Small Scale Helicopter Analysis and Controller Design for

Non-Aggressive Flights
Carlos Castillo, Wendy Alvis, Mauricio Castillo-Effen, Kimon Valavanis, Wilfrido Moreno
Center for Robot Assisted Search and Rescue
Department of CSE and EE
University of South Florida
Tampa FL 33620

In the design of controllers for small helicopters, the

Abstract – A simplified analysis is presented for controller
controlling variables are the throttle, the collective pitch,
design of small scale helicopters for non-aggressive flight
the antitorque pedals, and the cyclic pitch. The throttle
scenarios. A MATLAB based detailed approach was
control is used to maintain the r.p.m. of the main rotor.
followed to analyze the mathematical characteristics of the
The collective pitch control is used to control the
system, and to aid the design process of PID/PD and fuzzy
magnitude of the main rotor thrust. The cyclic pitch
logic controllers for the yaw, the pitch, the roll (Euler
control produces the lateral and the longitudinal tilting of
angles) and the height variables for hover and slow flight.
the main rotor disc and it is represented in the model by
Both controllers have demonstrated promising results and
two independent inputs, the lateral and longitudinal cyclic
they are simple and mathematically sound.
control inputs. These inputs control the roll and the pitch
Keywords: small scale helicopters, PID, fuzzy logic, angles that determine the forces created by the main rotor
MATLAB based design. in the lateral and longitudinal direction.

1 Nomenclature It was found in the literature review, that frequency

based approach to controller design was used as a common
φ Euler roll angle
method for comparing different controller designs.
θ Euler pitch angle
However, this approach is not straightforward due to the
ψ Euler yaw angle
complexity of the helicopter’s transfer function based
p Angular rate in the x axis in the body-fixed frame
models. Classical methodologies based on linear controller
q Angular rate in the y axis in the body-fixed frame
design such as the well known “PID” controller are still
r Angular rate in the z axis in the body-fixed frame
prevalent [1]. In addition, fuzzy logic based control offers
u Longitudinal speed in body-fixed frame
an alternative to design controllers for nonlinear systems,
v Lateral speed in the body-fixed frame
including helicopters [4], [11].
δlat Cyclic lateral control input
δlon Cyclic longitudinal control input In this paper, a MATLAB based simulation approach
δped Pedal control input was followed to analyze the mathematical characteristics of
δcol Collective control input the helicopter system model and to design classical PID/PD
x State variable matrix and fuzzy logic based controllers. Considering “a zero
y State space output vector angle approximation” of the Euler angles and utilizing
A State space system matrix SISO techniques, a simpler approximate stability analysis
B State space input matrix was performed on the system. Controllers were designed
C State space output matrix in the MATLAB environment for the yaw, the pitch, the
IAE Integral of the Absolute Error roll and the height variables for hover and slow flight.
I Identity matrix
Since classical controllers offer the advantage of
2 Introduction simpler stability analysis and in order to gain insight
information related to simplified model characteristics, a
Helicopters are nonlinear coupled multiple-input, Simulink based helicopter model was used to compare the
multiple-output (MIMO) systems with unique different outputs. Comparisons were performed among the
characteristics that makes the controller design a different designs, classical and fuzzy based controllers,
challenging problem. Conversely, model linearization including simulation results in which the IAE criterion was
results in controller design simplification. When dealing used for parameter optimization.
with non-aggressive flight scenarios, linearized models
suitable for hovering, up-down, forward-backward flights
offer a technically sound alternative for controller design,
testing and implementation.
3 Helicopter System Modeling φsp δlon
θ sp δlat

