THE LEFT AND RIGHT BLOCK POLE PLACEMENT COMPARISON STUDY: APPLICATION TO FLIGHT DYNAMICS

Informatics Engineering, an International Journal (IEIJ), Vol.4, No.1, March 2016
DOI : 10.5121/ieij.2016.4105 41
THE LEFT AND RIGHT BLOCK POLE PLACEMENT
COMPARISON STUDY: APPLICATION TO FLIGHT
DYNAMICS
BEKHITI Belkacem1
DAHIMENE Abdelhakim1
NAIL Bachir2
and HARICHE Kamel1
1
Electronics and Electrotechnics Institute, University of Boumerdes, 35000 Algeria.
2
Technology and sciences Institute, University of Djelfa, Algeria
ABSTRACT
It is known that if a linear-time-invariant MIMO system described by a state space equation has a number
of states divisible by the number of inputs and it can be transformed to block controller form, we can
design a state feedback controller using block pole placement technique by assigning a set of desired Block
poles. These may be left or right block poles. The idea is to compare both in terms of system’s response.
KEYWORDS
MIMO, Block Controller Form, State Feedback Controller, Block Pole Placement Technique, Left and/or
Right Block Poles.
1. INTRODUCTION
In most applied mathematical research, the aim is to investigate and control a given system. A
system is defined to mean a collection of objects which are related by interactions and produce
various outputs in response to different inputs.Examples of such systems are chemical plants,
aircrafts, spacecraft, biological systems, or even the economic structure of a country or regions
see [1, 4 and 17]. The control problems associated with these systems might be the production of
some chemical product as efficiently as possible, automatic landing of aircraft, rendezvous with
an artificial satellite, regulation of body functions such as heartbeat or blood pressure, and the
ever-present problem of economic inflation [18].To be able to control a system, we need a valid
mathematical model. However practical systems are inherently complicated and highly non-
linear. Thus, simplifications are made, such as the linearization of the system. Error analysis can
then be employed to give information on how valid the linear mathematical model is, as an
approximation to the real system [17].
A large-scale MIMO system, described by state equations, is often decomposed into small
subsystems, from which the analysis and design of the MIMO system can be easily performed.
Similarity block transformations are developed to transform a class of linear time-invariant
MIMO state equations, for which the systems described by these equations have the number of
inputs dividing exactly the order of the state, into block companion forms so that the classical
lines of thought for SISO systems can be extended to MIMO systems [19].

42
Such systems can be studied via the eigen structure, eigen values and eigenvectors, of the state
matrix A. The eigenvalues and eigenvectors can determine system performance and robustness
far more directly and explicitly than other indicators. Hence their assignment should improve
feedback system performance and robustness distinctly and effectively [1].Eigen structure
assignment (EA) is the process of applying negative feedback to a linear, time-invariant system
with the objective of forcing the latent-values and latent-vectors to become as close as possible to
a desired eigen structure. EA, in common with other multivariable design methodologies, is
inclined to use all of the available design freedom to generate a control solution. It is a natural
choice for the design of any control system whose desired performance is readily represented in
terms of an ideal eigen structure. Many research works has been done on EA [20, 21, 22, 23 and
24] and more specifically on flight control systems [25, 26, and 27].
The design of state feedback control in MIMO systems leads to the so-called matrix polynomials
assignment [2]. The use of block poles constructed from a desired set of closed-loop poles offers
the advantage of assigning a characteristic matrix polynomial rather than a scalar one [3]. The
desired characteristic matrix polynomial is first constructed from a set of block poles selected
among a class of similar matrices, and then the state feedback is synthesized by solving matrix
equations. The forms of the block poles used in our work are the diagonal, the controller and the
observer forms. Robustness is assessed, in each case, using the infinity norm, the singular value
of the closed loop transfer matrix and the condition number of the closed-loop transfer matrix.
Time response is assessed by plotting the step response and comparing the time response
characteristics [1]. A comparison study is conducted to determine, in light of the above criteria,
the best choice of the form of the block poles.
In thepresent paper, firstly we have started the work by introducing some theoretical preliminaries
on matrix polynomials, after that a theoretical background on robustness and sensitivity
analysisin term of responses is illustrated and briefly discussed, it is then followed by an
application to flight dynamic system by doing a comparison study in term of block roots form. As
a fifth section a discussion of the obtained results is performed, and finally the paper is finished
by a comparison study and a conclusion.
2. PRELIMINARIES
2.1. Definition of a polynomial matrix
Definition 1: given a set of 𝑚 × 𝑚complex matrices {𝐴0, 𝐴1, … , 𝐴𝑙}the following matrix valued
function of the complex variableλ is called matrix polynomial of degree (index)𝑙 and order𝑚:
(𝐴(λ): is called also λ-matrix.)
𝐴(𝜆) = 𝐴0 𝜆𝑙
+ 𝐴1 𝜆𝑙−1
+ ⋯ + 𝐴𝑙−1 𝜆 + 𝐴𝑙 (1)
Consider the system described by the following dynamic equation:
{
𝑥̇( 𝑡) = 𝐴𝑥(𝑡) + 𝐵𝑢(𝑡)
𝑦(𝑡) = 𝐶𝑥(𝑡)
(2)
Assuming that the system can be transformed to a block controller form, this means:

