Robust Adaptive Integral Backstepping Control of A 3-DOF Helicopter
Robust Adaptive Integral Backstepping Control of A 3-DOF Helicopter
Robust Adaptive Integral Backstepping Control of A 3-DOF Helicopter
ARTICLE
International Journal of Advanced Robotic Systems
DOI: 10.5772/50864
© 2012 Fang et al.; licensee InTech. This is an open access article distributed under the terms of the Creative
Commons Attribution License (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use,
distribution, and reproduction in any medium, provided the original work is properly cited.
Abstract Unmanned aerial vehicles have enormous Compared with a fixed‐wing flying vehicles, a rotary‐
potential applications in military and civil fields. A wing flying vehicle holds advantages in takeoff, landing
Quanser’s 3‐DOF helicopter is a simplified and and hovering in small‐scale area. Hence, it possesses
benchmark experimental model for validating the specific applications such as building surveillance,
effectiveness of various flight control algorithms. The emergency transportation and geographical information
attitude control of the 3‐DOF helicopter is a challenging collection. In many countries, governments have invested
task since the helicopter is an under‐actuated system with vast resources into research on this kind of vehicle. A
strong coupling and model uncertainty characteristics. In Quanser’s 3‐degree‐of‐freedom (DOF) helicopter is a
this paper, an adaptive integral backstepping algorithm is simplified and benchmark experimental platform for
proposed to realize robust control of the 3‐DOF helicopter. testing the effectiveness of various flight control
The proposed control algorithm can estimate model algorithms. The control objective is to enable the three
uncertainties online and improve the robustness of the Euler angles of the helicopter to converge into the desired
control system. Simulation and experiment results values. However, the high performance control of 3‐DOF
demonstrate that the proposed algorithm performs well helicopter is a challenging task because of the existence of
in tracking and under model uncertainties. the following control difficulties. Firstly, it is an under‐
actuated system with two control inputs and three outputs.
Keywords Adaptive control, Backstepping algorithm, Secondly, some strong coupling characteristics appear
Integral action, 3‐DOF helicopter, Model uncertainties between pitch motion and travel motion. Thirdly, it is
difficult to get the accurate mathematical model and
parameters of the helicopter. Besides, sometimes the model
1. Introduction parameters may change. Therefore, these difficulties
prevent some general control algorithms from performing
In recent years, fixed‐wing and rotary‐wing unmanned well at the non‐equilibrium points or under model
aerial vehicles (UAVs) have been rapidly developed and uncertainties. Hence, it is an arduous problem to develop a
been equipped with intelligent flight control devices. [1, 16]. robust control algorithm which can not only control the
www.intechopen.com IntZheng
J Adv Fang,
Robotic Sy, 2012,
Weinan GaoVol.
and9,
Lei79:2012
Zhang: 1
Robust Adaptive Integral Backstepping Control of a 3-DOF Helicopter
Figure. 2. Cascade control architecture
2.1 Integral Backstepping Control
From the dynamic model of the helicopter it can be seen
that all of the three differential equations can be described
as the following equation:
(b). Free‐body diagram q1 q 2
(3)
Figure 1. Benchmark experimental platform of 3‐DOF helicopter q 2 ku h
where Define two tracking errors e1 and e2 :
La K f M w Lw g M f La g Mb La g
k21 k22 e1 q1d q1
M w L2w (M f Mb )L2a M w L2w (M f Mb ) L2a (4)
e2 q2 d q2
Kf Mb g M f g
k41 k42 =
M f Lh Mb Lh M f Lh Mb Lh
The setting value of q1 can be given by software running
( La M f La Mb Lw M w ) g
k62 = on a PC, the feedbacks q1 and q2 can be measured by
M w L2w (M f Mb )( L2h L2a )
sensors on the helicopter. In order to balance the numbers
and the state vector is defined by X [ , , p, p , t , t] , the of inputs and outputs, a virtual input q2d is introduced:
input vector is defined by U [u , ut ] , the output vector is
Y [ , p, t ] . Since the number of inputs is less than outputs q 2 d c1e1 q1 d (5)
in the system, it is a so‐called under‐actuated system.
where parameter indicates the integral of tracking error
From equation (2) we can conclude that the 3‐DOF
helicopter is a strong coupling, nonlinear and under‐ e1 . Then, the derivatives of two tracking errors are
actuated control system with model uncertainties. The calculated:
general control algorithm cannot obtain a pleasing
performance when the mass of the counterweight is e1 q1 d q1 q1d q 2 c1e1 e2
varying, therefore, adaptive control methods are used to (6)
e2 q 2 d q 2 c1e1 q1d e1 ku h
estimate and compensate the model uncertainties.
