section in spite of faults in
The inertia matrix of the satellite system is considered the control actuator. From the plots of control input as
as J = diag(20, 21, 22)kg − m2 . The values of con- given in Fig-(2), it can be said that the control input is
stants are chosen as p1 = 5 , p2 = 5, p = 1.5 and free from chattering which can further be explained from
α1 = 4.5. The external disturbance is considered as the continuous nature of designed control input. Despite
10−2 [0.62sin(t) 0.79cos(1.5t) 1.03(sin(2t)+1)]T N −m. changes in the initial conditions of the angular rates as
The value of Lipschitz constant (L) is considered as 1. For for Case-b, the plots of angular rates and control input
the considered value of Lipschitz constant, the values of can be justified in a similar way as for Case-a.

488

Authorized licensed use limited to: BIRLA INSTITUTE OF TECHNOLOGY AND SCIENCE. Downloaded on September 16,2020 at 12:27:20 UTC from IEEE Xplore. Restrictions apply.
R EFERENCES
Basin, M., Panathula, C. B. and Shtessel, Y. (2016). Mul-
tivariable continuous fixed-time second-order sliding
mode control: design and convergence time estimation,
IET Control Theory & Applications 11(8): 1104–1111.
Blanke, M., Izadi-Zamanabadi, R., Bøgh, S. A. and
Lunau, C. P. (1997). Fault-tolerant control systemsa
holistic view, Control Engineering Practice 5(5): 693–
702.
Fridman, L. and Levant, A. (2002). Higher order sliding
modes, sliding mode control in engineering, barbot, j.
p and perruguetti, w. eds.
Fig. 2. Plots of control input for fixed-time fault-tolerant super-twisting
sliding mode control for Case-a. Han, Z., Zhang, K., Yang, T. and Zhang, M. (2016).
Spacecraft fault-tolerant control using adaptive non-
singular fast terminal sliding mode, IET Control Theory
& Applications 10(16): 1991–1999.
Lo, S.-C. and Chen, Y.-P. (1995). Smooth sliding-
mode control for spacecraft attitude tracking maneu-
vers, Journal of Guidance, Control, and Dynamics
18(6): 1345–1349.
Nagesh, I. and Edwards, C. (2014). A multivari-
able super-twisting sliding mode approach, Automatica
50(3): 984–988.
Perruquetti, W. and Barbot, J. P. (2002). Sliding mode
control in engineering, Vol. 11, M. Dekker.
Polyakov, A. and Poznyak, A. (2009a). Lyapunov
function design for finite-time convergence analy-
Fig. 3. Plots of angular rates for fixed-time fault-tolerant super-twisting
sliding mode control for Case-b. sis:twisting controller for second-order sliding mode
realization, Automatica 45(2): 444–448.
Polyakov, A. and Poznyak, A. (2009b). Reaching
time estimation for super-twisting second order sliding
V. C ONCLUSIONS
mode controller via lyapunov function designing, IEEE
Transactions on Automatic Control 54(8): 1951–1955.
This paper implements a fault-tolerant fixed-time super- Shtessel, Y., Edwards, C., Fridman, L. and Levant,
twisting sliding mode control for angular-rates stabi- A. (2014). Sliding mode control and observation,
lization of satellite system. Finite-time stability and the Springer.
expression of convergence time are calculated. Numerical Utkin, V. (1977). Variable structure systems with slid-
simulations at different values of initial conditions vali- ing modes, IEEE Transactions on Automatic control
date the effectiveness of the designed control algorithm. 22(2): 212–222.
Utkin, V. (2013a). On convergence time and disturbance
rejection of super-twisting control, IEEE Transactions
on Automatic Control 58(8).
Utkin, V. I. (2013b). Sliding modes in control and
optimization, Springer Science & Business Media.
Venkataraman, S. and Gulati, S. (1991). Terminal sliding
modes: a new approach to nonlinear control synthesis,
Fifth International Conference on Advanced Robotics’
Robots in Unstructured Environments, IEEE, pp. 443–
448.
Yu, X., Li, P. and Zhang, Y. (2017). The design of fixed-
time observer and finite-time fault-tolerant control for
hypersonic gliding vehicles, IEEE Transactions on In-
dustrial Electronics 65(5): 4135–4144.
Fig. 4. Plots of control input for fixed-time fault-tolerant super-twisting
sliding mode control for Case-b.

489

Authorized licensed use limited to: BIRLA INSTITUTE OF TECHNOLOGY AND SCIENCE. Downloaded on September 16,2020 at 12:27:20 UTC from IEEE Xplore. Restrictions apply.

You might also like