43
i. The number
𝑛
𝑚
= 𝑙 is an integer.
ii. The matrix 𝑊𝑐 = {𝐵, 𝐴𝐵, … , 𝐴𝑙−1
𝐵}is of full rank n.
Then we use the following transformation matrix:
𝑇𝑐 =
[
𝑇𝑐𝑙
𝑇𝑐𝑙 𝐴
⋮
𝑇𝑐𝑙 𝐴𝑙−2
𝑇𝑐𝑙 𝐴𝑙−1
]
Where 𝑇𝑐𝑙 = [𝑂 𝑚 𝑂 𝑚, … , 𝐼 𝑚][𝐵, 𝐴𝐵, … , 𝐴𝑙−1
𝐵]−1
(3)
The new system becomes:
{
𝑥̇( 𝑡) = 𝐴 𝑐 𝑥(𝑡) + 𝐵𝑐 𝑢(𝑡)
𝑦(𝑡) = 𝐶𝑐 𝑥(𝑡)
(4)
With
𝐴 𝑐 = 𝑇𝑐 𝐴𝑇𝑐
−1
, 𝐵𝑐 = 𝑇𝑐 𝐵, 𝐶𝑐 = 𝐶𝑇𝑐
−1
𝐴 𝑐 =
[
𝑂 𝑚
𝑂 𝑚
⋮
𝑂 𝑚
−𝐴𝑙
𝐼 𝑚
𝑂 𝑚
⋮
𝑂 𝑚
−𝐴𝑙−1
…
…
…
…
…
𝑂 𝑚
𝑂 𝑚
⋮
𝐼 𝑚
−𝐴1
]
, 𝐵𝑐 =
[
𝑂 𝑚
𝑂 𝑚
⋮
𝑂 𝑚
𝐼 𝑚
]
, 𝐶𝑐 = [𝐶𝑙 𝐶𝑙−1 ⋯ 𝐶1]
2.2.Matrix transfer function
The matrix transfer function of this open-loop system is given by:
𝑇𝐹𝑅(𝑠) = 𝑁𝑅(𝑠)𝐷 𝑅
−1
(𝑠)(5)
Where:
 𝑁𝑅(𝑠) = [𝐶1 𝑠 𝑙−1
+ ⋯ + 𝐶𝑙−1 𝑠 + 𝐶𝑙]
 𝐷 𝑅(𝑠) = [𝐼 𝑚 𝑠 𝑙
+ 𝐴1 𝑠 𝑙−1
+ ⋯ + 𝐴𝑙]
This transfer function is called the Right Matrix Fraction Description (RMFD); we need to use it
in the block controller form.
It should be noted that the behavior of the system depends on the characteristic matrix
polynomial𝐷 𝑅(𝑠).
2.3. Concept of solvents (block roots)
A root for a polynomial matrix is not well defined. If it is defined as a complex number it may not
exist at all. Then we may consider a root as a matrix called block root.

44
2.3.1. Right solvent
Given the matrix polynomial of order𝑚 and index 𝑙defined by:
𝐷 𝑅(𝑠) = 𝐼 𝑚 𝑠 𝑙
+ 𝐴1 𝑠 𝑙−1
+ ⋯ + 𝐴𝑙 = ∑ 𝐴𝑖 𝑠 𝑙−𝑖
𝑙
𝑖=0
where𝐴0 = 𝐼 𝑚 (6)
A right solvent, denoted by𝑅, is a 𝑚 × 𝑚 matrix satisfying:
𝐷 𝑅(𝑅) = 𝐴0 𝑅 𝑙
+ 𝐴1 𝑅 𝑙−1
+ ⋯ + 𝐴𝑙−1 𝑅 + 𝐴𝑙 = 0 𝑚 (7)
2.3.2. Left solvent
A left solvent of the matrix polynomial 𝐷(𝑠) defined above, denoted by 𝐿, is 𝑚 × 𝑚matrix
satisfying:
𝐷 𝑅(𝐿) = 𝐿𝑙
𝐴0 + 𝐿𝑙−1
𝐴1 + ⋯ + 𝐿𝐴𝑙−1 + 𝐴𝑙 = 0 𝑚 (8)
A right solvent, if exist, is considered as a right block root. A left solvent, if exist, is considered as
a left block root.
2.3.3. Latent root and latent vector
 A complex number𝜆satisfying det(𝐷 𝑅(𝜆)) = 0 is called a latent root of𝐷 𝑅(𝜆).
 Any vector 𝒙𝒊associated with the latent root satisfying 𝐷 𝑅(𝜆𝑖)𝒙𝒊 = 0 𝑚is a right latent
vector of𝐷 𝑅(𝜆).
The relationship between latent roots, latent vectors, and the solvents can be stated as follows:
Theorem: If𝐷(𝜆)has 𝑛 linearly independent right latent vectors 𝑝1, 𝑝2, … , 𝑝 𝑛 (left latent
vectors𝑞1, 𝑞2, … , 𝑞 𝑛) corresponding to latent roots𝑝1, 𝑝2, … , 𝑝 𝑛, then 𝑃Ʌ𝑃−1
, (𝑄Ʌ𝑄−1) is a right
(left) solvent. Where:𝑃 = [𝑝1, 𝑝2, … , 𝑝 𝑛], (𝑄 = [𝑞1, 𝑞2, … , 𝑞 𝑛] 𝑇) and Ʌ = 𝑑𝑖𝑎𝑔(𝜆1, 𝜆2, … , 𝜆 𝑛).
Proof: see [16]
Theorem: If 𝐷(𝜆)has 𝑛 latent roots𝜆1, 𝜆2, … , 𝜆 𝑛and the corresponding right latent
vectors𝑝1, 𝑝2, … , 𝑝 𝑛has as well as the left latent vectors𝑞1, 𝑞2, … , 𝑞 𝑛are both linearly independent,
then the associated right solvent 𝑅 and left solvent 𝐿 are related by: 𝑅 = 𝑊𝐿𝑊−1
, Where
𝑊 = 𝑃𝑄 and 𝑃 = [𝑝1, 𝑝2, … , 𝑝 𝑛], (𝑄 = [𝑞1, 𝑞2, … , 𝑞 𝑛] 𝑇). and“T “stands for transpose.(Proof:
see [16] )
2.3.4. Complete set of solvents
Definition 1: Consider the set of solvents {𝑅1, 𝑅2, … , 𝑅𝑙}constructed from the eigenvalues
(𝜆1, 𝜆2, … , 𝜆 𝑛)of a matrix𝐴 𝑐,{𝑅1, 𝑅2, … , 𝑅𝑙}is a complete set of solvents if and only if:

45
{
∪ 𝜎(𝑅𝑖) = 𝜎(𝐴 𝑐)
𝜎(𝑅𝑖) ∩ 𝜎(𝑅𝑗) = ∅
det(𝑉𝑅(𝑅1, 𝑅2, … , 𝑅𝑙)) ≠ 0
(10)
Where:
𝜎 denotes the spectrum of the matrix.
𝑉𝑅 is the block Vandermonde matrix corresponding to{𝑅1, 𝑅2, … , 𝑅𝑙} given as:
𝑉𝑅(𝑅1, 𝑅2, … , 𝑅𝑙) =
[
𝐼 𝑚
𝑅1
⋮
𝑅1
𝑙−1
𝐼 𝑚
𝑅2
⋮
𝑅2
𝑙−1
…
…
⋱
…
𝐼 𝑚
𝑅𝑙
⋮
𝑅𝑙
𝑙−1
]
(11)
The conditions for the existence and uniqueness of the complete set of solvents have been
investigated by P. Lancaster [15] and Malika Yaici [3].
Remark: We can define a set of left solvents in the same way as in the previous theorem.
2.4. Constructing a matrix polynomial from a complete set of solvents
We want to construct the matrix polynomial defined by 𝐷(λ)from a set of solvents or a set of
desired poles which will determine the behavior of the system that we want.Suppose we have a
desired complete set of solvents. The problem is to find the desired polynomial matrix or the
characteristic equation of the block controller form defined by:
𝐷(𝜆) = 𝐷0 𝜆𝑙
+ 𝐷1 𝜆𝑙−1
+ ⋯ + 𝐷𝑙−1 𝜆 + 𝐷𝑙
We want to find the coefficients𝐷𝑖for 𝑖 = 1, … , 𝑙
a. Constructing from a complete set of right solvents:
Consider a complete set of right solvents {𝑅1, 𝑅2, … , 𝑅𝑙}for the matrix polynomial𝐷(𝜆), If 𝑅𝑖is a
right solvent of 𝐷(𝜆)so:
𝑅𝑖
𝑙
+ 𝐷1 𝑅𝑖
𝑙−1
+ ⋯ + 𝐷𝑙−1 𝑅𝑖 + 𝐷𝑙 = 𝑂 𝑚 ⇒ 𝐷1 𝑅𝑖
𝑙−1
+ ⋯ + 𝐷𝑙−1 𝑅𝑖 + 𝐷𝑙 = −𝑅𝑖
𝑙
Replacing 𝑖 from 1 𝑡𝑜 𝑙 we get the following:
[𝐷 𝑑𝑙, 𝐷 𝑑(𝑙−1), … , 𝐷 𝑑1] = −[𝑅1
𝑙
, 𝑅2
𝑙
, … , 𝑅𝑙
𝑙
]𝑉𝑅
−1
(12)
Where 𝑉𝑅is the right block Vander monde matrix

46
b. Constructing from a complete set of left solvents:
Consider a complete set of left solvents{𝐿1, 𝐿2, … , 𝐿𝑙}for the matrix polynomial 𝐷(𝜆) If 𝐿𝑖is a left
solvent of 𝐷(𝜆)so:
𝐿𝑖
𝑙
+ 𝐿𝑖
𝑙−1
𝐷1 + ⋯ + 𝐿𝑖 𝐷𝑙−1 + 𝐷𝑙 = 𝑂 𝑚 ⇒ 𝐿𝑖
𝑙−1
𝐷1 + ⋯ + 𝐿𝑖 𝐷𝑙−1 + 𝐷𝑙 = −𝐿𝑖
𝑙
Replacing 𝑖 from 1 𝑡𝑜 𝑙 we get the following:
[
𝐷 𝑑𝑙
𝐷 𝑑(𝑙−1)
⋮
𝐷 𝑑1
]
= −𝑉𝐿
−1
[
𝐿1
𝑙
𝐿2
𝑙
⋮
𝐿𝑙
𝑙
]
(13)
Where 𝑉𝐿is the left block Vandermonde matrix defined by:
𝑉𝐿(𝐿1, 𝐿2, … , 𝐿𝑙) =
[
𝐼 𝑚
𝐼 𝑚
⋮
𝐼 𝑚
𝐿1
𝐿2
⋮
𝐿𝑙
…
…
⋱
…
𝐿1
𝑙−1
𝐿2
𝑙−1
⋮
𝐿𝑙
𝑙−1
]
(14)
2.5. State feedback design
Consider the general linear time-invariant dynamic system described by the previous state space
equation (2). Now applying the state feedback 𝑢 = −𝐾𝑥(𝑡)to this system, where: 𝐾 is a 𝑚 ×
𝑛gain matrix.
After using the block controller form transformation for the system, we get:𝑢 = −𝐾𝑐 𝑥 𝑐(𝑡)
Where: 𝐾 = 𝐾𝑐 𝑇𝑐 = [𝐾𝑐𝑙, 𝐾 𝑐(𝑙−1), … , 𝐾𝑐1]𝑇𝑐and𝐾𝑐𝑖 ∈ 𝑅 𝑚×𝑚
for 𝑖 = 1, … , 𝑙
Then the resulting closed loop system is shown below:
{
𝑥̇ 𝑐 = (𝐴 𝑐 − 𝐵𝑐 𝐾𝑐)𝑥 𝑐
𝑦𝑐 = 𝐶𝑐 𝑥 𝑐
Where:
(𝐴 𝑐 − 𝐵𝑐 𝐾𝑐) =
[
𝑂 𝑚
𝑂 𝑚
⋮
𝑂 𝑚
−(𝐴𝑙 + 𝐾𝑐𝑙)
…
…
…
…
…
𝑂 𝑚
𝑂 𝑚
⋮
𝐼 𝑚
−(𝐴1 + 𝐾𝑐1)
]