And we can obtain: The derivative of V is:
e 2 c12 e1 c1 c1 e 2 q1 d e1 ku h (7) V ce 2 2
1 1 c2e2 (10)
When e2 c2 e2 e1 (c2 0) , we can obtain:
Based on the Lyapunov theory, the system can keep
1 h stable when c1 , c2 , are positive. Similarly, we can obtain
u [(1 c12 )e1 (c1 c2 )e2 c1 q1d ]
k k (8) the integral backstepping control laws of elevation, pitch
and travel angle:
Construct a Lyapunov function as follows:
1 1 1
V 2 e12 e22 (9)
2 2 2
U 1 {[MwL2w (M f Mb )L2a ][(1 c21 )e1 (c1 c 2 )e 2 c1 d ] MwLw g (M f Mb )La g}
La K f
1 (11)
Up {(M f Lh Mb Lh )[(1 cp1 p )ep1 (cp1 cp2 )ep2 cp1p p pd ] Mb g M f g}
2
K f
2 2 2
p wLw (M f Mb )(Lh La ) [(1 c2 t )e (c c )e c t ]
M
d (La M f La Mb LwMw )g
t1 t1 t1 t2 t2 t1 t t d
The control input of elevation loop can be rewritten as:
for Elevation Motion
1
U {[Mˆ w L2w (M f M b ) L2a ][(1 c21 )e 1
The proposed integral backstepping method can La K f (13)
perform well in general situations, however, when the
(c 1 c 2 )e 2 c 1 d ] Mˆ w Lw g (M f M b ) La g}
mass of counterweight is varying, this method cannot
get a satisfactory performance. Since the real value of
M w cannot be measured precisely, we will replace it Take (13) into (7), the derivative of the angle velocity
ˆ . Therefore, the estimation tracking error is:
with the estimated value M w
LK
a f MwLwg Mf Lag MbLag
e2 c21e1 c1 c1e2 d e1 2 2
u
M L (Mf Mb)L
w w a MwL2w (Mf Mb)L2a
M Lg ˆ L2 (M M )L2 ][(1c2 )e (c c )e c ] (14)
[Mw w f b a 1 1 1 2 2 1 d
c21e1 c1 c1e2 d e1 w w
MwL2w (Mf Mb)L2a MwL2w (Mf Mb)L2a
M
c2e2 e1 w
{L2w[(1c21 )e1 (c1 c2)e2 c1 d ]Lwg}
M L (Mf Mb)L2a
2
w w
Construct a Lyapunov function:
1 2 12 12 1 2( 0) (15)
V
e1 e2 Mw 1
2 2 2 21[MwL2w (Mf Mb)L2a]
The derivative of V is:
M
V e1e1 e 2e 2 w ˆ
M
1[M L (Mf Mb )L2a ] w
2
w w
(16)
M ˆ
M
2 2
cz1ez1 cz2ez2 w
{Lwe 2[(1c1 )e1 (c1 c 2 )e 2 c1 d Lwg] w}
2 2
MwL2w (Mf Mb )L2a 1
4 Int J Adv Robotic Sy, 2012, Vol. 9, 79:2012 www.intechopen.com
-10
In order to make system stable, choose Mˆ w as:
-15
E lev ation(deg)
Mˆ w 1L2we 2[(1 c21 )e1
-20
(17) Desired
-30
0 5 10 15 20 25 30
Time(s)
Then we can obtain:
V c1e21 c 2e22 (18)
6
T rav el(deg)
4
Desired
3. Simulaiton and Experiment Results 2 Actual
0
In this section the proposed algorithm is compared with 0 5 10 15 20 25 30
both integral backstepping and PID control algorithms by
Time(s)
(a). Simulation results
simulation and experiment implementations. Our work -10
focuses on following three aspects:
(i) The control stability of the system. Though the -15
E le v a t io n (d e g )
stability of the control system has been theoretically -20
analysed, it is necessary to give some validations by Desired
experiments. -25 Actual
(ii) The tracking ability of the system.