A helicopter has six degrees of freedom, is nonlinear u

v
ψ sp
h sp=-Zesp
δped
δcol

and input-output coupled. Linearization is essential to TRANSFORM

FROM BODY

derive simplified working models, considering the inherent OUTER

TRANSLATIONAL
CONTROLLER
INNER
ATTITUDE
CONTROLLER
HELICOPTER
INCLUDING
FRAME
REFERENCE
TO INERTIAL

instability under hover and slow flight conditions. Small u sp

v sp
ACTUATOR
DYNAMICS
FRAME
REFERENCE

scale helicopters have very similar characteristics; φ

θ
ψ
therefore, it is beneficial to derive a generalized linear h

model for such types of helicopters, common to all ROLL, PITCH, YAW ANGLES AND HEIGHT

controller designs. Mettler [1] has used the Comprehensive

Identification from the FrEquency Responses (CIFER) Figure 1: Block Diagram of the Control System
package, along with system analysis, to develop a
linearized thirteenth order state matrix model of the X-Cell This paper focuses only on the inner loop controller.
60 helicopter with parameters that vary over different flight The inner loop controller is composed of four SISO
envelopes. The general form used for the state matrix controllers: the altitude, the yaw, the pitch and the roll
model is defined below as: controllers. The outer loop controller is composed of two
SISO controllers: the u and the v body-fixed frame velocity
x& = Ax + Bu controllers. The set points for the pitch and roll controllers
(1)
y = Cx are provided by the outputs of the u and the v controllers.
The set points for the u velocity, the v velocity, the yaw and
Where the A matrix represents acting forces on the the altitude depend on the trajectory to track. The set points
helicopter’s center of gravity, the B matrix represents used for the trajectory following demonstration presented
different physical effects on the inputs under different are supposed to be known for the purpose of this paper and
flight conditions [1] and the C matrix relates the measured their calculation is not investigated here.
helicopter system outputs to the body reference output
variables. The C matrix depends on the physical sensor 5 System Analysis
system being used. The u-vector refers to the set of control
inputs, i.e. the height, the lateral and the longitudinal cyclic 5.1 Transfer Functions Based on a Zero
(controls the sideways and forward-backward motion by
adjusting the pitch and roll angles of the helicopter), and Angle Approximation
the pedal (controls the heading, or the yaw angle) of the The Euler equations that were used to transform the
helicopter. Outputs of interest for controller design are the body reference frame to the inertial reference frame
yaw, the pitch, the roll angles and the height, all of which introduced nonlinearities into the model. These equations
can be actually measured. are given in equation (2) for roll, pitch, yaw and z. The
height output of the system is defined as -z.
4 The Control Scheme
Helicopters are maneuvered by controlling attitude φ= ∫ ( p + q sin θ tan θ + r cos φ tan θ )dt
angles (roll, pitch, and yaw) and altitude. Newton's laws of θ = ∫ ( q cos φ − r sin φ ) dt
motion govern the behavior of the helicopter in flight. (2)
Solving the Newton equations of motions in the body-fixed ψ = ∫ ( q(sin φ / cos θ) + r(cos φ / cos θ) ) dt
frame makes it simpler to manipulate them compared to the
z = ∫ ( u sin θ + v cos θ sin φ + w cos θ cos φ )
inertial frame. Hence, the most common approach to
implement helicopter control is the projection of the inertial
velocity to the body-fixed frame. The state matrix form of the system was modified
slightly to include the Euler equations. A block diagram
To follow trajectories in the inertial frame it is form of the modified state space model is given in Figure 2.
necessary to implement controllers for the u and v body-
fixed frame velocities, the attitude angles and the inertial linearized helicopter model

altitude. It is apparent that at least two controllers will be

necessary to track a trajectory in the inertial frame. The u(t) x& x y& y

most common control scheme for a small helicopter

B ∫ C ∫
follows a two cascaded controller design, as shown in
Figure 1, with an inner and an outer loop controller. A

