Symbolic Pole Assignment of 2DoF Conventional

Helicopter

TarÕk Veli Mumcu Kutluhan Kürúat ÇandÕr

Department of Electrical and Electronics Engineering Department of Control and Automation Engineering
Istanbul University YÕldÕz Technical University
Istanbul, Turkey Istanbul, Turkey
[email protected] [email protected]

ølker Üsto÷lu
Department of Control and Automation Engineering
YÕldÕz Technical University
Istanbul, Turkey
[email protected]

Abstract— Unmanned air vehicles and their control is a method is based on Linear Quadratic Gaussian theory and
popular research field. The next decade will determine the furthermore an optimal state estimator in terms of Kalman
control perfection of this vehicles and their role in our daily lives filter is developed and a loop transfer recovery procedure is
will increase. The major advantages of the unmanned rotor
systems can be given as their ability to hang on a particular
also proposed for the recovery of full state feedback
altitude and their manoeuvre agility to move and escape. properties. A H’ controller for MIMO systems developed in
Whereas there are many studies of control of drones, quadrotors [2] to develop a robust controller for an unmanned air vehicle
or conventional helicopter as well as full rank state feedback for attitude regulation. A recursive approach is used in [3] to
method, the studies considering combination of these methods obtain the parameters of the controller by the combination of
are rare. Even in symbolic case it is either not detailed so that it sliding mode, back-stepping control for altitude control. The
can easily be followed and applied or missing. This study aims to
assign the poles of two degree of freedom conventional helicopter.
parameter definitions and helicopter modelling in [4] obtains
The pole placement method used here is full rank state feedback. the constants by recursive Newton Euler approach as a result
Thus, the helicopter model for multi input multi output systems of ordinary least solution of over determined problem.
is explained, the stages of the control method are described and Another H’ control problem is studied in [5] by rotor state
they are simulated with in Mathematica. The outputs and feedback control to delay with model uncertainties and
stability analysis illustrates the effectiveness of the study. emphasize the applicability of the control on the real model.
Index Terms— State feedback, symbolic, full rank, 2 DoF The combination of PD for position control and a fuzzy
helicopter. adaptive PD control for attitude control can be found in [6] to
develop a controller for unmanned air vehicle as an under
I. INTRODUCTION actuated system as. An efficient model predictive controller is
The autonomous control of unmanned air vehicles such as developed in [7] for the control of a small unmanned
drones and helicopters is a popular research area of today and quadrotor helicopter. Instead of dealing with large calculations
will even be more within next decade. Recently their this study aims to reduce the computational requirement
application fields and purposes are multiplied, varied and needed for conventional model predictive controller with the
differentiated from post disaster recovery to entertainment. help of fewer predictions. The algorithm performance is given
This numerous application fields and popularity brings new as the tracking of a conventional quadrotor. There are not so
challenges to these vehicles. many studies within literature concerning the pole assignment
In the literature it is very common to find papers on both problem of a conventional single or multi rotor copter.
single and multirotor helicopters at any size from micro to This paper aims to focus on a symbolic pole assignment
small or medium. The papers often focus on multi input multi problem for a two degree of freedom (2 DoF) helicopter
output control (MIMO) systems. Besides, designing system and investigates its behaviour under multi-loop PI +
controllers for MIMO systems is a research area where pole full-rank state feedback control method. Section II gives the
assignment problem and minimizing the coupling between nonlinear model of 2 DoF helicopter and at the end of this
different inputs-output pairings are of special importance. In section the linear state space model of the system is also
[1] a MIMO controller design procedure is given where the presented. Section III develops the state feedback gains

978-1-5386-1880-6/17/$31.00 ©2017 IEEE

symbolically based on the so-called full-rank pole assignment The generalized forces of the system are given as in Eq.6 -
method. Section IV discusses the relative gain array and Eq.7.
develops the multi-loop PI controller for the system under state
feedback. The paper ends with some concluding remarks. Q1 = τ p (Vm, p ,Vm, y ) − Bpθ
(6)
II. MATHEMATICAL MODELLING OF 2 DOF HELICOPTER Q2 = τ y (Vm, p ,Vm, y ) − Byψ
The simple free body diagram of the 2 DoF helicopter on τ p (Vm, p ,Vm, y ) = K ppVm, p + K pyVm, y
which the movement is considered only on yaw and pitch axes (7)
can be seen on Fig. 1. τ y (Vm, p ,Vm, y ) = K ypVm, p + K yyVm, y

The torque constants are obtained by using equations

given as in Eq.8.