47
The characteristic matrix polynomial of this closed loop system is:
𝐷(𝜆) = 𝐼 𝑚 𝜆𝑙
+ (𝐴1 + 𝐾𝑐1)𝜆𝑙−1
+ ⋯ + (𝐴𝑙 + 𝐾𝑐𝑙)
From a set of desired eigenvalues, we construct the solvents then we construct the desired
characteristic matrix polynomial in the form:
𝐷 𝑑(𝜆) = 𝐼 𝑚 𝜆𝑙
+ 𝐷 𝑑1 𝜆𝑙−1
+ ⋯ + 𝐷 𝑑𝑙 (15)
By putting 𝐷 𝑑(𝜆) = 𝐷(𝜆)we get the coefficients 𝐾𝑐𝑖as follows:
𝐾𝑐𝑖 = 𝐷 𝑑𝑖 − 𝐴𝑖for 𝑖 = 1, … , 𝑙 (16)
After that we find the gain matrix by the following formula 𝐾 = 𝐾𝑐 𝑇𝑐.
3. ROBUSTNESS AND SENSITIVITY ANALYSIS
3.1. The sensitivity of eigenvalues (robust performance)
Robust performance is defined as the low sensitivity of system performance with respect to
system model uncertainty and terminal disturbance. It is well known that the eigenvalues of the
dynamic matrix determine the performance of the system then from that the sensitivities of these
eigenvalues determine the robustness of the system (2).
Theorem:Let 𝜆 𝑎𝑛𝑑 𝜆′
be the eigenvalues of the matrices 𝐴 𝑎𝑛𝑑 𝐴 + ∆𝐴 respectively, and let
𝑉 be the right eigenvectors matrix of 𝐴 , then Wilkinson has derived the variation in eigenvalues
as follows:
min𝑖(𝜆𝑖 − 𝜆𝑖
′
) ≜ min𝑖(∆(𝜆𝑖)) ≤ 𝜅(𝑉). ‖∆𝐴‖ (17)
‖. ‖ Stands for the matrix norm and 𝜅(. )Is the condition number. (Proof:see[1])
Theorem:Let 𝜆𝑖, 𝒗𝒊and𝒕𝒊be the 𝑖 𝑡ℎ
eigenvalue, right and left eigenvectors of a
matrix 𝐴 respectively(𝑖 = 1, … , 𝑛), let𝜆𝑖 + ∆𝜆𝑖be the 𝑖 𝑡ℎ
eigenvalue of the matrix 𝐴 + ∆𝐴, then
for small enough‖∆𝐴‖
∆𝜆𝑖 ≤ ‖𝒗𝒊‖‖𝒕𝒊‖‖∆𝐴‖ ≜ 𝑠(𝜆𝑖)‖∆𝐴‖ (18)
Such that: 𝑠(𝜆𝑖) = ‖𝒗𝒊‖‖𝒕𝒊‖.(Proof: see [1].)
This theorem shows that the sensitivity of an eigenvalue is determined by its corresponding left
and right eigenvectors and it is valid for small perturbations in the matrix𝐴.
3.2. Relative change
Let 𝜆𝑖 𝑎𝑛𝑑 𝜆𝑖
′
be the eigenvalues of the matrices 𝐴 𝑎𝑛𝑑 𝐴 + ∆𝐴 respectively.The relative
change𝑟𝑖of the eigenvalue 𝜆𝑖is defined as follows:

48
𝑟𝑖 =
|𝜆𝑖 − 𝜆𝑖
′
|
|𝜆𝑖|
=
|∆𝜆𝑖|
|𝜆𝑖|
𝑖 = 1, . . , 𝑛 (19)
3.3. Robust Stability
The stability of a system is the most wanted property. So its sensitivity to uncertainties is very
important when analyzing and designing the system. Stability is affected by the system
eigenvalues of the dynamic matrix so the sensitivity of these eigenvalues directly affects the
robust stability of the system (2). There are three robust stability measures using the sensitivity of
this system eigenvalues defined as follows:
Definition: Let{𝜆1, 𝜆2, … , 𝜆 𝑛}be the set of eigenvalues of an𝑛 × 𝑛matrix denoted by 𝐴and
assuming that all the eigenvalues are stable (𝑖. 𝑒: 𝑅𝑒{𝜆𝑖} < 0 ∀𝑖 ) and all the eigenvalues are
already arbitrary assigned for guaranteed performance, the three robust stability measures are
defined by:
1. 𝑀1 = min0<𝜔<∞{𝜎(𝐴 − 𝑗𝜔𝐼)}, 𝜎denotes the smallest singular value.
2. 𝑀2 = (𝜅(𝑉))
−1
. |𝑅𝑒{𝜆 𝑛}|such that: |𝑅𝑒{𝜆 𝑛}| ≤ ⋯ ≤ |𝑅𝑒{𝜆1}| and 𝑉is the right
eigenvector matrix of 𝐴 .
3. 𝑀3 = min0≤𝑖≤𝑛 {(𝑠(𝜆𝑖))
−1
|𝑅𝑒{𝜆𝑖}|}
4. LEFT/RIGHT BLOCK POLE PLACEMENT COMPARISON STUDY
APPLICATION TO THE B747AIRCRAFT
When the controlled system is multi-input multi-outputthen an infinite number of gain matrices
K may be found which will provide the required stability characteristics. Consequently, an
alternative and very powerful method for designing feedback gains for auto-stabilization systems
is the right and/or left block pole placement method. The method is based on the manipulation of
the equations of motion in block state space form and makes full use of the appropriate
computational tools in the analytical process.
The motion of the aircraft is described in terms of force, moment, linear and angular velocities
and attitude resolved into components with respect to the chosen aircraft fixed axis system. For
convenience it is preferable to assume a generalized body axis system in the first instance [5].
Whenever the aircraft is disturbed from equilibrium the force and moment balance is upset and
the resulting transient motion is quantified in terms of the perturbation variables. The perturbation
variables are shown in Fig. 1.