Figure 2: Block diagram of the modified state space model

 The transformation between the two reference frames equations and the transfer functions calculated using the
defines the C matrix followed by an additional integration, zero angle approximations. These step inputs were applied
see equation (3). Equation(4) can then be used to calculate one at time to allow for the transfer functions representing
the transfer functions from the time domain state matrix. the coupling effects to be verified as well. It was observed
that at some time instant after the step input was applied,
x& = Ax + Bu the pitch and roll angles increased, causing the model to
(3)
∫
y = (Cx )dt vary slightly from the small angle approximation transfer
functions. This difference did not occur until the roll or
pitch increased to approximately 2.5o. Therefore, keeping
Y(s) 1 the magnitude of pitch and roll angles less than 2.5o is a
T(s) = = C(sI − A) −1 (4) reasonable constraint for non-aggressive flight scenarios.
U(s) s
More detailed results of this comparison can be found in
[15].
The American Design Standard (ADS-33 [1]) states
that the hover speed range for rotorcraft is from 0 to 7
m/sec. The model used reaches a steady-state inertial 5.2 Diagonal Dominance
forward speed of 7.4 m/sec when the pitch angle is -1o. In order for a MIMO system to be broken into several
The standard rate for a turn is three degrees per second SISO systems for the purpose of using classical control
[10]. Based on the airspeed of 7.4m/s, and the rule of design methods, the amount of coupling within the system
thumb for a standard turn given in [10], the largest roll must be considered. Diagonal dominance allows the
angle that could be expected under hover flight conditions designer to determine if the system is tightly coupled.
is approximately 2o. For non-aggressive flight, it is One such way is to place the outputs and the control inputs
reasonable to require that the aircraft stay within a range along the diagonal of the open loop transfer function matrix
close to this hover limit. Considering these facts, it is and observe the effect of the other terms in the same row.
apparent that, for hover and slow speed flights, the Euler Diagonal dominance, φc(jw), is defined in equation (6).
angles will be small. If the angles are assumed to be zero, Zrc(s) represents the off diagonal elements of the
then several of the terms in the C matrix become zero and corresponding row, while Zcc(s) represents the diagonal
the equation is simplified and linearized as well. It has element of the row.
been concluded that this approximation is reasonable
considering the speed range in which the helicopter will be n
flying. This zero angle approximation is used to simplify
the calculation of the matrix of transfer functions, equation ∑z
r =1
rc ( jω)

(5). The cross-coupling between the outputs is accounted r ≠c

φc ( jω) = (6)
by terms present in the off-diagonal elements of the matrix. z cc ( jω)

⎡ φ φ φ φ ⎤ In order for SISO design methodologies to be used

⎢δ (s) (s) (s) (s) ⎥
δlon δped δcol successfully, the diagonal dominance must be less than one
⎢ lat ⎥
⎢ θ (s)
θ
(s)
θ
(s)
θ
(s) ⎥
[2]. This measure depends on frequency and it is
⎢δ δlon δped δcol ⎥ calculated across the bandwidth of the system. The transfer
T(s) = ⎢ lat ⎥ (5) functions were modified to include the actuator dynamics
⎢ ψ (s)
ψ
(s)
ψ
(s)
ψ
(s) ⎥ before calculating the diagonal dominance. All of the
⎢ δlat δlon δped δcol ⎥
⎢ z ⎥ diagonal elements met the requirement except for the
z z z height output variable. The maximum diagonal dominance
⎢ (s) (s) (s) (s) ⎥
⎢⎣ δlat δlon δped δcol ⎥⎦ for the height/collective was around 1.2. This is to be
expected due to the fact that when the cyclics are used to
MATLAB was used to calculate each one of the open create forward or sideways motion, this will also cause a
loop transfer functions. These transfer equations did not change in the height variable. However, the value is not far
include the actuator dynamics. The open loop dynamics above the requirement and, although using a single loop for
are defined in the transfer function form, and therefore, can this controller without any decoupling may decrease the
be multiplied by the actuator transfer function, Ga(s), in performance slightly, it should still give an acceptable
order to account for these effects. After calculating the response.
open loop transfer functions, the actuator effects were
included in both the model and all the system analysis. 5.3 Stability Analysis
It is possible to derive the closed loop transfer
Each of the open loop transfer functions, including the
functions using Mason’s rule based on the detailed flow
coupling terms, were calculated based on the small angle
diagram. Because of the complexity of the flow graph, this
approximations and built in Simulink. The same step inputs
calculation was extremely involved. The standard equation
were applied to both the state space model with the Euler
for deriving the closed loop transfer function for a SISO
system is given in equation(7). Gc(s) is the controller Bode Plot for yaw

