See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/270961027

Evolutionary Algorithms for Robust Cessna Citation X Flight Control

Conference Paper in SAE Technical Papers · September 2014

DOI: 10.4271/2014-01-2166

CITATIONS READS
20 249

4 authors:

Yamina Boughari Ruxandra Mihaela Botez

École de Technologie Supérieure École de Technologie Supérieure
10 PUBLICATIONS 57 CITATIONS 551 PUBLICATIONS 5,225 CITATIONS

SEE PROFILE SEE PROFILE

Georges Ghazi Florian Theel

École de Technologie Supérieure École de Technologie Supérieure
38 PUBLICATIONS 262 CITATIONS 8 PUBLICATIONS 60 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

criaq505 View project

CRIAQ MDO505 View project

All content following this page was uploaded by Ruxandra Mihaela Botez on 26 January 2015.

The user has requested enhancement of the downloaded file.

 Robust Flight Control of the Cessna Citation X using
Evolutionary algorithms

Yamina Boughari
ETS, Laboratory of Active Controls, Avionics and AeroServoElasticity LARCASE
1100 Notre Dame West, Montreal, Que., Canada, H3C-1K3

Ruxandra Mihaela Botez

ETS, Laboratory of Active Controls, Avionics and AeroServoElasticity LARCASE
1100 Notre Dame West, Montreal, Que., Canada, H3C-1K3

Georges Ghazi
ETS, Laboratory of Active Controls, Avionics and AeroServoElasticity LARCASE
Montreal, Que., Canada, H3C-1K3

Florian Theel
ETS, Laboratory of Active Controls, Avionics and AeroServoElasticity LARCASE
Montreal, Que., Canada, H3C-1K3

Abstract:

The main goal of this flight control system is to achieve good performance with acceptable flying quality
within the specified flight envelope while ensuring robustness for model variations, such as mass variation
due to the fuel burn.
The Cessna Citation X aircraft linear model is presented for different flight conditions to cover the aircraft’s
flight envelope, on which a robust controller is designed using the H-infinity method optimized by two
different heuristic algorithms. The optimal controller was used to achieve satisfactory dynamic
characteristics for the lateral stability control augmentation systems with respect to this aircraft’s flying
quality requirements. The weighting functions of the H-infinity method were optimised by using both genetic
and differential evolution algorithms. The evolutionary algorithms gave very good results. This is the first
time these algorithms have been used in this form to optimize H-infinity controllers on a business aircraft,
thereby respecting both flying quality requirements and robustness criteria as objective functions and
avoiding the use of other computationally complicated algorithms.

Keywords: flight control; H-infinity; optimal control; robust control.

Nomenclature

H_infinity = H-infinity

= aircraft positions

= state space parameter of the system

= roll angle and roll rate

 = pitch angle and pitch rate

= speeds along the Ox, Oy, Oz axes

= inertial moments of the body axis system

= total aircraft speed

= elevator, aileron and rudder deflections

= state space gain

= transfer function of the system

= sensitivity and robustness weighting functions

I. Introduction

Several modern control laws are used for the design of aircraft stability augmentation systems in the presence of
uncertainties, such as the LMI approach, to achieve a robust control design on an uncertain aircraft system [1], and
the H_2 control for gust load alleviation on flexible aircraft [2]. Another control law is the H_infinity theory,
proposed by Zames in 1983 [3], which has more recently gained in popularity as a way to ensure robustness in the
presence of uncertainties. H_infinity is a multiple input multi-output method to determine a feedback controller on
an uncertain system, and has been further developed for several control fields. Many applications of this control
method have been incorporated in the aeronautical domain, mostly on fighter jets, as in [4], where a scheduled
H_infinity controller was used on VSTOL longitudinal control, and in [5], where it was used on the lateral control
of an F-14.

The H_infinity control law is based on the selection of weighting functions, which is the key to the successful design
of a robust controller. There is no specific formula to determine these weighting functions. An H-infinity controller
design using the gain scheduling approach was successfully applied on a flexible aircraft, where the weighting
functions were not optimized and were only determined using engineering intuition[6]. To overcome the lack of
reference formulas, some guidelines to selecting weighting functions are given in [7, 8], however this procedure
may takes many iterations using a trial and error process, and the result may not satisfy all of the required
conditions (flying conditions in the case of aircraft). Tuning the weighting function until it meets the mandatory
requirements is then necessary. A lack of design experience on the part of the engineer can lead to a system with a
poor stability margin, even if it does give a satisfactory response.

Several weight optimisation methods are based on mathematical algorithms, in which trade-offs are arranged
between maximizing the stability margin and minimizing the H_infinity norm of the final closed-loop transfer
function [9]. These algorithms are often frequency-dependent optimisations, in which the iteration process takes up a
considerable amount of memory allocation. To overcome this frequency-dependent optimisation problem, a state
space weight optimisation was developed in [10]. However, that algorithm does not guarantee a global minimum
convergence, which can lead to a poor stability margin, which is especially a concern in a system operating in a
large envelope, such as aircraft.

