You are on page 1of 5

IEEE INDICON 2015 1570186471

1  
2  
3   Fast Terminal Sliding Mode Control for Twin Rotor
4  
5  
6  
Multi-Input Multi-Output System
7  
8  
9  
Shantanu Singh Dr. S Janardhanan Dr. Mashoq-un-Nabi
10  
11   Department of Electrical Engineering Department of Electrical Engineering Department of Electrical Engineering
12   Indian Institute of Technology Indian Institute of Technology Indian Institute of Technology
13   New Delhi 110016 New Delhi 110016 New Delhi 110016
14   Email:singh shantanu@hotmail.com Email: janas@ee.iitd.ac.in Email: mnabi@ee.iitd.ac.in
15  
16  
Abstract—Twin rotor multi-input and multi-output Section 2 illustrates the dynamics of the twin rotor multi-input
17  
system(TRMS) is used as the nonlinear multi-input multi- and multi-output system. Section 3 illustrates the controller
18   output(MIMO) system and a robust control approach based design both using terminal sliding manifold and fast terminal
19   on fast terminal sliding surface is proposed in the paper. The sliding manifold. Section 4 details the stability analysis, Sec-
20   nonlinear TRMS system is reduced to a simpler model and the tion 5 presents a comparative study of results using the two
21   error is handled by fast terminal sliding mode control. The
proposed control inputs. Section 6 is the conclusion of the
22   results are provided which prove the advantage of using fast
terminal sliding mode control. paper.
23  
24  
25   Index Terms- Terminal Sliding Mode (TSM), Fast Terminal
26   Sliding Mode (FTSM), Finite time convergence, Robustness,
Uncertainty.
27  
28  
29   I. I NTRODUCTION
30   TRMS is a highly nonlinear system which leads to a
31   complex design process of the controller [4], [5], [12]. The
32   nonlinear system is reduced to less complex system with
33   uncertainty. A controller needs to be designed which is ro-
34   bust and can provide good performance despite of uncer-
35   tainties/perturbation. Sliding mode control which is a class
36   of variable structure control has been studied considerably
37   over the past few decades for its efficient robustness in ap-
38   plications like robotic manipulators, air-crafts, power systems,
motors, etc [1], [3], [14], [13]. Linear sliding surface used for
39  
TRMS have proved effective but there has been a problem
40   of chattering which is dealt by using chattering suppressing Fig. 1. Twin Rotor MIMO System
41   techniques like boundary layer solution, regular form solution
42   and more [1], [2], [3]. There has been a lot of discussion on II. DYNAMIC M ODEL OF TRMS
43   the convergence property of sliding mode control and terminal
44   The TRMS has two rotors placed on a beam together with
sliding surface offers fast convergence and finite time solution.
45   a counterbalance whose arm has a weight at its end which
It has been observed that when the system states are far from
is fixed to the beam at the pivot and it determines a stable
46   the equilibrium point, terminal sliding mode control is not so
equilibrium position. The beam is pivoted on its base in such
47   effective in terms of convergence rate as compared to linear
a way that it can rotate freely both in the horizontal and
48   sliding surface. Fast terminal sliding manifold is a modification
vertical planes. Both of the rotors are driven by DC motors.
49   of terminal sliding manifold and contains an additional linear
The main rotor produces a lifting force allowing the beam to
50   term which dominates the convergence rate when states are
rise vertically. The tail rotor is used to change the yaw position
51   far from the equilibrium point [6], [8], [10]. The finite time
of the twin rotor system. Apart from the mechanical unit, the
52   convergence is still assured. In this paper a fast terminal sliding
electrical unit is used to send control signals via an I/O card.
surface is used to compute the control input for the twin rotor
53   The system is multi-variable, nonlinear and significant cross-
system. The system is highly nonlinear thus a simplified model
54   of TRMS is used and some of the nonlinearities which are
coupling between the actions of the rotors. Pitch and yaw
55   present are approximated as system uncertainties [5], which is
angle denoted by ψ and φ respectively are measured using
56   compensated by fast terminal sliding mode control. This paper
high resolution encoders.
57   compares the convergence rate and time of terminal sliding The states of the beam is described by four variables:
60   mode and fast terminal sliding mode. The results demonstrate horizontal and vertical angles and the two corresponding
61   effective tracking performance of the states. angular velocities.
62  
63  
64  
65  