transfer function; Ga(s) is the actuator function and 50

OUTPUT(s)/δ is the open loop transfer function for the 0

Magnitude (dB)
output of interest. Because the system was determined to be -50

diagonally dominant, deriving the transfer functions using -100

the standard SISO calculation, equation(7), was considered -150

for stability analysis. 0

MIMO Transfer Function
SISO Transfer Function
-90

Phase (deg)
OUTPUT
G c (s)G a (s)
(s) -180

T(s) = δ (7) -270

OUTPUT 10
-1 0
10 10
1
10
2
10
3 4
10

1 + G c (s)G a (s) (s) Frequency (rad/sec)

δ
Figure 5: Bode plot for yaw frequency response
In order to verify the above approach, the closed loop
transfer functions were also derived using both equation(7) While the ability to simplify the calculations of the
and Mason’s rule. The results were compared for each of transfer functions using equation(7) may not always be the
the four closed loop transfer functions. The transfer case for all controls systems, it has been found to be
functions for height were found to be mathematically sufficient for determining the phase and gain margin with
identical using both methods. This was not the case for the this particular model. Further details can be found in [15].
roll, the pitch and the yaw transfer functions. It was found
that there were differences in the transfer functions for the
yaw, the height and the pitch. A bode plot was used to 6 Classical PID/PD Control Design
compare the frequency response of the transfer function Classical PID controllers are simple to tune if the
derived with Mason’s rule and the transfer function derived dynamics of the process to control can be modeled using a
using the simplification for roll, pitch and yaw. The first or a second order transfer function. When this is not
frequency response of the systems matched extremely well. the case, as with the helicopters, it is possible to tune their
Figure 3, Figure 4 and Figure 5 displays the bode plot for parameters manually. However if a model of the controlled
the pitch, roll and yaw transfer functions respectively. system is available then a more practical approach is to use
an optimization algorithm that automatically tests different
Bode Plot for Pitch
parameter values until it finds the set of parameters that
50

0
gives the best possible value for a certain performance
function (objective function or fitness function). In this
Magnitude (dB)

-50

-100 case, a series of tests were implemented to obtain the

-150

-200
tuning parameters of PID controllers. As it was mentioned
-250
above, the diagonal dominance of the system was less than
90

0
MIMO Transfer Function one except for the height/collective which was close to one.
SISO Transfer Function
Based on these observations, it was decided that the
Phase (deg)

-90

-180 simplicity of SISO design outweighed any degradation of

-270
the controllers' performance in non-aggressive scenarios.
-360
-2 -1 0 1 2 3 4
10 10 10 10 10 10 10
Frequency (rad/sec)
The optimization algorithm was implemented using the
MIMO model. To obtain the parameters for the altitude,
Figure 3: Bode plot for pitch frequency response the yaw, the pitch and the roll controllers, the objective
function used for the optimization algorithm was to
minimize the Integral of the Absolute Error (IAE) [14].
Bode Plot for Roll
50

0
The IAE is a performance criterion that considers the
Magnitude (dB)

-50
difference between the set point and the output that exists
when a system is excited by a step input. The Optimization
-100

-150

-200
Toolbox of MATLAB 7.0 was used to obtain the
90
MIMO Transf er Function controller's gain for the PID/PD controllers.
0 SISO Transfer Function
Phase (deg)

-90

-180 Figure 6 shows the helicopter system’s block diagram used

-270
to obtain the PID parameters for the set of controllers
composing the inner loop controller.
-360
-2 -1 0 1 2 3 4
10 10 10 10 10 10 10
Frequency (rad/sec)