The algorithms proposed here aim to optimize the Cessna Citation X aircraft‘s lateral flight control using the
H_infinity method Both Genetic and Differential Evolution search algorithms have been used to get the optimal
weighting functions; and K_infinity, which is also used in business aircraft stability and lateral control augmentation
system design, is used to achieve robust control performance.

This paper is organized as follows: The next section presents a brief description of the Cessna Citation X, the fastest
business aircraft in the world, and then describes the nonlinear and linear lateral aircraft models, followed by the
Cessna Citation X’s flying requirements. A brief presentation of the H_infinity theory is given in section three. The
weight-selection methods are described in the fourth section, both the differential evolution and the genetic
algorithm method, followed by their application to the H-infinity problem in section five. Our simulation and
results’ analysis are presented in section six, followed by our conclusions.

II. Cessna Citation X business aircraft

The Cessna Citation X is the fastest civil aircraft in the world, and operates at Mach number of 0.935. The
longitudinal and lateral motions that model the business aircraft are described, as well as its flight envelope,
followed by the flying qualities requirements.

The Cessna Citation X aircraft was selected for this work in part because the Aircraft Flight Research Simulator
level D is available in our laboratory. The availability of this flight simulator makes it possible to validate the
numerical results with real flight test data.

In order to analyze the stability of an aircraft, its model must first be identified. Model identification can be done by
using a combination of fuzzy logic and neural network methods to obtain a nonlinear aircraft model, for example,
for the military aircraft F/A –18 [11]. Recently, a new system identification for the business Cessna Citation X
aircraft was developed in [12]; this system was compared with a linear Cessna citation X model obtained from a
linearization method using aircraft simulator data.

A. Aircraft dynamics
The model of the Cessna Citation X aircraft’s longitudinal and lateral motions are represented in Figure 1.

Figure 1 Longitudinal and lateral models of the Cessna Citation X business aircraft

The motion of an aircraft is represented with a nonlinear model given in [13], as follows:

The rates of change of positions x, y and z are:

and the rates of change of angular positions q, p and r:

̇ ̇

̇ ̇

̇ ̇
 ̇

̇ ⁄

and the rates of change of the speeds are:

To design a controller for any aircraft, a linearization of the nonlinear model of the aircraft is required, for flight
conditions within the flight envelope given by the designer. Following the decoupling of the linearized aircraft
motion in two modes (longitudinal and lateral), the equations are represented in the form of the following state space
system:

This system is decomposed into two sub-systems representing the aircraft’s longitudinal and lateral motions. The
aircraft’s longitudinal motion dynamics are given by the state space equation, using the elevator as input:

̇ ̇ ̇ ̇

( ) ( )

where the state vector and control vector (t) are given by:

The aircraft’s lateral motion dynamics are given by the state space equation, using the aileron and the rudder as
inputs:

̇
 ⁄ ⁄ ⁄ ⁄ ⁄ ⁄

( ) ( )
where the state vector and control vector are given by:

The linear model of the Cessna Citation X is obtained for 36 flight conditions based on the Aircraft Flight Research
Simulator tests performed at the LARCASE laboratory [12]. The linearized model is further decomposed in LFR
models [14] using the bilinear interpolation method. These models in turn give 72 flight conditions, described in the
following section, for 12 weight conditions represented in Figure 4.

B. LFR models design by interpolation in the flight points

The interpolation of linear models using Linear Fractional Transformation (LFT) [14] helps to facilitate the variation
of the state space matrices versus the altitude and the TAS. Given the data extracted from the Aircraft Flight
Research Simulator provided by CAE Inc., the aircraft dynamics can be described for any conditions in the flight
envelope.

Figure 2 shows the 36 flight points chosen within the flight envelope limits. They represent the aircraft models
obtained for each 5000 ft in altitude, at 4 different speeds.

Figure 2 Cessna Citation X flight envelope

Before carrying out the interpolation, two steps need to be performed. The first step is to define the region for an
altitude and a range of TAS where the interpolation will be performed; the four corners of the region then form the
vertices. Each of these ranges has a lower and upper value which are the bounds. The second step is the
normalization of these bounds in order to attribute each coordinate of the vertices to a value equal to 1 or -1.

To optimize the level of accuracy, the smallest possible regions have been defined, containing only 3 or 4 flight
points to use as reference points for the interpolation. This definition only allows a bilinear interpolation, for which
4 coefficients must be found, using equations (20, 21, 22).
The Least Square (LS) method is employed to minimize the relative error in these reference points. From Table 1 it
can be observed that the maximum errors found for the state space matrices A and B are negligible; therefore the
results are good.

