Professional Documents
Culture Documents
org
Published in IET Control Theory and Applications
Received on 4th September 2007
Revised on 23rd April 2008
doi: 10.1049/iet-cta:20070335
ISSN 1751-8644
Abstract: The decoupling control of a twin rotor multi-input multi-output (MIMO) system is studied and
proposed to apply robust deadbeat control technique to this nonlinear system. First, the nonlinear problem is
identied and system model is developed. Then, it is shown that the system is able to be decoupled into two
single-input single-output (SISO) systems, and the cross couplings can be considered as disturbances to each
other. Finally, a robust deadbeat control scheme is applied to the two SISO systems and a controller is
designed for each of them. This design is evaluated in simulations, and the nal result is tested in a twin
rotor MIMO system. Comparing with a traditional system with two proportional, integral and derivative (PID)
controllers, this method is easy to follow, and the results show that the proposed scheme has less overshoot,
shorter settling time and is more robust to cross-coupling disturbances.
Introduction
www.ietdl.org
networks with well-known orthogonal least square algorithm
to model a twin rotor multi-input multi-output (MIMO)
system (TRMS).
This study addresses the decoupling and robust deadbeat
control of a TRMS based on the available models and
techniques. First, the system model is identied. It is
shown, that the identied system is able to be decoupled
into two single-input single-output (SISO) systems, and
the crossing couplings can be considered as disturbances to
each of the SISO systems. Then, a PID-based robust
deadbeat control scheme is applied to the SISO systems,
and a deadbeat controller is designed for each of them.
These robust deadbeat controllers can tolerate system
parameter changes for up to 50% [16]. This feature is used
to surpass the cross-coupling effects between the main and
tail rotors. This design is investigated in simulations using
Simulink, and compared with a system with two
independent PID controllers. The results show that the
proposed scheme has less overshoot, shorter settling time
and is more robust to cross-coupling disturbances. The
nal result is tested in a TRMS in the lab.
The paper is organised as follows. The TRMS is introduced
and described in the system and modelling section. The
system model is then presented. This is followed by the
robust deadbeat control method and procedure section,
where the details of the robust deadbeat control scheme and
our design are addressed. The system is then simulated
using simulink in the simulation and experiment section. In
this section, we also analysed and compared the simulated
results. Finally, the proposed method is tested in a TRMS
and the main ndings of this study are summarised in the
conclusion.
www.ietdl.org
The parameters in the simplied four point-mass system are
Mvl is the return torque corresponding to the force of
gravity, Mmr is the mass of the DC motor within the
main rotor, Mm is the mass of the main part of the
beam, Mtr is the mass of the DC motor within tail
rotor, Mt is the mass of the tail part of the beam, Mcb is
the mass of the counter weight, Mb is the mass of the
counter-weight beam, Mms is the mass of the main
shield, Mts is the mass of the tail shield, Im is the length
of the main part of the beam, It is the length of the tail
part of the beam, Ib is the length of the counter-weight
beam, Icb is the distance between the counter-weight and
joint, and g is the gravitational acceleration.
Consider the rotation of the beam in the vertical plane
(around the horizontal axis). The driving torques are
produced by the propellers, and the rotation can be
described in principle as the motion of a pendulum. We
can write the equations describing this motion as follows.