Figure 4: Bode plot for roll frequency response

 determines instability for a MIMO system that is behaving
stably, indicates a phase margin on the MIMO system that
Lateral
HELICOPTER
MODEL
TRANFORM
is very close to zero. Using the SISO system as the
reference point for stability will ensure that the controllers
Roll angle INCLUDING
set point Roll angle cyclic FROM
ACTUATOR
controller MODELLING BODY FRAME

Longitudinal
REFERENCE
TO
are stable; therefore the controller gains were reduced by a
factor of 0.3 to allow for a reasonable gain margin of at
Pitch angle
set point Pitch angle cyclic INERTIAL
controller REFERENCE
least 40o. The reduction of the controller gains did not
degrade the controller response. The IAE for the altitude,
Yaw angle
set point Yaw angle Pedal
controller
both before and after the modifications, was 5.426. The
IAE for the yaw was 0.8223. The final gains and phase
Altitude
set point Altitude Collective
controller
margins for the altitude and yaw controllers are shown in
Table 2.

As with the height controller, the roll and pitch Bode

plots derived from the SISO approximation indicated that
the system was unstable with the parameters gains obtained
Figure 6: Block Diagram Used with the Optimization by the optimization process. Each of the roll controller
Algorithm for the Inner Controller gains were reduced by a factor of 0.55 and the pitch
controller gains were each reduced by a factor of 0.8 in
The procedure used was to apply a step function at the order to insure stability. The IAE value for pitch was
set point of the controller under optimization and keep the raised slightly after the modification, from 8.383 to 10.07.
set point of the other controllers in an appropriate steady- The roll IAE was increased from 7.283 to 11.95; however,
state value (zero degrees for angles and 2 meters for the increase was a reasonable trade off for ensuring
height). The optimization algorithm was applied stability. The final gains and phase margins for the pitch
sequentially in order to get the optimal parameters for the and roll controllers are shown in Table 2.
controllers.
Table 1: Controller Values
6.1 Optimization Results for the PID/PD
Controllers KP KI KD

Table 1 presents the final inner controller values obtained PITCH 0.605 1.925 0.165
for the roll, the pitch, the yaw and the altitude controllers.
It is important to notice that cross-coupling between the 0.96 2.8 0.24
ROLL
inputs of the helicopter was neglected.
HEIGHT 1.2526 0 0.3393
For both, the height and the yaw controllers, the
optimization algorithm produced extremely small values
for the integral constant in the PID controller, KI. For this YAW 28.2370 0 0.9969
reason, it was decided to use PD controllers instead for the
height and the yaw controllers. The optimization routine
was performed a second time to determine the values of KD
and KP for each of the new PD controllers. Table 2: Controller Analysis
PHASE GAIN
Equation(7) was used to calculate the SISO transfer MARGIN MARGIN
functions for the closed loop yaw, altitude, pitch and roll. (degrees) (dB)
The Bode plots for each of the closed loop transfer
functions were obtained using MATLAB’s tools. The yaw
PITCH 51.5 5.86
was found to be stable using the gains obtained via
optimization. The height closed loop transfer was
determined to be unstable from the Bode plots using the ROLL 40.1 5.8
gains obtained from the optimization tool. In contrast with
the SISO stability analysis for the height, the MIMO HEIGHT 42.9 11.9
system behaves stably without any kind of oscillation for
overall system response time less than 1000 seconds. It is YAW ∞ 123
possible that the system is stable with the initial optimized
gains but is not determined to be stable because the
stability analysis neglects coupling. However it is
important to note that having a SISO approximation that
7 Fuzzy Logic Controllers Table 3: Optimized values of the scaling factor for the
fuzzy PD-like controllers
Fuzzy control systems have also been used to control
helicopters [3], [4], [5]. Following the analysis presented Se Sce Su
in the previous section, it was decided to use the equivalent
type of fuzzy PID/PD-like controllers [7] for the same HEIGHT 0.5704 0.3425 1.1306
controlled variables. Mann, Hu and Gosine [6] present the
three types of fuzzy structural elements needed to design a YAW 3.0386 0.6509 4.7436
fuzzy PID-like controller. Figure 7 shows the typical block
diagram of a fuzzy PD-like controller [8] and Figure 8
shows the typical block diagram of a fuzzy PID-like
controller where e(t) is the error defined as the difference
between the desired signal value, set point, and the real
value of the controlled variable; ∆e(t) is change of the Table 4: Optimized values of the scaling factors of the
error. fuzzy PID-like controller
Se Sce S1 S2
e( t ) ê
Se Fuzzy uˆPD uPD PITCH 0.6859 0.3505 3.5181 0.6159
PD Su
d
dt
Sce Ɛ rule base ROLL 0.8962 0.2546 3.5576 0.5923