Table 1. Maximum relative error

Longitudinal mode Lateral mode

A 1.04e11 % 1.97e11 %
B 3.05e 12
% 3.97e11 %

From these results, 26 regions are obtained, which cover a large part of the flight envelope. The mesh is valid for all
of the weight and balance conditions presented in Figure 4. Figure 3 shows that some of the regions superimpose
others (darker zones) due to the common reference points, and in many cases there is not only the interpolation but
also the extrapolation.

Figure 3 Region definition

 4
x 10 Centrage du Cessna Citation X

3.6

3.4 #12

#11

3.2 #10

#8 #9
3

Poids (lb)
#6 #7
2.8

#4 #5
2.6

2.4 #2 #3

2.2 #1

2
10 15 20 25 30 35 40
x cg (%)

Figure 4 Cessna Citation X Weight/Xcg conditions

Observing each vertex of these 26 regions leads to 72 different flight points that can be analyzed. Figure 5 shows
these 72 conditions obtained by means of the LFR models, a step that make it possible to more closely approximate
the flight envelope limits.

Figure 5 Flight points obtained by LFR models

C. Flying qualities’ requirements

For aircraft longitudinal motion, two modes are perceived: 1) those that correspond to the short period, and 2) those
that correspond to the long period or phygoïd mode. Three modes are perceived in lateral aircraft motion, the Dutch
roll mode, the roll mode, and the spiral mode. Each of these modes exhibits some of the desired criteria. These
criteria are required for satisfactory flight performance, and are expressed in terms of damping [13], time constant,
overshoot, steady state error, and settling time, as shown in Table 2.

Table 2. Aircraft flying qualities criterion

Criterion Type Limits

Overshoot temporal OS<35%

Steady state error temporal ess≤2%

Settling time temporal Ts≤4s

Short period damping modal 0.3 ≤  sp ≤ 2

Long period damping modal 0.04 ≤  ph
Dutch roll damping modal 0.3 ≤  dr ≤ 2
Roll time constant temporal Tr <1.4 sec

III. H_infinity theory

H∞ represents a modern approach to characterize closed-loop performance by measuring the size of the closed-loop
transfer function matrices and the way in which the control objectives can be fixed to minimize closed-loop transfer
functions [15].

A. Standard H_infinity robust control problem

The LFT technique is used to formulate the Standard H_infinity configuration, as shown in Figure 6. P(s) denotes
the generalized plant which contains the disturbance model and the interconnection structure between the plant and
the controller K(s). In Figure 6, w denotes all of the external inputs (disturbances, noise and command) and z denotes
all of the external outputs, expressed in terms of error signals, to be minimised, including both the performance and
the robustness measures. The control input is denoted by y, and u denotes the control signal’s vectors. The objective
is to find a stabilizing controller that minimises the output z, which means that it minimises the H_infinity norm of
the closed loop transfer function from w to z.

w z

P(s)

u y

K(s)

Figure 6 Standard H_infinity configuration

The generalized plant P(s) can be written as:

[ ]

The transfer function between z and w can be written as follows [16]:

where is the lower linear fractional transformation of P and K. The optimization problem design is then
formulated as [16]:

‖ ‖

B. Mixed sensitivity H_infinity problem

The mixed sensitivity H_infinity optimization, with the configuration as shown in Figure 7, is one of several
practical optimization problems in industry. The cost function is a combination of two other cost functions, such as a
control signal’s energy limitation and a good tracking reference. In order to keep the system internally stable, these
types of specifications will be optimized for a set of stabilizing controllers with the state space gain defined as [17]:

‖[ ]‖ (
∞

z1
z2
r + e u y
K(s) G(s)
-

Figure 7 Mixed sensitivity H_infinity configuration

Weighting functions are often used to respect the design specifications in a closed-loop system (control input signal
limitations and good tracking). Thus, equation (27) can be rewritten as:

‖[ ]‖
∞
In practice, a stabilizing controller is found by iteration, using a lowest-achievable value . Equation (28) then
becomes:

‖[ ]‖
∞

where is given as the maximum value of the of the system’s closed-loop transfer function, which is
given as 1[16].

In equation (29), is used to shape the sensitivity function , and to shape the complementary
sensitivity function given as , which characterizes the disturbances and control use. is the state space
gain calculated from the H-infinity method, and and represent the weighting functions, appropriately chosen
as:

These weighting functions are the object of optimization, using both Genetic and Differential Evolution algorithms,
which are defined in the following section.

IV. Differential Evolution and Genetic algorithms

Differential Evolution
The Differential Evolution (DE) algorithm was developed in 1995 by Price and Storn [18, 19], and has been used in
global optimization in many disciplines [20]. The DE algorithm is a heuristic optimization algorithm that uses real
values (it does not need coding and encoding operations) to represent problem parameters. The key concept of DE is
its use of a differential operator to generate the donor vector. This section gives a detailed presentation of the DE
algorithm, along with a presentation of genetic algorithms.