-30
(iii) The adaption effect of estimating the model 0 5 10 15 20 25 30
uncertainty. Time(s)
15
Both simulations and experiments centre on the three Desired
interesting aspects. Firstly, we test the dynamic Actual
T ra v e l(d e g )
10
performance of step response. Then, different signal
inputs are used to test the tracking performance of the 5
proposed algorithm. Furthermore, the adaptive capability
is compared by varying the mass of counterweight. The 0
0 5 10 15 20 25 30
main initial conditions of the helicopter are given as Time(s)
follows: (b). Experimental results
• Initial attitude of the helicopter (deg): [x10,x30,x50] = Figure 3. Step response of PID
[−27.5, 0, 0]
• Initial mass of the counterweight (g): 1830 -10
• Sample time (ms): 1
-15
E le v a t io n (d e g )
3.1 Step Response Result Comparison
-20 Desired
Actual
In the following experiments, the dynamic and static -25
performance of the adaptive integral backstepping
controller is compared with the PID controller and -30
0 5 10 15 20 25 30
integral backstepping controller when the control inputs Time(s)
are step signals. Fig. 3, Fig. 4 and Fig. 5 depict the
8
simulation and experiment results of the three control
algorithms respectively. From the figures we can see that 6
T ra v e l(d e g )
-10 3.2 Tracking Performance Comparison
E le v a t io n (d e g )
-15
The following task is to test and compare the tracking
-20
ability of the three controllers. Fig. 6 and Fig. 7 show the
Desired
-25
Actual
control performance of elevation motion using three
-30
control algorithms given sawtooth and square signals
0 5 10 15 20 25 30 respectively. From the experiment results, it can be seen
Time(s)
that the adaptive integral backstepping control algorithm
15 gives better response speed.
Desired
E le v a tio n (d e g )
10 Actual
PID
-5
5
E levation(deg)
-10
-15
0 Desired
0 5 10 15 20 25 30
-20 Actual
Time(s)
(b). Experimental results -25
0 5 10 15 20 25 30 35 40
Figure 4. Step response of integral backstepping Time(s)
Integral Backstepping
-5
-5 E levation(deg)
-10
-10 -15
E le v a t io n (d e g )
-10
Time(s)
-15
8
-20 Desired
6 Actual
-25
T ra v e l(d e g )
0 5 10 15 20 25 30 35 40
Time(s)
4
Desired (a). Simulation results
2 PID
Actual
-10
Desired
0
E le v a t io n (d e g )
0 5 10 15 20 25 30 Actual
Time(s) -15
(a). Simulation results
-10
-20
E le v a t io n (d e g )
0 5 10 15 20 25 30 35 40
Time(s)
Actual
-30 -15
0 5 10 15 20 25 30
Time(s)
-20
0 5 10 15 20 25 30 35 40
15 Time(s)
Adaptive Integral Backstepping
Desired -10
T ra v e l(d e g )
10 Actual
Desired
E le v a t io n (d e g )
Actual
-15
5
0 -20
0 5 10 15 20 25 30 35 40
0 5 10 15 20 25 30
Time(s)
Time(s)
(b). Experimental results
(b). Experimental results
Figure 5. Step response of adaptive integral backstepping Figure 6. Tracking performance for sawtooth inputs
PID PID
-5 -14.5
Desired
E lev ation(deg)
E lev ation(deg)
-10
Actual
-15 -15
Desired
-20
Actual
-25 -15.5
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
Time(s) Time(s)
Integral Backstepping Integral Backstepping
-5 -14.5
E lev ation(deg)
E lev ation(deg)
-10
-15 -15
Desired Desired
-20 Actual Actual
-25 -15.5
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
Time(s) Time(s)
Adaptive Integral Backstepping
Adaptive Integral Backstepping
-14.5
-5
E lev ation(deg)
E lev ation(deg)
-10
-15
-15
Desired Desired
-20 Actual Actual
-15.5
-25 0 5 10 15 20 25 30 35 40
0 5 10 15 20 25 30 35 40
Time(s)
Time(s)
(a). Simulation results
(a). Simulation results
PID PID
-5
-12
E lev ation(deg)
-10
Elevation(deg)
-14
-15
-10 -14
-5
-14
E lev ation(deg)
-10
-16 Desired
-15 Actual
-18
0 5 10 15 20 25 30 35 40
-20 Desired
Actual
Time(s)
-25 (b). Experimental results
0 5 10 15 20 25 30 35 40
Figure 8. Elevation control of three control algorithms when the
Time(s)
(b). Experimental results mass is changed
Figure 7. Tracking performance for square inputs
integral backstepping algorithms cannot realize good
dynamic response in the elevation motion when some
3.3 Mass Variation
uncertainties exist in the model, but the adaptive integral
backstepping algorithm has excellent performance. The
In this section, the mass of counterweight is changed to
estimated mass value of the system is shown in Fig.9. As
examine the compensation and estimation performance
we can see, the estimated value is very close to the real
of the proposed algorithm. A 200g object is hung on the
value.
helicopter at 10 seconds and 30 seconds, and is removed
at 20 seconds. From Fig.8, it can be seen that the PID and
1.86
Simulation
[3] J.P. Su, W.J. Lu. (2001). Composite Sliding Mode
Control and Its Application to a Twin‐rotor multi‐
1.85
input multi‐output system. AASRC Transactions of
1.84
the Aeronautical and Astronautical Society of the
R.O.C, pp.275–282.