Figure 7: Continuous-time structure of the fuzzy two-term

PD-like controller
8 Results
e(t ) ê An outer loop for the helicopter control system was
Se Fuzzy
•
uˆ PI uPI
PI S1 1 designed and simulated using techniques similar to those
Sce Ɛ
S
d rule base presented in this paper in order to allow for several
dt uPID
+ specified trajectories to be flown [15]. These trajectories
+
were selected in order to evaluate the inner loop response
Fuzzy
over several different sequences of inputs. The most
uˆ PD uPD
PD S2 complex of these trajectories was the double loop. Both the
rule base fuzzy controllers and the PID/PD controllers handled this
flight trajectory with minimal error, Figure 9.
Figure 8: Continuous-time structure of the fuzzy two-term
PID-like Controller
120

Se is the scaling factor for the error, e(t). Sce is the 100

scaling factor for the change of the error, ∆e(t). S1 and S 2

80
ze (m)

60 Desired Trajectory

are the output scaling factors of the PI-like and PD-like

PID trajectory
Fuzzy trajectory

40

controller which constitute the fuzzy PID-like controller. 20

0
-60

-40

7.1 Modifications of the Inputs and the -20

0
100
150

20 50

Outputs
0
40 -50
-100
60 -150

ye (m)
xe (m)

The adjustment of the input and output gains or scaling

factor was selected as the tuning approach for the fuzzy PD-
Figure 9: Double Loop Trajectory
like and PID-like controllers implemented in this study. An
optimization algorithm based on the IAE criterion was used 8.1 Comparison of controllers
as the objective function to get the optimal values for the
gains. The procedure followed was equivalent to the The optimization processes were based on the
procedure described in the optimization section 6. Table 3 minimization of the IAE criterion. Different values were
shows the scaling factors obtained for the fuzzy PD-like obtained for the classical PID/PD controllers and for the
controllers. Table 4 shows the scaling factors obtained for fuzzy PID/PD-like controllers. Table 5 shows a comparison
the fuzzy PID-like controllers. of the IAE for the classical and the fuzzy controllers.
 Table 5: IAE values obtained during optimization 9
Pitch Response Comparison
Pitch Set Point
Pitch Response with Fuzzy PID-like Controller

Classical Fuzzy
8 Pitch Response with PID Controller

PITCH 10.07 5.053

Pitch angle (deg)

5

ROLL 11.95 3.826 3

HEIGHT 5.426 9.863 1

YAW 0.8223 0.7994

-1
0 5 10 15 20 25
time (sec)

Figure 12: Comparison of the responses of the pitch

Roll Response
Figure 10, Figure 11, Figure 12, and Figure 13 shows the 9
Optimized Fuzzy PID-like Controller

responses of the altitude, yaw, pitch and roll variables

Roll Set Point
Roll Response with Fuzzy PID-like Controller
8 Roll Response with PID Controller

respectively. 7

Altitude Response Comparison

Roll angle (deg)

5
16
Altitude Set Point
Altitude Response with Fuzzy PD-likeController 4
Altitude Response with PD Controller
14
3

12 2

1
10
Altitude (m)

8
-1
0 5 10 15 20 25
time (sec)
6

Figure 13: Comparison of the responses of the roll

2
0 1 2 3 4 5 6 7 8
time (sec)