Kt , y Kt , p
K pp = K f , p rp ; K yy = K f , y ry ; K py = ; K yp = (8)
Rm , y Rm , p

By using Eq.8 equations we reached the torque equations

given in Eq.9, Eq.10 and Eq.11.

D (q )q + C (q, q )q + g (q) = τ (9)

ª J + mheli lCM
2
0 º
Fig. 1. 2 DoF helicopter modelling as simple free body diagram [8]. D(q) = « eq , p » (10)
«¬ 0 J eq , y + m l 2
heli CM cos θ »¼
2

The centre of mass length is obtained by using motor,

propellers and shield weights as in Eq.1. ª 2
mheli lCM sin θ ׺
« Bp »
× cos θψ 
( mm , p + mshield ) rp + ( mm , y + mshield ) ry C (q, q ) = « » (11)
lCM = (1) « −2m l 2 sin θ × »
mm , p + mm , y + 2mshield « heli CM
By »
«¬× cos θψ »¼
The kinetic energy of the center of mass is acquired ª m gl cos θ º ª K ppVm, p + K pyVm, y º
by using velocity of the x, y , z directions as in Eq.2. g (q ) = « heli CM » ;τ = « » (12)
¬ 0 ¼ ¬ K ypVm, p + K yyVm, y ¼
1
Tt = mheli ((−lCMψ sinψ cos θ − lCM θ cosψ sin θ ) 2 Where D denotes mass matrix, C denotes
2 (2) Christoffel symbols and g denotes gravitational acceleration.
1
+ (−lCMψ cos θ + lCM θ sinψ sin θ ) 2 + lCM
2 2
θ cos θ 2 ) 2 The controller added feedback system of the 2DoF helicopter
can be seen on Fig. 2.
The linear time invariant mathematical model of 2 DoF
Lagrange-Euler method is used to model the
Helicopter is of the form x = Ax + Bu , y = Cx where A is
helicopter system is as in Eq.3.
given as in Eq.13 and B is given as in Eq.14.
L= K−P (3)
ª0 0 1 0 º
Where K denotes kinetic energy and P denotes « »
potential energy. The Lagrange-Euler equation is represented «0 0 0 1 »
« Bp »
in Eq.4. 0 −
A = «0 2
0 » (13)
« J eq , p + l CM m »
∂ ∂L ∂ « »
− L = Q1 «0 Bp »
∂t ∂q1 ∂q1 0 0 −
(4) « J eq , y + l C2M m »
∂ ∂L ∂ ¬ ¼
− L = Q2
∂t ∂q2 ∂q2

The states of the system are shown in Eq.5.

T
q = [ q1 q4 ] = ª¬θ ψ θ ψ º¼
T
q2 q3 (5)

978-1-5386-1880-6/17/$31.00 ©2017 IEEE

 ª 0 0 º gains which assign the poles of the system to desired locations
« » are calculated by the following formula Eq.19.
« 0 0 »
« K pp K py » 
K = Q −1 KT (19)
B = « J + l2 m 2 »
J eq , y + l C M m » (14)
« eq , p C M
« » where Q is a p × p nonsingular matrix, having nonzero rows
« K yp K yy »
« J + l2 m J eq , y + l C2M m » of B and K is the full-rank state-feedback matrix determined
¬ eq , p C M ¼
in the new basis such that A − BK   is block triangular or