A.DE Initialization phase

In this phase, the number of generations is chosen as the stopping criterion. The problem dimension is then set
according to the number of parameters forming the fitness function. Next, the parameters to be optimized are
represented in a vector form (equivalent to one chromosome in the genetic algorithm); at each generation, the ith
vector is described as:

⃗ [ ]

The population is initialized randomly within the search space constrained by the lower and higher boundaries for
each parameter. These boundaries are represented in vectors given by equations (33) and (34):

⃗ [ ]
⃗ [ ]

The jth component of the ith vector is initialized as:

[ ]
 where [ ]
Once the initialization phase is completed, the next step is the mutation operation

B. DE Mutation
“Mutation” is the operation of changing parameters between different vectors. In the DE algorithm, the “target
vector” is equivalent to the “parent vector” in the genetic algorithm, and the “donor vector” is obtained from the
differential mutation operation. The “trial vector”, called the “offspring” in a genetic algorithm, results from the
combination of the “trial” and the “donor” vectors. A random sampling of three different parameter vectors
is performed in the current population; each ith “target” vector is used to create its corresponding
“donor” vector. For each “mutant” vector , three different indices are randomly chosen from the
range [ ] where NP is the population number. In order to obtain the “donor” vector , the difference between
two different vectors is weighted by a randomly chosen scalar as defined in equation (36):

⃗ ⃗ ⃗

C. DE Crossover

To improve the diversity of the population, a “crossover” operation is performed, from which the “donor” and the
“target” vectors exchange their components to create the “trial” vector⃗⃗⃗ :
⃗⃗⃗ [ ].
There are two types of crossover operation: the exponential (two points modulo) and the binomial (uniform).
In the exponential crossover two integers and are chosen randomly from the range[ ], where the dimension D
refers to the number of parameters to be optimized, and then the trial vector is given as follows:

for
〈 〉 〈 〉 〈 〉

and
[ ]
where 〈 〉 denotes the modulo function with modulus .

The trial vector in the binomial crossover is given as:

if [ ]

or

otherwise

When the population diversity step is finished, a selection operation is performed as detailed next.

D. DE Selection

Using the “selection” operation, we can determine if the “trial” or “target” vectors survive in the next generation or
not, and thus maintain a constant population size. The selection operation is outlined as:
 ⃗ ⃗ (⃗ ) (⃗ )
else
⃗ ⃗ (⃗ ) (⃗ )
where (⃗ ) is the objective function or the” fitness” to be converged using the iteration process.

E. Iteration

The operations listed above are repeated until the stopping criteria have been met. These criteria consist of:
1. The maximum number of generations; and
2. The convergence of the fitness functions.

Genetic Algorithm
The genetic algorithm (GA) is an evolutionary computation and a powerful stochastic search and optimisation
technique that has become the most-recognised and most-utilised technique in the last few years ([21-24]). The GA
does not require a priori knowledge of the model to be optimized; it is an iterative process carried out until the
desired solution is found, in which each iteration represents a generation; and where the best chromosome is
examined according to its fitness. In this research the real coded genetic algorithm is considered; in which the
chromosome is a string of real parameters, with no need to perform binary coding or encoding.

To illustrate the iteration process of a genetic algorithm search, the following steps are listed below:

Step 1: The weighting functions are represented as a chromosome of fixed length; some parameters are defined,
such as the size of the chromosome population, the crossover and the mutation probabilities.

Step 2: The performance of an individual is quantified by defining a fitness function, which selects chromosomes
that will be mated.

Step 3: The random initial population is set.

Step 4: The H_infinity norm of each chromosome (individual) is computed and the gain control K is found.

Step 5: The fitness of each individual is evaluated.

Step6: A pair of individuals is selected according to the probability of their fitness.

Step 7: The next generation is reproduced by creating a pair of offspring chromosomes.

Step8: The new chromosomes are inserted into the new population.

Step 9: Starting at step 6, the process is repeated until the sizes of the new and the initial populations are equal.

Step10: The initial population is replaced with the new population.

Step 11: Return to step 4 and repeat the process until the termination condition has been satisfied.

V. DE and GA algorithms for the H-infinity problem

Trial and error and other types of methods have been performed on control parameter optimisation using heuristic
algorithms such as the Genetic Algorithm (GA) ([25], [26]), the swarm particle optimization PSO ([27, 28]), the
Fruit Fly optimization algorithm [29]. Other nonlinear methods such as fuzzy logic and neural network methods
have been applied to identification and control( [30],[31]), includinghybrid fuzzy logic ([32],[33]). All of these
methods were developed with the aim of reducing the computation time while achieving satisfactory results.

For this work, the DE and GA algorithms were selected to optimize the weighting function parameters, i.e., for
longitudinal control. The optimal controller is found using the following algorithm:

Set a population number as NP formed by the weighting functions and , the parameters from the initial vector
are:

⃗ [ ]

Each of these parameters i.e. [ ] belongs to an interval in which represents the lower bound
and represents the higher bound. The optimal gain is calculated by choosing the appropriate and
parameters and then simulating the control system in the time domain to obtain satisfactory characteristics of the
system’s response. If the satisfactory characteristics are not reached, the iteration process continues, until one of
stopping conditions has been achieved.

