Professional Documents
Culture Documents
PD Backsteping
PD Backsteping
PD Backsteping
of a Tilt-rotor UAV
Arindam Bhanja Chowdhury, Anil Kulhare and Gaurav Raina
I. I NTRODUCTION
HERE are various places where, owing to dangerous or
T onerous conditions, UAVs have to replace humans. The
main uses of UAVs are surveillance and reconnaissance. Figure 1. Tilt rotor showing the motor thrust and tilt angles of the two rotors
There are several types of UAV which are developed to with reference and body frames [4].
reach the above requirements, like fixed-wing UAV, flapping-
wing UAV and Quad-rotor UAV. Each of these has its own
merits and demerits. Our focus is to develop a UAV which speed of the rotors leads to roll control. Increasing the speed of
can take-off and land vertically as well as reach the disaster the right rotor (rotor 2) and decreasing the speed of left rotor
site quickly. Hence a Tilt-rotor UAV is an excellent choice. (rotor 1) will cause the UAV to left-roll and to move towards
The Tilt-rotor has two rotors mounted on the two sides of its the left. Doing the opposite will have the reverse effect. For
airframe, which can be tilted to provide lift and forward thrust. pitch control the rotors are tilted to the front simultaneously,
Some examples of Tilt-rotor aircraft are the Arizona State which also provides forward motion. Yaw is controlled by
University’s HARVee [1], Compigne University’s BIROTAN changing the tilt angles αR and αL of the right and left rotors
[4], large scale versions like Boeing’s V22 Osprey [2] and in opposite directions.
Bell’s Eagle Eye [1]. In [5] the control strategy used for a Quad-Tilt-rotor in-
One clear advantage of the Tilt-rotor, with respect to other volves a Linear-Quadratic-Regulator (LQR) concept, where
multi-rotors, is that it requires only two motors, allowing a the system model has been linearized about some operating
reduction in weight, volume and energy consumption. More- point. In this way a storage function is developed so that
over, with the help of two rotors it can move faster than a the error between the desired and actual values of the states
fixed-wing airplane. The two rotors (rotor 1 and 2 as shown in goes to zero in finite time. But the linearized model assumes
Figure 1) rotate in opposite directions, canceling the reaction that the system stays near the operating point, which may not
torque which is generated due to their motion, and this keeps always be the case. In [4], the stabilization of the Tilt-rotor is
the UAV stable. achieved by a non-linear control strategy, but its control law
The control aspects of a Tilt-rotor include altitude control, design involves nested saturations. The control law we will
forward motion and pitch control, lateral motion and roll use here is also a non-linear one, but is much simpler to use,
control and, yaw control. Altitude control is achieved by and also does not involve linearization, thereby removing the
varying the speed of both rotors simultaneously. Changing the assumption considered in [5].
In [5] a back-stepping based Proportional-Integral-
{Arindam Bhanja Chowdhury, Anil Kulhare} are with the Department of
Electrical Engineering, Indian Institute of Technology Madras, Chennai 600 Derivative (PID) strategy is used for a Quad-rotor, while
036, India disregarding the constraint that the value of the expression
{bhanja.abc.ind19, hereanil} @gmail.com inside the parenthesis of sin−1 in equation (25) cannot go
Gaurav Raina is with the faculty of the Department of Electrical Engineer-
ing, Indian Institute of Technology Madras, Chennai 600 036, India beyond ±1. Under such circumstances the expression of φd
gaurav@ee.iitm.ac.in will become undefined. But this will not arise in our proposed
978-1-4577-2074-1/12/$26.00 2012
c IEEE 3475
controller. There is no translational force in the yb direction. So the
We use the Euler-Lagrange approach to develop the model body force vector (Fb ) is given as
of the Tilt-rotor UAV [6]. A back-stepping based Proportional-
Derivative (PD) control law is implemented to restore the Fb = [Fxb , 0, Fzb ]T (4)
system when disturbed from its desired hovering position. We
Torques applied to the body expressed in the E-frame for
have constructed a Lyapunov function to show that the error
changing φ, θ, ψ are given by τφ , τθ , τψ respectively. Their
in roll, pitch and yaw decreases monotonically.
expressions are
The rest of the paper is organized as follows. In Section II,
we outline the mathematical model of the Tilt-rotor aircraft. ⎡⎤ ⎡ ⎤
In Section III, we deal with the control law design. In Section τφ [fL cos(αL ) − fR cos(αR )]l
IV, we perform simulations which shows how the system τ = ⎣ τθ ⎦ = ⎣ [fL sin(αL ) + fR sin(αR )]h ⎦ ,
returns to its desired position when disturbed. In Section V, τψ [fL sin(αL ) − fR sin(αR )]l
we summarize our work. (5)
where l is the distance of each rotor from point O, and h is
II. M ATHEMATICAL M ODELING the vertical distance of the COG from O, as shown in Figure
1.