49
Figure: 1 Motion variables notation
The basic dynamics using Newton’s laws are, assuming that the mass isn’t variable:
𝐹⃗ = 𝑚 𝒱𝑐
𝐼⃗⃗⃗⃗⃗̇
𝑎𝑛𝑑 𝑀⃗⃗⃗ = 𝐻 𝐼⃗⃗⃗⃗⃗̇
(20)
Where: 𝐼is indicating that the vector is written in the base of the inertial axes.
𝐹⃗is the resultant force.𝑚 is the mass of the body.𝒱𝑐
⃗⃗⃗⃗⃗̇ is the acceleration of the body.
𝐻 𝐼⃗⃗⃗⃗⃗is the angular momentum.𝑀⃗⃗⃗is the resultant moment.
Referring to the body fixed frame the equations can be rewritten as
𝐹⃗ = 𝑚 𝒱𝑐
𝐵⃗⃗⃗⃗⃗⃗̇
+ 𝒲 𝐵𝐼⃗⃗⃗⃗⃗⃗⃗⃗⃗ × 𝒱𝑐
𝐵⃗⃗⃗⃗⃗⃗ 𝑎𝑛𝑑 𝑀⃗⃗⃗ = 𝐻 𝐵⃗⃗⃗⃗⃗⃗̇
+ 𝒲 𝐵𝐼⃗⃗⃗⃗⃗⃗⃗⃗⃗ × 𝐻 𝐵⃗⃗⃗⃗⃗⃗ (21)
With 𝒲 𝐵𝐼
= [
𝑃
𝑄
𝑅
] 𝑎𝑛𝑑 𝒱𝑐
𝐵
= [
𝑈
𝑉
𝑊
]
Where:𝐵 denotes the body axes and 𝒲 𝐵𝐼⃗⃗⃗⃗⃗⃗⃗⃗⃗indicates the angular vector from base 𝐼to base 𝐵.
By symmetry, we have𝐼 𝑥𝑦 = 𝐼 𝑦𝑧 = 0 𝑎𝑛𝑑 𝐻 𝐵
= [
𝐻 𝑥
𝐻 𝑦
𝐻𝑧
] = [
𝐻 𝑥𝑥 0 𝐻 𝑥𝑧
0 𝐻 𝑦𝑦 0
𝐻 𝑥𝑧 0 𝐻𝑧𝑧
] [
𝑃
𝑄
𝑅
]

50
The state description of the lateral motion of B747 was developed in the ref [5]. In this section we
will apply the results to the case of the Boeing 747 for an altitude of 12192m and a speed of
235.9m/s and assuming that Θ0 = 0.The state space model of its lateral motion (Lateral motion
of the B747) is given by the next state space equations:
[
𝑣̇
𝑝̇
𝑟̇
𝜙̇
]
=
[
−0.0558
−0.0127
0.0036
0
0
−0.4351
−0.0061
1
−235.9
0.4143
−0.1458
0
9.81
0
0
0
] [
𝑣
𝑝
𝑟
𝜙
]
+ [
0
−0.1433
0.0038
0
1.7188
0.1146
−0.4859
0
][
𝛿 𝑎
𝛿 𝑟
]
Where:𝑦1 = 𝑣:Lateral speed rate perturbation𝑦2 = 𝑝:Roll rate perturbation
𝑦3 = 𝑟:Yaw rate perturbation𝑦4 = 𝜙:Roll angle.
𝑢1 = 𝛿 𝑎:Ailerons deflection𝑢2 = 𝛿 𝑟:Rudder deflection
The dimension of the matrix 𝐴 is 4 × 4 and the number of inputs is2 . The rank of the matrix
[𝐵 𝐴𝐵]is 4, and then the system is block controllable of index 2. Therefore we can convert the
system into block controller form by the following transformation matrix𝑇𝑐:
𝑇𝑐 = [
[𝑂2 𝐼2][𝐵 𝐴𝐵]−1
[𝑂2 𝐼2][𝐵 𝐴𝐵]−1
𝐴
] =
[
0.0070
0.0088
−0.0003
−0.0004
0.0007
0.0008
−7.0203
−0.0543
0.0250
0.0312
−1.6573
−2.0723
−7.0199
−0.0538
0.0688
0.0860
]
We obtain the following:
𝐴 𝑐 = [
𝑂2 𝐼2
−𝐴 𝑐2 −𝐴 𝑐1
] =
[
(
0 0
0 0
) (
1 0
0 1
)
(
−0.0561 9.6066
0.0041 −0.7736
) (
−0.4589 1.7627
−0.0161 −0.1778
)
]
𝐵𝑐 =
[
𝑂2
𝐼2
]
=
[
(
0 0
0 0
)
(
1 0
0 1
)]
𝐶𝑐 = 𝐶 [
[𝑂2 𝐼2][𝐵 𝐴𝐵]−1
[𝑂2 𝐼2][𝐵 𝐴𝐵]−1
𝐴
]
−1
=
[
(
−0.8588 114.8295
0 0
) (
0 1.7188
−0.1433 0.1146
)
(
−0.0058 −0.0167
−0.1433 0.1146
) (
0.0038 −0.4859
0 0
)
]
We want to design a state feedback using block pole placement for the following set of desired
eigenvalues:−1 ± 1.5𝑖, −1.5 and − 2 or 𝜆1,2,3,4 = {−1 + 1.5𝑖 , −1 − 1.5𝑖 , −1.5 , −2}