A. Objective function

The objective function to be minimized in order to obtain the optimal solution is calculated by the H_infinity
algorithm, in which the fitness function gives the specifications for the desired time response of the closed-loop
system using overshoot , integral square error and robustness criterion as shown in the next equation:

Two flowcharts summarize the evolutionary algorithms, presented below in Figures 8 and 9.
 Start

Initial random of population

Mutation
Formulation of new parameter
for each individual

No
New parameter
respects bounds

Yes

Crossover

Fitness test

Fitness value
No (gamma+ISE+OS)
Selection
The trial individual Control
gives best solution signalU(s)
+
_ Hinfinity G
Actual output
Yes - Y(s)

H_infinity Plant
Atribuate the trial individual controller
to the new generartion

No The trial individual

respond to the
requirements

Yes

Optimal solution

Stop

Figure 8 H_infinity optimization using the Diffrential Evolution algorithm

 Start

Initial random of population

Generation=0

Gen=Gen+1

Evaluate the fitness value for

each chromosome

Genetic Oporator

Selection

Fitness value
Crossover (gamma+ISE+OS)

Control
signalU(s)
+
_ Hinfinity G
Mutation Actual output
- Y(s)

H_infinity Plant
controller

Gen> max generation

No

Yes

Optimal solution

Stop

Figure 9 H_infinity optimization using the Real-valued Genetic Algorithm

 VI. Results analysis
The open loop of the Cessna Citation X business aircraft is composed of aircraft dynamics, actuators and sensors,
while in the aircraft’s closed loop, it is the actuators’ limits, or the performances weighting functions that are
considered in the Control Augmentation System (CAS), as in the simulation shown in Figure 10

The business aircraft Cessna Citation X is represented in the state space form for its longitudinal and lateral motions.
Robust control using the H_infinity design is then applied on the Cessna Citation X to improve its stability and its
time response.

Closed-loop simulations of the Cessna Citation X lateral aircraft mode were performed for the whole flight
envelope. The results presented below were obtained for 12 centering configurations, using 36 flight conditions
extracted from the Cessna Citation X Flight Simulator, and were further generalized for 72 flight conditions
obtained using the interpolation method.

Outer loop

Autopilot

Inner loop 1

CAS Actuators Sensors Outputs

Pilot Inputs

Cessna Citation X
dynamics

Figure 10 Closed loops representation of Cessna Citation X business aircraft

A. Results analysis

Table 3 represents the weighting functions given by the optimization using the DE algorithm. The optimization is
performed for a population size of 50, and the search range for each longitudinal weighting parameter
and lateral weighting parameter is defined as[ ]. The results
are shown for the 20th generation.
 Table 3. Optimization results

weight function DE Algorithm Genetic Algorithm GA

Range of [ ] [ ]
Solution and

Range of [ ] [ ]
Solution and

Generation number

The mean gamma value of 72 flight conditions for each centering is given in Table 4. It can be seen that the
robustness criterion limit is respected for all centering, which means that the resulting controllers are
robust against uncertainties.

Table 4. Mean gamma value

Centring 1 2 3 4 5 6 7 8 9 10 11 12
conditions
0.99 0.99 0.99 0.99 0.99 0.99 0.99 0.99 0.99 0.99 0.99 0.99

0.96 0.96 0.96 0.97 0.97 0.99 1.00 1.04 1.04 1.09 1.09 1.09

1. Results for 36 flight conditions

The closed-loop responses for pitch rate and roll angle control using the H_infinity controller are given for 36
flight conditions and all centerings in Figures 11 and 14, respectively, and the resulting pitch angle and roll angle
rate is shown in Figures 12 and 15, respectively. It is obvious that the pitch rate follows the impulse input, and
the roll angle follows the step input with a maximum steady state error (ess) of 2% and a settling time (Ts) of less
than 2sec. The aircraft reached the desired pitch angle rate , and roll angle on time without having any
overshoot (OS) for most flight conditions, except for high altitudes and speeds. The resulting pitch angle and roll
angle rate are analyzed in the same way. In addition, the magnitudes of both the elevator and the aileron control
surfaces shown in Figures 13 and 16 do not exceed the limits (given by the designer), and behave as desired.
 1.2 5

1
4

0.8

0.6
q deg/sec

teta deg
2

0.4

0.2

0
0

-0.2 -1
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
time sec time sec

Figure 11 Closed loop pitch angle rate control response Figure 12 Closed loop pitch angle response