The dynamic model of the Tilt-rotor is derived using
the energy-based Euler-Lagrange approach. The modeling
involves two right-handed reference frames. We consider the C. Equations of motion
inertial frame (E-frame), represented by the variables x, y, The translational kinetic energy of the system is given as
z, and the body frame (B-frame), represented by xb , yb , zb .
1 1 1
The origin of the B-frame is attached to the point O on Ttrans = mẋ2 + mẏ 2 + mż 2 , (6)
the airframe. The positive xb -axis points to the front of the 2 2 2
airframe, the positive yb -axis points to the left rotor (rotor 1) where m is the mass of the system. The body angular velocity
and the positive zb -axis points upwards. Positive roll (φ), pitch [6] [ωφ , ωθ , ωψ ]T of the system is given by
(θ) and yaw (ψ) are obtained by right-handed rotation about ⎡ ⎤ ⎡ ⎤
ωφ (φ̇ − ψ̇sθ)
the positive x, y, and z axes respectively. ⎣ ωθ ⎦ = (RT Ṙ)∨ = ⎣ (θ̇cφ + ψ̇cθsφ) ⎦ ,
So, the generalized coordinates for representing the Tilt-
ωψ (ψ̇cθcφ − θ̇sφ)
rotor model are q = [x, y, z, φ, θ, ψ]T , where (x, y, z)
represents the position of the Center Of Gravity (COG), and where (..)∨ means converting the skew symmetric matrix
(φ, θ, ψ) represents the orientation of the rotor-craft relative RT Ṙ into vector form.
to the E-frame. The kinetic energy due to rotation is given as
A. Rotation matrix 1 1
Trot = Ixx (φ̇ − ψ̇sθ)2 + Iyy (θ̇cφ + ψ̇cθsφ)2
Any vector vb in the B-frame can be transformed into a 2 2
1
vector v in the E-frame by the relation + Izz (ψ̇cθcφ − θ̇sφ)2 , (7)
2
v = Rvb , where Ixx , Iyy , Izz are the moments of inertia about the xb ,
where R is the Rotation matrix of the B-frame relative to the yb , zb axes respectively.
E-frame. The expression of R is The total potential energy of the system is defined as
⎡ ⎤ U = mgz. (8)
cψcθ cψsθsφ − sψcφ cψsθcφ + sψsφ
R = ⎣ sψcθ sψsθsφ + cψcφ sψsθcφ − sφcψ ⎦ , (1) Now with the potential and kinetic energy expressions, we find
−sθ cθsφ cθcφ out the Lagrangian L = Ttrans + Trot − U for the system.
where cθ and sθ are the abbreviated forms for cos θ and sin θ Expression for the Lagrangian using (6), (7) and (8) is
respectively.