1
The momentum equation for the vertical system is given τ1 = kτ1 u1 + Δτ1 (s) (14)
as
I1 ψ̈ = M1 − MF G − MBψ − MG (1) τ2 = kτ2 u2 + Δτ2 (s) (15)

where The nonlinear static characterstic where


ki
M1 = a1 τ12 + b1 τ1 (2) kτi =
Ti1
The gravity momentum, friction forces momentum and gyro- kc (T0 − Tp )s
scopic momentum is given by ΔMR (s) = τ1 (s)
Tp s + 1
MF G = Mg sin ψ (3) ki (Ti1 − Ti0 − Ti1 s)
Δτi (s) = ui (s) (16)
MBψ = B1ψ ψ̇ + B2ψ sign(ψ̇) (4) Ti1 (Ti1 s + Ti0 )
MG = kgy M1 φ̇ cos ψ (5) The uncertainties in equations (13), (14), (15) can be compen-
sated by the controller by choosing higher gain value.
The motor and the electric control circuit are approximated
by a first order transfer function thus in Laplace domain the States variables and input vector are given below,
motor momentum is described by
x = [ψ φ ψ̇ φ̇]T
k1 v = [x1 x2 ]T
τ1 = (6)
T11 s + T10 w = [x3 x4 ]T
The momentum equation for the horizontal movement is given u = [u1 u2 ]T
by
(17)
I2 φ̈ = M2 − MBφ − MR (7)
State space equation is given by
The nonlinear static characteristic
v̇ = w
M2 = a2 τ22 + b2 τ2 (8) ẇ = F (x, u) + Δ(x) (18)
Friction force momentum where
MBφ = B1φ φ̇ + B2φ sign(φ̇) (9)
F (x, u) =
and MR is the cross reaction momentum approximated by  −1 
I1 (ku1 Mu1 − Mg sin x1 − B1ψ x3 − B2ψ sign(x3 ))
kc (T0 s + 1) I2−1 (Mu1 − B1φ x4 − B2φ sign(x4 ) − kc kτ1 u1 )
MR = τ1 (10)
(Tp s + 1) Δ(x) =
 −1 
The tail DC motor with electrical circuit is given by I1 (1 − Kgy x4 cos(x1 ))(a1 Δ2τ1 + (2a1 Kτ1 u1 + b1 )Δτ1
k2 I2−1 (a2 Δ2τ2 + (2a2 kτ2 u2 + b2 )Δτ2 − kc Δτ1 − ΔMr )
τ2 = u2 (11) (19)
T21 s + T20
ku1 = 1 − Kgy x4 cos(x1 )
The output is given by
Mu1 = A1 u21 + B1 u1
y = [ψ φ]T (12) Mu2 = A2 u22 + B2 u2
where ψ, φ, τ1 , τ2 are pitch angle, yaw angle, main motor Ai = ai kτ2i
torque and tail rotor torque respectively. u1 , u2 are inputs to B i = bi k τ i
the system and are equal to the voltage provided to the two
DC motors. Ii is the moment of inertia of rotor, ai and bi are Inputs to the system are of quadratic form. This is not desirable
static characteristic parameters, Mg is moment of gravitational for sliding mode controller design, hence we approximate
force, Biθ is friction momentum parameters, ki are motor Mu1 , Mu2 by a linear function of inputs plus a error term
gains, TTi1
i0
are motor time constants, kc is cross coupling which will be compensated by the controller.
gain, T0 , Tp are cross coupling time constants. The system
parameters as given in the manual are: I1 = 0.068kgm2 , I2 = Mu 1 = c 1 u 1 + d 1 + Δ u 1 (20)
0.02kgm2 , a1 = 0.0135, b1 = 0.0924, a2 = 0.02, b2 = Mu 1 = c 2 u 2 + d 2 + Δ u 2 (21)
0.09, Mg = 0.32N m, B1ψ = 0.006N ms/rad, B2ψ =
0.001N ms/rad, B1φ = 0.1N ms/rad, B2φ = Now substituting equations (19), (20), (21) in equation (18),
0.01N ms/rad, Kgy = 0.05s/rad, k1 = 1.1, k2 = 0.8, T11 = we get
1.1, T10 = 1, T21 = 1, T20 = 1, Tp = 2, T0 = 3.5, kc = −0.2. v̇ = w
Equations (10), (6), (11) can be written in a simplified form ẇ = f (x) + B(x)u + Δ(x, u) (22)
which is linear in torques τ1 , τ2 [5] and approximated as
MR = kc τ1 + ΔMR (s) (13) where