0.8

0.6

0.4

0.2
delta deg

0
e

-0.2

-0.4

-0.6

-0.8

-1
0 1 2 3 4 5 6 7 8 9 10
time sec

Figure 13 Elevator deflection responses to pitch rate control

1.2 8

7
1

0.8
5
roll rate deg/sec

0.6
4
phi deg

3
0.4

2
0.2

0
0

-1
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
time sec time sec

Figure 14 Closed loop roll angle control response Figure 15 Closed loop roll rate control response
 70

60

50

40

aileron angle deg

30

20

10

-10

-20
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
time sec

Figure 16 Closed loop aileron angle control response

2. Results for 72 flight conditions:

The simulation was performed for 72 flight conditions for each centering as shown in Figures 17, 18, 19 and 20.
Some responses show an overshoot (OS) of a maximum of 28% for pitch angle rate , and 18% for roll angle .
Globally, the aircraft longitudinal and lateral motions are stabilized with the H_infinity controller. For both controls,
the pitch angle rate and roll angle , the resulting response satisfies the handling quality requirements with a
steady state error less than 2%, and OS of less than 30%. The H∞ method thus gives a controller that approximates
the handling quality requirements for both longitudinal and lateral motions as given in Table 2, improving aircraft
stability and performance.

3.5

2.5

2
amplitude

1.5

0.5

-0.5
0 1 2 3 4 5 6 7 8 9 10
time (sec)

Figure 17 Pitch angle rate responses for the first centering (22000 lb/33%)
 1.4

1.2

0.8

q deg/sec
0.6

0.4

0.2

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
temps (s)

Figure 18 Pitch angle rate responses for the 12th centering (34000 lb/20%)

1.4

1.2

1
roll angle phi deg

0.8

0.6

0.4

0.2

0
0 1 2 3 4 5 6 7 8 9 10
temps (s)

Figure 19 Roll angle responses for the first centering (22000 lb/33%)
 1.4

1.2

roll angle phi deg

0.8

0.6

0.4

0.2

0
0 1 2 3 4 5 6 7 8 9 10
temps (s)

Figure 20 Roll angle responses for the 12th centering (34000 lb/20%)

These results have been validated for almost all of the flight conditions using a linear model, except for a few flight
conditions which represent 0.5% for the pitch angle rate and 2% for roll angle , as indicated in Figures 21 and 22,
where they present unacceptable responses. Their details are generated by the Matlab code in text format as shown
on Tables 5 and 6. From these results, it can be deduced that the uncontrollable flight conditions belong either to
very high loads or to low loads with flight points outside of the aircraft’s envelope limit; which is understandable;
also, when the interpolation was performed to create more flight conditions, the trim condition for these additional
flight points may have been lost.

4
x 10
6
uncontrollable flight condition

4
Altitude (ft)

0
100 150 200 250 300 350 400 450 500 550
TAS (kts)

Figure 21 Unacceptable pitch angle rate responses for linear model validation
 Table 5. Uncontrollable flight conditions for pitch angle rate

Region Vertex Centering

23 1 1
25 4 1
26 4 1
8 1 12
11 3 12

4
x 10
6

uncontrollable flight condition

4
Altitude (ft)

0
100 150 200 250 300 350 400 450 500 550
TAS (kts)

Figure 22 Unacceptable roll angle responses for linear model validation

Table 6. Uncontrollable flight conditions for roll angle

Region Vertex Centering

22 1 1
23 4 3
26 1 3
1 1 10
26 4 11
11 1 12
25 4 12
26 1 5
26 2 5
26 3 5
26 4 5
25 2 6
 25 4 6
17 1 11
17 3 11
18 1 11

3. Non-linear validation

Finally, to prove the efficiency of the optimized controller and its robustness against uncertainties, a nonlinear
validation was performed using the Cessna Citation X aircraft’s nonlinear model. A simulation of the pitch angle
rate and roll angle control responses were performed, with the results shown in Figures 23 and 25, respectively,
for the altitude of 2000 ft, TAS of 230 knots and load of 26000 lb, and with varying mass shown in Figures 24 and
26. It can be seen that the pitch angle rate and roll angle responses remain stable during the simulation, despite
the presence of uncertainties (mass variation), and that all the performance criteria are reached; therefore the system
is robust as desired.

0.8

0.6
q deg/sec

0.4

0.2

-0.2
0 1 2 3 4 5 6 7 8 9 10
time sec

Figure 23 Pitch angle rate hold control responses using nonlinear aircraft model

4
x 10 Altitude (ft)
2.04

2.02

1.98
0 1 2 3 4 5 6 7 8 9 10

True airspeed (kts)

231

230.5

230

229.5
0 1 2 3 4 5 6 7 8 9 10

Heading(deg)
118

117.5

117
0 1 2 3 4 5 6 7 8 9 10

4
x 10 Mass(lb)
2.6

2.595

2.59
0 1 2 3 4 5 6 7 8 9 10
Time(sec)

Figure 24 Altitude, true airspeed, heading and mass variation responses using nonlinear aircraft model
 1.4