with variable transforms Eq.15 diagonal with each block containing a part of the polynomial
with the roots as the desired poles [9]. Compared to the
α = J eq , p + l 2 m , β = J eq , y + l 2 m , methods based on Ackermann’s or Bass- Gura’s formulae this
CM CM
algorithm returns with lower gain values to be applied to the
γ = K py K yp − K pp K yy (15) system for states, in other words the norm of the state
feedback matrix is much lower. In this case K is a 2 × 4
Note that the outputs of the system are y1 = θ and y2 = ψ symbolic matrix as given as in Eq.20.
§ k11 k12 k13 k14 ·
( )
T
and the state vector is x = θ ψ θ ψ as given in Eq.5. K =¨ ¸ (20)
© k21 k22 k23 k24 ¹
III. POLE ASSIGNMENT OF 2 DOF CONVENTIONAL HELICOPTER
where the entries depend on system parameters and desired
Recall that the poles of the system can be assigned to any
pole locations as given in Eq.21 and Eq.22.
locations on the complex s-plane if the pair ( A, B) is
controllable, i.e. the so-called controllability matrix is of full p1 p2α K py K yp p3 p4α
column rank and the control law is given by a linear k11 = −
combination of the states which can be formulated as K pp K pp γ
u = Kx + v , with v being an auxiliary input to the system. The
state space representation of the system after static state
feedback is given as in Eq.16 and also can be seen in Fig.2. K py p3 p4α
k12 =
γ
x = ( A + BK ) x + Bv , y = Cx (16)
Bp Bp
(− p1 − p2 − )α K py K ypα (− p3 − p4 − )
The new transfer function Gˆ ( s ) of the plant will be as α β
k13 = −
given as in Eq.17. K pp K pp γ

Gˆ ( s ) = C ( sI − A − BK ) −1 B . (17) Bp
K pyα (− p3 − p4 − )
β
k14 = (21)
γ
K yp p3 p4α
k 21 =
γ

K pp p3 p4α
k22 = −
Fig. 2. Full state control feedback system of 2 DoF helicopter
γ
Bp
A simple test on controllability of 2 DoF Helicopter K ypα (− p3 − p4 − )
β
model reveals that the system is controllable. Let the desired k 23 =
pole locations be as given as in Eq.18. γ
Bp
σ ( A + BK ) = { p1 , p2 , p3 , p4 } (18) K ppα (− p3 − p4 − )
β
k24 = − (22)
γ
The system model is converted to Luenberger
controllable canonical form by applying transformation
A = TAT −1 and B = TB and resulting matrices will be in
special block-structured forms. The full-rank state feedback

978-1-5386-1880-6/17/$31.00 ©2017 IEEE

 IV. MULTI-LOOP PI CONTROL OF OVERALL SYSTEM Ky p Thrust torque constant acting on yaw axis 0.0219 N m / V
from pitch propeller
Using full rank state feedback method the poles are assigned Thrust torque constant acting on yaw axis
Kyy 0.0720 N m / V
as desired but note that the 2 DoF helicopter is a 2 × 2 system, from yaw propeller
so before closing the loop for system outputs the best input- Vm , p Voltage apply to pitch motor ±24V
output pairings must be established for a reasonable frequency Vm , y Voltage apply to yaw motor ±15V
band. Here, the relative gain array (RGA) method is used as
an operator to find the best pairings of the system [10]. The
operator RGA at zero frequency is defined as
RGA = Gˆ (0) D Gˆ −T (0) . It should be remembered that for non- With this parameter set the system has four poles located at
square systems pseudo inverse of Gˆ ( s ) is used [11]. Here, for {0, 0, − 9.24979, − 8.76343} so if the desired poles are chosen
2 DoF helicopter model desired pole location dependent RGA as {−12, − 10, − 9.25, − 8.76} the resulting state feedback
is calculated as in Eq.23. matrix is determined as:

ª 73 p3 p4 73 p3 p4 º § 51.227 − 3.278 5.446 − 0.374 ·

K =¨ ¸
«1 + 7127 p p −
7127 p1 p2 » © −10.556 98.333 − 1.205 11.221 ¹
RGA = « »
1 2
(23)
« 73 p3 p4 73 p3 p4 »
«− 1+ » Finally the PI controller for both input-output pairing is
¬ 7127 p1 p2 7127 p1 p2 ¼
1
designed as PI ( s) = 30(1 + ) . The system outputs for unit
0.1 s
where the parameters are taken from Table I. As can be seen step inputs are given in Fig. 4 and Fig. 5, respectively.
at zero frequency there is no interaction between different
control channels. The best input-output pairing is {1, 1} and
{2, 2}. Since the system is diagonally dominant there is no
prefilter design step required to make the system channels
decoupled. Fig. 3. represents how the system interconnection
should be such that the system behaviour is satisfactory.