2
 
I1−1 ku1 c1 0 Control input is calculated using equations (27) and (29) as
B(x) =
−I2−1 kc kτ1 I2−1 c2
p
u = −B −1 (f (x) + Ksign(s) + β w̃ + α ṽ (p−q)/q w̃) (30)
f (x) =  q
I1−1 (ku1 d1 − Mg sin x1 − B1ψ x3 − B2ψ sign(x3 ))
I2−1 (d2 − B1φ x4 − B2φ sign(x4 ) − kc kτ1 u1 ) The stability analysis is carried out in

Δ( x, u) =
⎡ ⎤
I1−1 ku1 (Δu1 + a1 Δ2τ1 + (2a1 kτ1 u1 + b1 )Δτ1 ) IV. S TABILITY A NALYSIS
⎣ I −1 (Δu + a2 Δ2 + (2a2 ktau u2 + b2 )Δtau − kc Δtau ⎦
2 2 τ2 2 2 1
−ΔMr ) Let us consider a Lyapunov function candidate as

All system uncertainties and nonlinearities are put together 1 T


V = s s (31)
as Δ(x, u) ∈ 2×1 . This approximation is done to decouple 2
the control inputs and to make the state space suitable for
controller design [5]. Differentiating V with respect to time

III. T ERMINAL AND FAST T ERMINAL S LIDING M ODE V̇ = sT ṡ


p
C ONTROL D ESIGN V̇ = sT (ẇ − ẇd + β w̃ + α ṽ (p−q)/q w̃)
q
We define vd = [ψd φd ]T , wd = v̇d = [ψ̇d φ̇d ]T ṽ = p
v − vd and w̃ = w − wd The standard linear sliding surface is V̇ = sT (f + B(x)u + Δ − w˙d + β w̃ + α ṽ (p−q)/q w̃)
q
replaced by a nonlinear sliding surface and is called terminal
sliding mode. It must be noted that wd = 0 since the twin rotor Now substituting u as given by equation (29)
is to stay at the desired pitch and yaw angles. This nonlinear
sliding manifold is given by V̇ = −sT (Ksign(s) − Δ)
s = w̃ + αṽ p/q (23) V̇ = −||s||(K − Δsign(s))
2×1
where v, w ∈  and α > 0, p and q are positive odd
It is assumed that the bound on Δ(x, u) is known. If we choose
integers such that q > p. Thus when s = 0 then
K such that ||Δmax || < K then it can be concluded that
w̃ = −αṽ p/q (24)
V̇ < − ||s||(K − ||Δmax ||)
For initial point v(0) = 0, v will reach equilibrium point in
finite time[6], [7] V̇ < − ||s||ρ
αq V̇ < − ρV 1/2 (32)
t = ṽ(0)(q−p)/q
(q − p)
where ρ = K − ||Δmax || and ||Δmax || is the known upper
Sliding manifold dynamics is given by
bound of Δ(x, u). It can be said from Lyapunov stability
ṡ = −Ksign(s) (25) theorem that the system is asymptotically stable. Equation (32)
means that s → 0 in finite time. Hence from definition of s
Control input is calculated from equations (21) and (24) as as given in equation (27), when s → 0 then ṽ → 0 or we can
p say vd → v.
u = B −1 (w˙d − f (x) + Ksign(s) + α ṽ (p−q)/q ) (26)
q
Similar method is applied for terminal sliding mode con-
It should be noted that for ||v|| > 1 the nonlinear term αv 1/3 trol, where for the same choice of Lyapunov function candi-
reduces the convergence rate. In order to overcome this we date, Lyapunov stability condition V̇ < 0 holds.
propose a modification of terminal sliding manifold, which is
given by
s = w̃ + βṽ + αṽ p/q (27) V. R ESULTS