1.2

phi (deg) 0.8

0.6

0.4

0.2

0
0 1 2 3 4 5 6 7 8 9 10
Time(sec)

Figure 25 Roll angle control responses on a nonlinear aircraft model

4
x 10 Altitude (ft)
2

2
0 1 2 3 4 5 6 7 8 9 10

True airspeed (kts)

230.01

230

229.99

229.98
0 1 2 3 4 5 6 7 8 9 10

Heading(deg)
118

117.5

117
0 1 2 3 4 5 6 7 8 9 10

4
x 10 Mass(lb)
2.6

2.598

2.596
0 1 2 3 4 5 6 7 8 9 10
Time(sec)

Figure 26 Altitude, true airspeed, heading and mass variation responses on a nonlinear aircraft model
 VII. Conclusion
An optimization was performed using two algorithms, the Differential Evolution (DE) and the Genetic algorithms.
The H_infinity method was used to optimize the control augmentation system design. Simulations were performed
using the resulting optimal gains for the Cessna Citation X aircraft’s longitudinal and lateral motion, for12 centering
(Xcg positions) and 72 flight conditions, selected to cover the whole flight envelope. The optimized feedback gains
enhanced the aircraft’s closed-loop performance, according to the handling qualities and specifications given in
Table 1, while the Differential evolution (DE) and Genetic Algorithms (GA) algorithms showed a high efficiency in
global optimization with minimum time convergence.

Acknowledgement(s)

This work was performed at the LARCASE (Laboratory of active controls, avionics and aeroservoelasticity
research). We want to thank CAE Inc for their support in the development of the Aircraft Research Flight Simulator
at the LARCASE laboratory. This simulator was obtained thanks to research grants that were approved by the
Canadian Foundation of Innovation (CFI) and the MDEIE. Thanks are also due to Mrs Odette Lacasse at the ETS
for her continuing enthusiasm and support.