1 1 1
L = mẋ2 + mẏ 2 + mż 2
B. Body forces and torques 2 2 2
1 1
The forces acting on the Tilt-rotor include the forward thrust + Ixx (φ̇ − ψ̇sθ) + Iyy (θ̇cφ + ψ̇cθsφ)2
2
along the positive xb axis (Fxb ) and the vertical thrust acting 2 2
1
along the positive zb axis (Fzb ). They are expressed as + Izz (ψ̇cθcφ − θ̇sφ)2 − mgz. (9)
2
Fxb = fR sin(αR ) + fL sin(αL ), (2) The Euler-Lagrange equation describing the model is
Fzb = fR cos(αR ) + fL cos(αL ), (3) d δL δL
− = F, (10)
where fR , fL are the thrust exerted by the right and left rotors dt δ q̇ δq
and αR and αL are their respective tilt angles with the positive where F = [Fe , τ ]T is a column vector consisting of the E-
zb axis. frame forces Fe = [Fx , Fy , Fz ]T and torques τ . Fx , Fy , Fz
10 0
19.5 −0.5
ẍ = ζ1 = −Kpx (x − xd ) − Kvx (ẋ − x˙d ) + x¨d , 0 10
Time (s)
20 0 10
Time (s)
20
z̈ 30 0
φ̈ = γ1 = −Kpφ (φ − φd ) − Kvφ (φ̇ − φ˙d ) + φ¨d , 29.5 −0.5
θ̈ = γ2 = −Kpθ (θ − θd ) − Kvθ (θ̇ − θ˙d ) + θ¨d , 0 10
Time (s)
20 0 10
Time (s)
20
φ (deg)
0 0 3.5 3.5
fR (Nm)
fL (Nm)
−50 −100 3 3
0 10 20 0 10 20
Time (s) Time (s) 2.5 2.5
dθ/dt (deg/s)
50 100
2 2
θ (deg)
0 10 20 0 10 20
0 0 Time (s) Time (s)
−50 −100 10 20
0 10 20 0 10 20
Time (s) Time (s) 10
αR (deg)
αL (deg)
0
dψ/dt (deg/s)
100 50
ψ (deg)
0
50 0 −10
−10
0 −50
0 10 20 0 10 20 −20 −20
Time (s) Time (s) 0 10 20 0 10 20
Time (s) Time (s)
Figure 3. Orientation and angular rate of the Tilt-rotor UAV while hovering Figure 5. Motor thrusts and tilt angles of the Tilt-rotor UAV for stabilizing
at the desired position. it at the desired hovering position.
Fzb (N)
0 6 Future work with this UAV may involve modifying the back-
−2 4 stepping algorithm so as to avoid the occurrence of sin−1 in
0 10 20 0 10 20 the above expressions. A logical extension of our work would
Time (s) Time (s)
0.05 0.1 be to develop an algorithm to track discontinuous trajectories.
τφ (Nm)
τθ (Nm)
0 0
VI. ACKNOWLEDGMENT
−0.05 −0.1 We are grateful to Tony White for help with the text.
0 10 20 0 10 20
Time (s) Time (s)
0.1 R EFERENCES
τ (Nm)
http://www.bellhelicopter.com/en/aircraft/military/bellEagleEye.cfm.
−0.1 [2] Boeing, V-22 Osprey, (2008, September 13). Available:
0 10 20
Time (s) http://www.Boeing.com/rotor-craft/military/v22/index.htm.
[3] J. J. Dickerson, D. Miles, O. Cifdaloz, Wells, V. L. Rodriguez, “Ro-
bust LPV H Gain-Scheduled Hover-to-Cruise Conversion for a Tilt-
Wing Rotor-craft in the Presence of CG Variations”, American Control
Figure 4. Forces and torques acting on the Tilt-rotor UAV while hovering Conference, 2007.
at the desired position. [4] F. Kendoul, I. Fantoni, R. Lozano, “Modeling and control of a small
autonomous aircraft having two tilting rotors”, 44th IEEE Conference on
Decision and Control, and the European Control Conference 2005.
The variations in fL , fR , αL and αR as calculated from [5] A. A. Mian, W. Daobo, “Non-linear Flight Control Strategy for an
the previous variations in Fxb , Fzb , τφ , τθ and τψ as obtained Underactuated Quad-rotor Aerial Robot”, IEEE International Conference
from equations (47) and (48) are shown in Figure 5. on Networking, Sensing and Control, 2008.
[6] R. M. Murray, Z. Li, S. S. Sastry, A Mathematical Introduction to Robotic
Manipulation, CRC Press 1994.
V. C ONTRIBUTIONS [7] K. T. Oner, E. Cetinsoy, M. Unel, M. F. Aksit, I. Kandemir, K. Gulez,
“Dynamic model and control of a new quad-rotor unmanned aerial vehicle
In this paper, we developed a detailed model of a Tilt-rotor with tilt-wing mechanism”, World Academy of Science, Engineering and
UAV using the Euler-Lagrange approach. A back-stepping Technology, 45, 2008.
based PD control is implemented to stabilize the system.
Simulations show that the rotor-craft was able to recover its