51
4.1. State Feedback Design Using Right Block Poles
4.1.1. Right solvents in diagonal form
 Construction of the feedback gain matrix:
The desired right block poles in diagonal form are constructed as follows:
𝑅1 = (
−1 1.5
−1.5 −1
) , 𝑅2 = (
−1.5 0
0 −2
)
Then the corresponding controllable feedback gain matrix, which is obtained using the equation
(16), is:
𝐾𝑐 = [(
2.0576 6.8794
2.6632 0.4082
) (
2.4502 0.3990
1.7566 2.4131
)]
𝐾 = 𝐾𝑐 𝑇𝑐 ⇔ 𝐾 = [(
0.0738 −17.2161
0.0208 −12.4610
) − (
4.6217 14.6110
7.8327 18.3891
)]
The norm of the feedback gain matrix is‖𝐾‖2 = 32.5970.
 Time specifications:
The following points summarize the time specifications (Settling time𝑇𝑠and Maximum peak𝑀 𝑝)
for the closed-loop system described by equation (15 and 16) and the response of the system to an
initial state𝑥0 = [1 1 1 1] 𝑇
.
a. Lateral speed rate perturbation (𝑣) ∶ 𝑇𝑠 = 10.1587 s and 𝑀 𝑝 = 396
b. Roll rate perturbation(𝑝) ∶ 𝑇𝑠 = 4.2328 s and 𝑀 𝑝 = 1
c. Yaw rate perturbation(𝑟) ∶ 𝑇𝑠 = 4.9735 s and 𝑀 𝑝 = 2.43
d. Roll angle(𝜙): 𝑇𝑠 = 3.5979 s and 𝑀 𝑝 = 1.27
The settling time𝑇𝑠is defined in our case as the time required for the outputs to remain within
±0.05 unity.
 Robustness:
Robust stability:Robust stability is determined using the measures defined in the previous
section. First we find the norms of the left and right eigenvectors associated to each eigenvalue.
The right eigenvector matrix associated to the closed-loop system is:
𝑉 = [
1.0000
(−0.0029 + 0.0003𝑖)
(0.0040 − 0.0065𝑖)
(0.0011 + 0.0012𝑖)
1.0000
(−0.0029 − 0.0003𝑖)
(0.0040 + 0.0065𝑖)
(0.0011 − 0.0012𝑖)
−0.9575
0.2397
−0.0127
−0.1598
1.0000
−0.0021
0.0086
0.0010
] and‖𝑉‖ = 1.9844
The left eigenvector matrix has norm equal to:
‖𝑇‖ = ‖𝑉−1‖ = 1018.8 ⟹ 𝑠(𝑉) = ‖𝑉‖‖𝑉−1‖ = 2021.6

52
The norms of all the right eigenvectors are 1. The associated left eigenvectors have norms as
follows:‖𝒕 𝟏‖2 = 473.0835, ‖𝒕 𝟐‖2 = 473.0835, ‖𝒕 𝟑‖2 = 3.6939, ‖𝒕 𝟒‖2 = 774.7516.
The sensitivity of each eigenvalue is:𝑠(𝜆𝑖) = {473.0835 , 473.6939 , 3.6939 , 774.7516}
Now the stability measures are:
 𝑀1 = min0≤𝜔<∞{𝜎(𝐴 − 𝐵𝐾 − 𝑗𝜔𝐼)} = 0.0025
 𝑀2 = (𝑠(𝑉))
−1
|𝑅𝑒{𝜆1,2}| = 4.9464 × 10−4
 𝑀3 = min0≤𝑖≤4 {(𝑠(𝜆𝑖))
−1
|𝑅𝑒{𝜆𝑖}|} = 0.0023
Robust performance:We generate a random small perturbation using MATLAB software, we
get the following:
∆𝐴 =
[
0.4218
0.9157
0.7922
0.9595
0.6557
0.0357
0.8491
0.9340
0.6555
0.7577
0.7431
0.3922
0.6787
0.1712
0.7060
0.0318
]
× 10−4
The eigenvalues of the matrix(𝐴 − 𝐵𝐾 + ∆𝐴)are:
−1.0595 + 1.4677𝑖; −1.0595 − 1.4677𝑖; −1.4999; −1.8809.
The relative change of the each eigenvalue is given below by the following:
𝑟𝑖(𝜆𝑖) = {0.0751 , 0.0751 , 6.6667 × 10−4
, 0.0595}
4.1.2. Right solvents in controllable form
The desired right block poles in controllable form are constructed as follows:
𝑅1 = (
−2 −3.25
1 0
) , 𝑅2 = (
0 1
−3 −3.5
)
(16), is:
𝐾𝑐 = [(
4.1627 11.8410
−2.0584 0.44510
) (
2.2286 2.1689
−0.6411 2.6347
)]
𝐾 = 𝐾𝑐 𝑇𝑐 ⟺ 𝐾 = [(
0.1315 −15.7512
−0.0114 4.3567
) − (
7.7148 29.5186
4.4348 −14.6081
)]

53
.
b. Roll rate perturbation(𝑝) ∶ 𝑇𝑠 = 3.7037 s and 𝑀 𝑝 = 1.58
 Robustness:
𝑉 = [
1.0000
(0.0006 + 0.0053𝑖)
(0.0041 − 0.0066𝑖)
(0.0022 − 0.0019𝑖)
1.0000
(0.0006 − 0.0053𝑖)
(0.0041 + 0.0066𝑖)
(0.0022 + 0.0019𝑖)
1.0000
−0.0028
0.0064
0.0019
−1.0000
0.0033
−0.0086
−0.0017
]and‖𝑉‖ = 2
‖𝑇‖ = ‖𝑉−1‖ = 2221 ⟹ 𝑠(𝑉) = ‖𝑉‖‖𝑉−1‖ = 1238.7
follows: ‖𝒕 𝟏‖2 = 345.5263, ‖𝒕 𝟐‖2 = 345.5263, ‖𝒕 𝟑‖2 = 958.4084, ‖𝒕 𝟒‖2 = 875.6205.
The sensitivity of each eigenvalue is: 𝑠(𝜆𝑖) = {345.5263 , 345.5263 , 958.4084 , 875.6205}
 𝑀1 = min0≤𝜔<∞{𝜎(𝐴 − 𝐵𝐾 − 𝑗𝜔𝐼)} = 0.0031
 𝑀2 = (𝑠(𝑉))
−1
|𝑅𝑒{𝜆1,2}| = 8.0732 × 10−4
 𝑀3 = min0≤𝑖≤4 {(𝑠(𝜆𝑖))
−1
|𝑅𝑒{𝜆𝑖}|} = 0.0016
Robust performance:The eigenvalues of the matrix(𝐴 − 𝐵𝐾 + ∆𝐴)are:
−0.9594 + 1.5188𝑖; −0.9594 − 1.5188𝑖; −1.6042; −1.9770.
𝑟𝑖(𝜆𝑖) = {0.0248 , 0.0248 , 0.0695 , 0.0115}