References

[1] S. Ibrir and R. Botez, "Robust stabilization of uncertain aircraft active systems," Journal
of Vibration and Control, vol. 11, pp. 187-200, 2005.
[2] N. Aouf, B. Boulet, and R. Botez, "H2 and H-optimal gust load alleviation for a flexible
aircraft," Proceedings of 2000 American Control Conference (ACC 2000), 28-30 June
2000, Danvers, MA, USA, 2000, pp. 1872-6.
[3] G. F. Zames, B.A. , "Feedback, minimax sensitivity, and optimal robustness " Automatic
Control, IEEE Transactions vol. 28, 1983
[4] R. A. Hyde and K. Glover, "VSTOL aircraft control system design using scheduled
<e1>H</e1>&infin; controllers," in Practical Methods for Robust Control System
Design, IEE Colloquium on, 1991, pp. 211-213.
[5] A. K. P. G.J.Balas, J.Renffrow,C.Mllaney, and R. T M'Closeky,, ""Control of the F-14
aircraft lateral-directional axis during powered approach"," Journal of Guidance,
Control, and Dynamics, vol. 21, pp. 899-908, 1998.
[6] N. Aouf, B. Boulet, and R. Botez, "A gain scheduling approach for a flexible aircraft," in
American Control Conference, 2002. Proceedings of the 2002, 2002, pp. 4439-4442
vol.6.
[7] Y. Ciann-Dong, J. Hann-Shing, and L. Shin-Whar, "Experimental design of
H_&infin; weighting functions for flight control systems," in American
Control Conference, 1994, 1994, pp. 2516-2520 vol.3.
[8] H. Jiankun, C. Bohn, and H. R. Wu, "Practical H_&infin; weighting
functions and their application to real-time control of a pilot plant," in American Control
Conference, 1999. Proceedings of the 1999, 1999, pp. 920-924 vol.2.
[9] A. Lanzon, "Weight optimisation in scriptHsignloop-shaping," Automatica, vol. 41, pp.
1201-1208, 2005.
[10] M. Osinuga, S. Patra, and A. Lanzon, "State-space solution to weight optimization
problem in H loop-shaping control," Automatica, vol. 48, pp. 505-13, 2012.
[11] N. Boely, R. M. Botez, and G. Kouba, "Identification of a non-linear F/A-18 model by
the use of fuzzy logic and neural network methods," Proceedings of the Institution of
Mechanical Engineers, Part G (Journal of Aerospace Engineering), vol. 225, pp. 559-74,
2011.
[12] C. Hamel, Sassi, A., Botez, R., Dartigues, C., , "Cessna Citation X aircraft global model
identification from fight tests," SAE International Journal of Aerospace vol. 6(1), pp.
106-114, 2013.
[13] R. C. Nelson, "Flight Stability & Automatic Control," 1997.
[14] C. Poussot-Vassal and C. Roos, "Flexible aircraft reduced-order LPV model generation
from a set of large-scale LTI models," in American Control Conference (ACC), 2011,
2011, pp. 745-750.
[15] G. Zames and L. Y. Wang, "Local-global double algebras for slow H adaptation. I.
Inversion and stability," IEEE Transactions on Automatic Control, vol. 36, pp. 130-42,
1991.
[16] P. H. P. a. M. M. K. D.-W. Gu, Robust Control Design with Matlab, 2005.
[17] D. J. WALKER, "Practical Aspects of Implementing H-Infinity
Controllers on a FBW Research Helicopter," RTO AVT Symposium, 2000.
[18] K. V. Price, "Differential evolution: a fast and simple numerical optimizer," in Fuzzy
Information Processing Society, 1996. NAFIPS., 1996 Biennial Conference of the North
American, 1996, pp. 524-527.
[19] R. Storn and K. Price, "Minimizing the real functions of the ICEC'96 contest by
differential evolution," in Evolutionary Computation, 1996., Proceedings of IEEE
International Conference on, 1996, pp. 842-844.
[20] Y. Boughari, R. M. Botez, F. Theel, G. Ghazi, "Optimal Flight Control on Cessna X
Aircraft using Diffrential Evolution," presented at the Modeling, Identification, and
Control, 2014.
[21] C. Wongsathan and C. Sirima, "Application of GA to design LQR controller for an
Inverted Pendulum System," in Robotics and Biomimetics, 2008. ROBIO 2008. IEEE
International Conference on, 2009, pp. 951-954.
[22] J. F. Wu and L. Xiao, "Research on LQR control optimized by elitist preserving genetic
algorithm," in 2nd International Conference on Information Science and Engineering,
ICISE2010, December 4, 2010 - December 6, 2010, Hangzhou, China, 2010, pp. 5327-
5329.
[23] H.-T. Wang, S.-W. Guo, P. Guo, and Z.-Z. Qin, "Application of genetic algorithms for
aerodynamic parameter estimation of large parachute," Yuhang Xuebao/Journal of
Astronautics, vol. 31, pp. 981-985, 2010.
[24] S. Yang, Q. Weiqi, W. Qing, and H. Kaifeng, "Aerodynamic Parameter Estimation Using
Genetic Algorithms," in Evolutionary Computation, 2006. CEC 2006. IEEE Congress on,
2006, pp. 629-633.
[25] M. J. Neath, A. Swain, U. Madawala, and D. Thrimawithana, "An Optimal PID
Controller for a Bidirectional Inductive Power Transfer System Using Multi-objective
Genetic Algorithm," Power Electronics, IEEE Transactions on, vol. PP, pp. 1-1, 2013.
 [26] M. K. Tan, Y. K. Chin, H. J. Tham, and K. T. K. Teo, "Genetic algorithm based PID
optimization in batch process control," in Computer Applications and Industrial
Electronics (ICCAIE), 2011 IEEE International Conference on, 2011, pp. 162-167.
[27] M. S. Rahimian and K. Raahemifar, "Optimal PID controller design for AVR system
using particle swarm optimization algorithm," in Electrical and Computer Engineering
(CCECE), 2011 24th Canadian Conference on, 2011, pp. 000337-000340.
[28] R. G. Kanojiya and P. M. Meshram, "Optimal tuning of PI controller for speed control of
DC motor drive using particle swarm optimization," in Advances in Power Conversion
and Energy Technologies (APCET), 2012 International Conference on, 2012, pp. 1-6.
[29] H. Jiuqi, W. Peng, and Y. Xin, "Tuning of PID controller based on Fruit Fly Optimization
Algorithm," in Mechatronics and Automation (ICMA), 2012 International Conference
on, 2012, pp. 409-413.
[30] G. Kouba, R. Botez, N. Boëly, "Identification of F/A-18 model from flight tests using the
fuzzy logic method," in Proceedings of the 47th AIAA Aerospace Sciences Meeting
Including The New Horizons Forum and Aerospace Exposition, Orlando, FL, USA.,
2009.
[31] N. Boëly, R. Botez, G. Kouba, "Identification of F/A-18 nonlinear model between control
and structural deflections," in Proceedings of the 47th AIAA Aerospace Sciences Meeting
Including The New Horizons Forum and Aerospace Exposition, Orlando, FL, USA, 2009.
[32] T. L. Grigorie, R. M. Botez, A. V. Popov, M. Mamou, and Y. Mebarki, "A hybrid fuzzy
logic proportionalintegral-derivative and conventional on-off controller for morphing
wing actuation using shape memory alloy Part 1: Morphing system mechanisms and
controller architecture design," Aeronautical Journal, vol. 116, pp. 433-449, 2012.
[33] A. V. Popov, L. T. Grigorie, R. M. Botez, M. Mamou, Y. Mebarki,, "Real time morphing
wing optimization validation using wind tunnel tests," in Journal of Aircraft vol.,vol.
47(4), ed, pp. 1346-1355.

View publication stats