where β > 0. Hence when s = 0 then The system dynamics are simulated in MATLAB. Simula-
tion results for both terminal and fast terminal sliding mode
w̃ = −βṽ − αṽ p/q (28) controllers are plotted and compared. We must first estimate
Linear term dominates the convergence rate when |v| > 1. the upper bound of Δ. The upper bound of Δτi is calculated
Both terminal and fast terminal sliding modes are finite time using equation(16). |Δτi | ≤ 0.25. ΔMr is calculated as well
stable [6], [8]. Fast terminal sliding mode control is applied to and it is safe to choose |ΔMr | ≤ 0.8. Δ = [δ1 δ2 ]T , and
twin rotor multi-input and multi-output system and results are δ1 , δ2 are calculated as |δ1 | ≤ |4.26 − 0.22x2 cos(x1 )|, |δ2 | ≤
shown in the next section. Sliding manifold dynamics is given 22. As required by the controller, K is chosen as K = 30.
by Abiding by the conditions in III given as α > 0, β > 0, it is
appropriate to choose α = 1, β = 1. The system is simulated
sṡ ≤ 0 ∀t > 0 (29) and results plotted as follows,

3
Pitch over time Pitch and Yaw over time

4 4

pitch
2
x1

0 0

−2 −2
0 2 4 6 8 10 0 2 4 6 8 10
time(sec) time(sec)
Yaw over time

4 4

2 2

yaw
x2

0 0

−2 −2
0 2 4 6 8 10 0 2 4 6 8 10
time(sec) time(sec)

Fig. 2. Twin Rotor MIMO System States Variables using terminal sliding Fig. 5. Convergence of pitch (ψ) and yaw (φ) using fast terminal (red) and
mode controller terminal sliding (blue) manifolds

Angular Velocities over time


2
FTSM

Pitch angular velocity


Pitch over time
TSM
4 0

2
−2
x1

0
−4
0 2 4 6 8 10
−2 time(sec)
0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
time(sec) 2
Yaw angular velocity

Yaw over time

4 0
3
2
−2
x2

1
0
−4
−1 0 2 4 6 8 10
time(sec)
1 2 3 4 5
time(sec)

Fig. 6. Convergence of (ψ̇) and (φ̇) using fast terminal and terminal sliding
Fig. 3. Twin Rotor MIMO System States Variables using fast terminal sliding manifolds
mode controller
It can be observed from Figure 2 and Figure 3 that the
states converge to equilibrium point xd = [0 0 0 0]T
from initial point x0 = [4 3 0 0]. There is negligible
Sliding Variable over time
2
overshoot though chattering phenomenon can be observed
near the equilibrium point. This is due to the discontinuous
0 nature of the control input and high gain of controller. The
sliding variables converge s = 0 in finite time though it does
s1

−2
not stay there and remains within a region, in this case
−4 < 10− 3. Figure 5, Figure 6 and Figure 7 show a comparison
0 2 4 6 8 10 between the convergence of the states to equilibrium point
time(sec) using fast terminal sliding mode control and terminal sliding
2 mode control, From Figure 5 and Figure 6 it is observed that
the states converge faster in case of fast terminal sliding mode
0
control. The linear term −α(vd − v) has significant effect on
the convergence rate when ||v − vd || > 1. Pitch angle(ψ) goes
s2

−2
to desired point in t = 2sec using FTSM control while in
−4 case of terminal sliding mode control t = 3.8sec . Yaw angle
0 2 4 6 8 10 converges to desired point for both FTSM and terminal sliding
time(sec)
mode in t = 2sec and t = 3.2sec respectively. It should be
noted that in dynamic systems like air-crafts and UAVs fast
Fig. 4. Sliding Variables for fast terminal sliding mode controller convergence means quick and better manoeuvring.