Fig. 3. PI Control of overall system.

Fig. 4. θ (t ) vs. time (in seconds) for unit-step reference signal.

The 2 DoF helicopter parameters are given in the table below

[8].
TABLE I. 2 DOF HELICOPTER BODY PARAMETERS.
g Gravitational constant 9.81m / s 2
lCM Center of mass length along helicopter body 0.1855 ms
from pitch axis
m Total moving mass of the helicopter 1.3872 kg
J eq , p Total moment of inertia about pitch axis 0.0384kg m2
J eq , y Total moment of inertia about yaw axis 0.0431kg m2
Bp Viscous damping about pitch axis 0.8 N / V
By Viscous damping about yaw axis 0.318 N / V

Kpp Thrust torque constant acting on pitch axis 0.2041 N m / V

from pitch propeller Fig. 5. ψ (t ) vs. time (in seconds) for unit-step reference signal.
Kpy Thrust torque constant acting on pitch axis 0.0068 N m / V
from yaw propeller

978-1-5386-1880-6/17/$31.00 ©2017 IEEE

 V. CONCLUSIONS [4] S. Rajappa, A. Chriette, R. Chandra and W. Khalil, “Modelling
and Dynamic Identification of 3 DOF Quanser Helicopter,”
In this paper a symbolic full rank state feedback pole 2013 16th International Conference on Advanced Robotics
assignment problem is discussed for 2 DoF helicopter model (ICAR), Montevideo, 2013, pp. 1-6.
and state feedback gains are presented depending on system [5] S. Panza and M. Lovera, “Rotor state feedback in helicopter
parameters and desired pole locations. Relative gain array is flight control: Robustness and fault tolerance,” 2014 IEEE
also calculated in terms of desired pole locations. Conference on Control Applications (CCA), Juan Les Antibes,
pp. 451-456, 2014
[6] H. Gao, C. Liu, D. Guo and J. Liu, “Fuzzy adaptive PD control
REFERENCES for quadrotor helicopter,” 2015 IEEE International Conference
on Cyber Technology in Automation, Control, and Intelligent
[1] E. Barzanooni, K. Salahshoor and A. Khaki-Sedigh, “Attitude Systems (CYBER), Shenyang, 2015, pp. 281-286
flight control system design of UAV using LQGLTR [7] Abdolhosseini, M., Zhang, Y.M. & Rabbath, C.A. J Intell Robot
multivariable control with noise and disturbance,” 2015 3rd RSI Syst (2013) 70: 27.
International Conference on Robotics and Mechatronics [8] Quanser 2-DOF Helicopter Laboratory Manual, Quanser Inc.,
(ICROM), Tehran, 2015, pp. 188-193. Ontario, Canada, 2011.
[2] J. P. Ortiz, L. I. Minchala and M. J. Reinoso, “Nonlinear Robust [9] Patel, R. V. and N. Munro, Multivariable System Theory and
H-Infinity PID Controller for the Multivariable System Design, Oxford: Pergamon Press, 1982.
Quadrotor,” IEEE Latin America Transactions, vol. 14, no. 3, [10] Bristol, E. H., “On a new measure of interaction for
pp. 1176-1183, March 2016. multivariable process control”, IEEE Transactions on Automatic
[3] S. H. Dolatabadi and M. J. Yazdanpanah, “MIMO sliding mode Control, AC-11, no. 1, 1966, pp. 133-134.
and backstepping control for a quadrotor UAV,” 2015 23rd [11] Glad, T. and L. Ljung, Control Theory: Multivariable and
Iranian Conference on Electrical Engineering, Tehran, 2015, Nonlinear Methods, London and New York, NY: Taylor and
pp. 994-999. Francis, 2000, pp. 220-228.

978-1-5386-1880-6/17/$31.00 ©2017 IEEE