The main rotor model
dSv
Im Sf Fv (vm ) Vv kv g((A B) cos av C sin av )
dt
1
(1)
V2h (A B C) sin 2av
2
da v
J v
Vv Sv tr t
dt
Jv
Mv
dSv
dt
(2)
(3)
(4)
da h
S Jmr vm cos av
Vh h
dt
Jh
Sh Jmr vm cos av
D sin2 av E cos2 av F
(5)
dSh
dt
(6)
Mh
( uvv uv )
dt
Tmr
(7)
vm Pv (uvv )
(8)
duhh
1
( uhh uh )
dt
Ttr
(9)
vt Ph (uhh )
(10)
vm Pv (uvv )
(11)
vt Ph (uhh )
(12)
F h F h (v t )
(13)
F v F v (v m )
(14)
www.ietdl.org
T (s) as an example
T (s)
T (s)
v4n
s4 avn s3 bv2n s2 gv3n s v4n
(15)
1
(s=vn )4 a(s=vn )3 b(s=vn )2 g (s=vn ) 1
(16)
1
s as bs2 gs 1
4
(17)
Equation (17) is the normalised, fourth-order and closedloop transfer function. For a higher-order system, the same
method is used to derive the normalised equation. The
coefcients of the equation a, b and g are selected from
Table 1. Taking the above fourth order system as an
example, with a required settling time of 0.95 s, we can
nd the normalised settling time from Table 1
vn Ts 4:81
Therefore we require
vn
TS0
2nd
1.82
4.82
3rd
1.90 2.20
4.04
4th
4.81
5th
4:81 4:81
5:063
Ts
0:95
a 2:20;
b 3:50;
g 2:80
IET Control Theory Appl., 2008, Vol. 2, No. 11, pp. 999 1007
doi: 10.1049/iet-cta:20070335
www.ietdl.org
Hence, the transfer function is
T (s)
657:1
s4 11:1386s3 89:71889s2 363:397s 657:1
(18)
where
G1 (s)
K [K3 (s2 Xs Y )]
Gc (s)
s
15:02
3:458s2 2:225s
15:02
G(s)
s(s 2:603)(s 0:8547)
G2 (s)
s3
H1 (s) (1 Kb s)
H2 (s) Ka
The closed-loop transfer function can be written as
C(s)
15:02K [K3 (s2 Xs Y )]
4
R(S) s {3:458 15:02Kb KK3 }s3 {2:225 15:02KK3
15:02KKb K3 X }s2 {15:02Ka 15:02KK3 X
15:02KKb K3 Y }s {15:02KK3 Y }
a 2:20;
vn
b 3:50;
g 2:80
Ts0
80% of the desired settling time Ts
vn
Ts0
4:81
3:00625
Ts 80% 1:6
{K3 Y } 657:1
1003
www.ietdl.org
we have
Hence
Kb 0:243;
K3 17;
X 14:21;
1:519
1:533s 1:046
1:519
G(s)
2
(s 0:6982)(s 0:04983s 1:498)
G2 (s)
s3
0:748s2
H1 (s) (1 Kb s);
K3 7:723;
Y 38:6529
G1 (s)
Kb 0:5;
Ka 45:848
H2 (s) Ka
Ka 2:5453
X 3:131;
Y 6:963
a 2:20;
b 3:50;
g 2:80
vn
Ts0
4:81
3:00625
Ts 80%
1:6
Therefore
s4 6:6138s3 31:6314s2 76:0735s 81:6771
Comparing the characteristic equation and let K equal to 1,
1004
& The Institution of Engineering and Technology 2008
www.ietdl.org
1005
www.ietdl.org
the settling time has been shortened to 20 s and the
overshoot has been reduced about 20% in the tail rotor and
completely eliminated from the main rotor. Comparing
these responses, we can clearly note the following
For the two SISO systems: The settling time has been reduced
to about 6 and 12 s in tail and main rotors, respectively; the
amount of overshoot has been reduced about 20%.
For the 2-DOF system: The settling time has been reduced
to 20 s; the amount of overshoot has also been reduced as well.
Conclusions
References
www.ietdl.org
[8] BLYTHE P.W., CHAMITOFF G.: Estimation of aircrafts
aerodynamic coefcients using recurrent neural
networks. Proc. 2nd Pacic Int. Conf. Aerospace Science
and Technology, Australia, 1995
[9] CHON K.H., COHEN R.J.: Linear and non-linear ARMA model
parameter estimation using an articial neural network, IEEE
Trans. Biomed. Eng., 1997, 44, (3), pp. 168 174
[10] KIM B.S., CALISE A.J.: Non-linear ight control using neural
networks, J. Guid. Control Dyn., 1998, 20, (1), pp. 26 33
[11] TALEBI H.A., PATEL R.V., ASMER H.: Dynamic modelling of
exible-link manipulators using neural networks with
application to the SSRMS. Proc. IEEE Int. Conf. Intelligent
Robots and Systems, Victoria, Canada, 1998
[12] LYSHEVSKI S.E. : Identication of non-linear ight
dynamics: theory and practice, IEEE Trans. Aerosp.
Electron. Syst., 2000, 36, (2), pp. 383 392
IET Control Theory Appl., 2008, Vol. 2, No. 11, pp. 999 1007
doi: 10.1049/iet-cta:20070335
1007