4
Pitch and Yaw over time [6] Lin Tie, Kai-Yuan Cai,”A general form and improvement of fast terminal
sliding mode” 8th World Congress on Intelligent Control and Automation
0.5 (WCICA), pp.2496-2501, 2010.
[7] S.T.Venkataraman,S. Gulati, ”Control of Nonlinear Systems Using Ter-
pitch

0 minal Sliding Modes”, American Control Conference, pp. 891-893, 1992.


[8] Shuanghe Yu, Xinghuo Yu, Man Zhihong, ”Robust global terminal
−0.5 sliding mode control of SISO nonlinear uncertain systems” Proceedings
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
of the 39th IEEE Conference on Decision and Control, vol.3 pp. 2198-
time(sec) 2203, 2000.
[9] A. Rahideh and M.H.Shaheed, ”Mathematical dynamic modeling of a
0.5 twin-rotor multiple input-multiple output system”, P I MECH ENG I-J
SYS, vol. 221 (I1), pp. 89-101, 2006.
0 [10] Xinghuo Yu and Man Zhihong, ”Fast Terminal Sliding-Mode Control
yaw

Design for Nonlinear Dynamical Systems” IEEE Transaction on Circuits


−0.5 and Systems I: Fundamental Theory and Application, vol. 49, no. 2,pp.
261-264, 2002.
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 [11] Man Zhihong; Xing Huo Yu, ”Terminal sliding mode control of MIMO
time(sec) linear systems” Proceedings of the 35th IEEE Conference on Decision
and Control, vol. 4, pp. 4619-4624, 1996.
Fig. 7. Convergence of pitch(ψ) and yaw(φ) when |ṽ| < 1. TSM in red and [12] Anup K.Ekbote, N.S. Srinivasan, Arun D.Mahindrakar,”Terminal Slid-
FTSM in blue ing Mode Control of a Twin Rotor Multiple-Input Multiple-Output
System”, 18th IFAC World Congress, Milano, 2011.
[13] Likun Hu, Pengfei Zhao, Ziguang Lu, ”Trajectory tracking control
In case of ||v − vd || < 1 the convergence time for based on global fast terminal sliding mode for 2-DOF manipulator”,
both terminal sliding mode and fast terminal sliding mode 2nd International Conference on Intelligent Control and Information
Processing (ICICIP), vol. 1 pp. 65-69, 2011.
is almost similar as observed from Figure 7. As mentioned
in III, fast terminal sliding mode controller has better transient [14] Shuanghe Yu, Xinghuo Yu, Bijan Shirinzadeh, Zhihong Man, ”Contin-
uous finite-time control for robotic manipulators with terminal sliding
characteristics as compared to terminal sliding mode controller mode” Automatica, vol.41, issue 11, pp. 1957-1964, November 2005.
when the states are far from equilibrium point.
Robustness is guaranteed in presence of system uncertain-
ties and perturbations as long as the upper bound of these
uncertainties and perturbations are well known.

VI. C ONCLUSION
This paper presents a robust approach to control of a non-
linear system with significant uncertainty and it was observed
that the control law proposed works effectively. It can be
concluded that modification of terminal sliding mode gives
better performance in terms of convergence rate. Future work
would include suppression of chattering phenomenon near the
equilibrium point and further improvement of control design to
adapt to the uncertainty in the system parameters and external
disturbances.

VII. ACKNOWLEDGEMENT
The authors thanks Science and Engineering Research
Board, Department of Science and Technology for funding
this work through the research project ”Investigation of Finite-
time, Optimal and Robust Non-Singular Control Techniques
with Application to Nonlinear Systems” (RP02903).

R EFERENCES
[1] J.E.Slotine and W.Li, Applied nonlinear control, Englewood CliDs,
NJ:Prentice-Hall, 1991.
[2] V.I.Utkin, Sliding modes in control and optimization, Berlin, Heidelberg,
Springer, 1992.
[3] V.I.Utkin, J.Guldner, Jingxin Shi, Sliding mode control in electromechan-
ical systems, London, Taylor and Francis, 1999.
[4] Feedback Instruments Ltd, Twin Rotor MIMO System Control Experi-
ment 33-949S, 2006.
[5] L. Huang, ”An approach for robust control of a twin-rotor multiple input
multiple output system”,IEEE International Conference on Robotics and
Automation (ICRA), pp. 4423-4428, 2011

You might also like