9 Conclusions
Figure 10: Comparison of the responses of the altitude
The analysis of the small scale helicopter model was
greatly simplified by the use of both zero angle and SISO
12
Yaw Response Comparison
approximations. Comparison between the simplified
transfer functions and the more complex model
Yaw Set Point
Yaw Response with Fuzzy PD-like Controller
Yaw Response with PD Conroller

demonstrated that these simplifications were adequate for

10

8
non-aggressive flight conditions.
Yaw angle (deg)

The inner loop controllers were designed using the

techniques presented in this paper and successfully
4

2 implemented with the Simulink model. In addition to

0
demonstrating desirable step responses, the controllers also
allowed for the successful following of complex
-2
0 1 2 3 4
time (sec)
5 6 7 8
trajectories.

The methods of analysis and design presented

Figure 11: Comparison of the responses of the yaw significantly reduce the work and time required for the
implementation of non-aggressive flight controllers in
small-scale helicopters. In addition, the controllers
designed require very little computational overhead to
implement in the physical system. This is an advantage for
small portable helicopters with very little load capacity.

References
[1] Bernard Mettler, Identification Modeling and
Characterization of Miniature Rotorcraft, Kluwer
Academic Publishers, 2003.
[2] Neil Munro, "Symbolic Methods in Control System [15] CRASAR VTOL technical report No. 01
Analysis and Design", The Institution of Electrical
Engineers, Londo, 1999.

[3] M. Sugeno, “Development of an Intelligent

Unmanned Helicopter”, Chapter 2 of ‘Fuzzy Modeling and
Control: Selected Works of M. Sugeno’, Hung T. Nguyen
and Nadipuram R. Prasad (eds), CRC Press, Boca Raton,
1999.

[4] B. Kadmiry and D. Driankov, “Fuzzy Control of an

Autonomous Helicopter”, IFSA World Congress and 20th
NAFIS International Conference, 2001, Joint 9th, pp. 2797-
2802, vol. 5, Vancouver, BC Canada, July 25-28,2001.

[5] D. Driankov, H. Hellendoorn, and M. Reinfrank, “An

Introduction to Fuzzy Control”, Springer-Verlag Berlin,
Second edition, 1996.

[6] George K. I. Mann, Bao-Gang Hu and Raumond G.

Gosine, “Analysis of Direct Action Fuzzy PID Controller
Structures”, IEEE Trans. Syste., Man, and Cyb. vol. 29,
No 3, June, 1999.

[7] Bao-Gang Hu, G. K. I. Mann, and R. G. Gosine, “A

Systematic Study of Fuzzy PID Controllers-Function-
Based Evaluation Approach”, IEEE Trans. Fuzzy Syst., vol.
9, No. 5, October 2001.

[8] Han-Xiong Li and H. B. Gatland, “Conventional

Fuzzy Control and Its Enhancement”, IEEE Trans. Syst.,
Man, and Cyb. Vol.26, NO. 5, October 1996.

[9] Wagtendonk, Walter J., "Principles of Helicopter

Flight", Aviation Supplies & Academics, Inc., First
Edition, 1996.

[10] U.S. Dept. of Transportation, Rotorcraft Flying

Handbook, FAA-H-8083-21.

[11] Tito G. B. Amaral, “ Automatic Helicopter Control

Motion using Fuzzy Logic”, In Proc. of the 2001 IEEE
International Fuzzy Systems Conference, pages 860-
863,December, 2-5, 2001.

[12] David H. Shim, H. Jin Kim, S. Sastry, “Control

System Design for Rotorcraft Unmanned Aerial Vehicles
using Time-domain System Identification”, In Proc. of the
IEEE international Conference on Control Applications,
pages 808-813, September 25-27, 2000.

[13] Chue Chien Lee, “Fuzzy Logic in Control Systems:

Fuzzy Logic Controller-Part I”, IEEE Trans. Syst., Man,
and Cyb.

[14] Smith, Carlos and Corripio, Armando., “Principles

and Practice of Automatic Process Control”, Second
Edition, pp. 322. Wiley, 1997.