Mass(kg)
1.83
[4] S. Bouabdalla, R. Siegwart. (2005). Backstepping and
1.82
Sliding‐mode Techniques Applied to an Indoor
1.81 Micro Quadrotor. Proceedings of IEEE International
1.8
Conference on Robotics and Automation, pp. 2247‐
0 5 10 15 20 25 30 35 40
Time(s)
2252.
(a). Simulation results [5] C.D. Yang, W.H. Liu. (2002). Robust nonlinear H
decoupling control off light vehicle in hovering. In
Exeriment
1.86
1.83
parameter uncertainties. In Proceedings of the
1.82
Estimate
American control conference, pp.3454‐3459.
Real [7] A.T. Kutay, A. J. Calise, M. Idan, N. Hovakimyan.
1.81
(2005). Experimental results on adaptive output
1.8
0 5 10 15 20 25 30 35 40
feedback control using a laboratory model helicopter.
time(s)
IEEE Transactions on Control Systems Technology,
(b). Experimental results vol.13, pp.196–202.
Figure 9. Mass estimation [8] B. Kadmiry, D. Driankov. (2004). A fuzzy gain‐
scheduler for the attitude control of an unmanned
4. Conclusions and Future Works helicopter. IEEE Trans. Fuzzy Syst.12 , pp.502–515.
[9] K.S. Kim, Y. Kim. Robust backstepping control for
In this paper, an adaptive integral backstepping slew maneuver using nonlinear tracking function.
algorithm has been proposed to realize the attitude IEEE Transactions on Control Systems Technology.
control of a 3‐DOF helicopter. Simulation and experiment 11 (2003) 822‐829.
results validate that the integral action can eliminate the [10] S. Bouabdallah and R. Siegwart. (2008). Design and
tracking errors of the system and the adaptive action can Control of a Miniature Quadrotor. Advances in
estimate and compensate the model uncertainty Unmanned Aerial Vehicles, pp.171‐210.
effectively. The proposed algorithm can reduce the [11] J. Yang, W. Hsu. (2009). Adaptive backstepping
overshoot and improve response speed. In the future, the control for electrically driven unmanned helicopter.
proposed control algorithm will be tested on other similar Control Engineering Practice, vol.17, pp.903–913.
experiment platforms. [12] Z. Fang, W. Gao. (2011). Adaptive Integral
Backstepping Control for a Micro‐Quadrotor. 2011
5. Acknowledgments 2nd International Conference on Intelligent Control
and Information Processing, pp. 910‐915.
The author wants to thank the State Key Lab of [13] B. Zheng and Y. Zhong. (2011). Robust Attitude
Synthetical Automation for Process Industries (SKL) and Regulation of a 3‐DOF Helicopter Benchmark:
the Fundamental Research Funds for the Central Theory and Experiments. IEEE Transaction on
Universities under Grant No.N100408003, National Industrial Electronics, vol.58, pp.660‐670.
Science Foundation of China under Grant 61040014 and [14] W. Wang, K. Nonami and M. Hirata. (2008). Model
the Applied Basic Research Fund of Shenyang Municipal Reference Sliding Mode Control of Small Helicopter
Science Technology Project under Grant No.F10‐205‐1‐50 X.R.B based on Vision, International Journal of
for the support on this project. Advanced Robotic Systems, Vol.5, No.3, pp.235‐242.
[15] J. Dua, K. Kondakb and M. Bernardb, et al. (2008).
6. References Model Predictive Control for a Small Scale
Unmanned Helicopter. International Journal of
[1] T. Bresciani. (2008). Modeling, Identification and Advanced Robotic Systems, Vol. 5, No. 4, pp. 433‐438
Control of a Quadrotor Helicopter. Thesis, Department R. D. Garcia and K. P. Valavanis. (2009). The
of Automatic Control,Lund University, Sweden. Implementation of an Autonomous Helicopter
[2] A. Tayebi, S. McGilvray. (2006). Attitude Stabilization of Testbed. Journal of Intelligent & Robotics Systems,
a VTOL Quadrotor Aircraft. IEEE Transactions on Vol.54, No.3, pp.423‐454.
Control Systems Technology, vol.14, pp.562‐571.