54
4.1.3. Right solvents in observable form
The desired right block poles in observable form are constructed as follows:
𝑅1 = (
−2 1
−3.25 0
) , 𝑅2 = (
0 −3
1 −3.5
)
(16), is:
𝐾𝑐 = [(
3.9439 7.6066
2.0041 0.6639
) (
3.5411 0.7627
1.7964 1.3222
)]
𝐾 = 𝐾𝑐 𝑇𝑐 ⟺ 𝐾 = [(
0.0929 −24.8926
0.0188 −12.6812
) − (
7.1135 27.7859
5.6464 13.8672
)]
The following point summarizes the time specifications (Settling time𝑇𝑠and Maximum peak𝑀 𝑝)
.
 Robustness:
The right eigenvector matrix associated to the closed-loop system is
𝑉 = [
1.0000
(−0.0015 + 0.0004𝑖)
(0.0040 − 0.0065𝑖)
(0.0006 + 0.0006𝑖)
1.0000
(−0.0015 − 0.0004𝑖)
(0.0040 + 0.0065𝑖)
(0.0006 − 0.0006𝑖)
−1.0000
−0.0023
−0.0062
0.0016
1.0000
0.0018
0.0085
−0.0009
] and‖𝑉‖ = 2
‖𝑇‖ = ‖𝑉−1‖ = 2314.8 ⟹ 𝑠(𝑉) = ‖𝑉‖‖𝑉−1‖ = 4628.8
follows: ‖𝒕 𝟏‖2 = 393.0204, ‖𝒕 𝟐‖2 = 393.0204, ‖𝒕 𝟑‖2 = 1464.9, ‖𝒕 𝟒‖2 = 1725.8.

55
 𝑀1 = min0≤𝜔<∞{𝜎(𝐴 − 𝐵𝐾 − 𝑗𝜔𝐼)} = 0.0042
 𝑀2 = (𝑠(𝑉))
−1
|𝑅𝑒{𝜆1,2}| = 2.1604 × 10−4
 𝑀3 = min0≤𝑖≤4 {(𝑠(𝜆𝑖))
−1
|𝑅𝑒{𝜆𝑖}|} = 0.0010
−1.0240 + 1.4553𝑖; −1.0240 − 1.4553𝑖; −1.7260 − 0.2201𝑖; 1.7260 + 0.2201𝑖.
𝑟𝑖(𝜆𝑖) = {0.0281 , 0.0281 , 0.2103 , 0.1757}
4.2. State feedback design using left block poles
4.2.1. Left solvents in diagonal form
The desired left block poles in diagonal form are constructed as follows:
𝐿1 = (
−1 1.5
−1.5 −1
) , 𝐿2 = (
−1.5 0
0 −2
)
(16), is:
𝐾𝑐 = [(
2.0576 6.9475
2.7314 0.4082
) (
2.4502 −0.4502
1.3475 2.4131
)]
𝐾 = 𝐾𝑐 𝑇𝑐 ⇔ 𝐾 = [(
0.0746 −17.1938
0.0214 −9.5890
) − (
3.7718 14.6499
7.1530 18.8959
)]
.

56
 Robustness:
𝑉 = [
1.0000
(−0.0027 + 0.0004𝑖)
(0.0040 − 0.0065𝑖)
(0.0007 + 0.0014𝑖)
1.0000
(−0.0027 − 0.0004𝑖)
(0.0040 + 0.0065𝑖)
(0.0007 − 0.0014𝑖)
1.0000
−0.0003
0.0063
0.0002
−0.9999
0.0096
−0.0087
−0.0048
] and‖𝑉‖ = 2
‖𝑇‖ = ‖𝑉−1‖ = 1026.1 ⟹ 𝑠(𝑉) = ‖𝑉‖‖𝑉−1‖ = 2052.1
follows: ‖𝒕 𝟏‖2 = 496.8456, ‖𝒕 𝟐‖2 = 496.8456, ‖𝒕 𝟑‖2 = 737.9024, ‖𝒕 𝟒‖2 = 201.6491.
 𝑀1 = min0≤𝜔<∞{𝜎(𝐴 − 𝐵𝐾 − 𝑗𝜔𝐼)} = 0.0021
 𝑀2 = (𝑠(𝑉))
−1
|𝑅𝑒{𝜆1,2}| = 4.8731 × 10−4
 𝑀3 = min0≤𝑖≤4 {(𝑠(𝜆𝑖))
−1
|𝑅𝑒{𝜆𝑖}|} = 0.0020
−1.0665 + 1.4825𝑖; −1.0665 − 1.4825𝑖; −1.3861; −1.9808.
𝑟𝑖(𝜆𝑖) = {0.0381 , 0.0381 , 0.0759 , 0.0096}
4.2.2. Left solvents in controllable form
The desired left block poles in controller form are constructed as follows:
𝐿1 = (
−2 −3.25
1 0
) , 𝐿2 = (
0 1
−3 −3.5
)
(16), is:

57
𝐾𝑐 = [(
3.9439 11.6066
−1.9959 0.6639
) (
3.5411 3.5752
−1.0161 1.3222
)]
𝐾 = 𝐾𝑐 𝑇𝑐 ⇔ 𝐾 = [(
0.1269 −25.0421
−0.0084 7.0609
) − (
12.8170 27.7592
1.0850 −14.0189
)]
.
b. Roll rate perturbation(𝑝) ∶ 𝑇𝑠 = 4.4444 s and 𝑀 𝑝 = 1.57
 Robustness:
𝑉 = [
1.0000
(−0.0012 + 0.0034𝑖)
(0.0040 − 0.0066𝑖)
(0.0019 − 0.0005𝑖)
1.0000
(−0.0012 − 0.0034𝑖)
(0.0040 + 0.0066𝑖)
(0.0019 + 0.0005𝑖)
1.0000
0.0041
0.0062
−0.0027
−0.9368
0.3127
−0.0145
−0.1563
] and‖𝑉‖ = 1.9763
‖𝑇‖ = ‖𝑉−1‖ = 884.8156 ⟹ 𝑠(𝑉) = ‖𝑉‖‖𝑉−1‖ = 1784.7
follows: ‖𝒕 𝟏‖2 = 390.9862, ‖𝒕 𝟐‖2 = 390.9862, ‖𝒕 𝟑‖2 = 698.5479, ‖𝒕 𝟒‖2 = 16.5739.
 𝑀1 = min0≤𝜔<∞{𝜎(𝐴 − 𝐵𝐾 − 𝑗𝜔𝐼)} = 0.0026
 𝑀2 = (𝑠(𝑉))
−1
|𝑅𝑒{𝜆1,2}| = 5.7187 × 10−4
 𝑀3 = min0≤𝑖≤4 {(𝑠(𝜆𝑖))
−1
|𝑅𝑒{𝜆𝑖}|} = 0.0021

58
−0.9529 + 1.5250𝑖; −0.9529 − 1.2550𝑖; −1.5919; −2.0022.
𝑟𝑖(𝜆𝑖) = {0.0296 , 0.0296 , 0.0613 , 0.0011}
4.2.3. Left solvents in observable form
The desired left block poles in observable form are constructed as follows:
𝐿1 = (
−2 1
−3.25 0
) , 𝐿2 = (
0 −3
1 −3.5
)
(16), is:
𝐾𝑐 = [(
4.1627 7.5441
2.2385 0.4451
) (
2.2286 1.1377
0.3901 2.6347
)]
𝐾 = 𝐾𝑐 𝑇𝑐 ⇔ 𝐾 = [(
0.0942 −15.6987
0.0185 −2.8802
) − (
5.7119 29.3762
6.0367 15.4846
)]
.
 Robustness:
𝑉 = [
1.0000
(−0.0016 − 0.0021𝑖)
(0.0039 − 0.0065𝑖)
(−0.0005 + 0.0014𝑖)
1.0000
(−0.0016 + 0.0021𝑖)
(0.0039 + 0.0065𝑖)
(−0.0005 − 0.0004𝑖)
−1.0000
0.0007
−0.0063
−0.0004
1.0000
−0.0013
0.0086
0.0007
] and‖𝑉‖ = 2

59
‖𝑇‖ = ‖𝑉−1‖ = 1383.9 ⟹ 𝑠(𝑉) = ‖𝑉‖‖𝑉−1‖ = 2767.8
follows: ‖𝒕 𝟏‖2 = 402.7999, ‖𝒕 𝟐‖2 = 402.7999, ‖𝒕 𝟑‖2 = 1180.7, ‖𝒕 𝟒‖2 = 717.5067.
The sensitivity of each eigenvalue is:
𝑠(𝜆𝑖) = {402.7999 , 402.7999 , 1180.7 , 717.5067}
 𝑀1 = min0≤𝜔<∞{𝜎(𝐴 − 𝐵𝐾 − 𝑗𝜔𝐼)} = 0.0025
 𝑀2 = (𝑠(𝑉))
−1
|𝑅𝑒{𝜆1,2}| = 3.6129 × 10−4
 𝑀3 = min0≤𝑖≤4 {(𝑠(𝜆𝑖))
−1
|𝑅𝑒{𝜆𝑖}|} = 0.0013
−1.0497 + 1.4968𝑖; −1.0497 − 1.4968𝑖; −1.3717; −2.0288.
𝑟𝑖(𝜆𝑖) = {0.0276 , 0.0276 , 0.0855 , 0.0144}
4. DISCUSSION
Small gains are desirable because they minimize the control energy and prevent saturation of the
controller elements and noise amplification. For time specifications, the smaller the settling time
and maximum peak the better the time response. The right solvents in canonical forms give
smaller settling times. The maximum peaks are practically the same except the case of the lateral
velocity 𝑣where it is better in the case of right solvents.For the sensitivities of the eigenvalues, we
choose the one that has the lowest sensitivity taking into account the distance of the eigenvalue
from the 𝑗𝜔axis. For the robust stability the greater the value of its measure the more robustly
stable the system, where 𝑀3is more accurate than 𝑀1and 𝑀2[1]. For robust performance, the
smaller the value of relative change the better the performance. In our case, the crucial criterion is
the robustness, because of the linearization of the model.
6. COMPARISON OF THE RESULTS
In order to do a comparison study between left and right block roots placement we gather the
preceding results in table: 1as shown below.To make the analysis easier and be clear table: 1 can
be summarized into table: 2.
From table: 2the choice of the block roots in our case is as follows:

60
 In diagonal form: the right solvents are more suitable.
 In controllable form: The left solvents are suitable.
 In observable form: The left solvents are more suitable.
Table: 1 shows the collected data to be compared.
Diagonal form Controllable form Observable form
Right Left Right Left Right Left
‖𝐾‖2
32.5970 31.3234 37.0470 41.9890 42.7086 37.3973
Timespecifications
𝑡 𝑠
10.1587 𝑠 10.1587 𝑠 8.8889 𝑠 10.0529 𝑠 9.8413 𝑠 8.9947 𝑠
𝑣 Max
peak
396 424 346 407 205 302
𝑡 𝑠
4.2328 𝑠 4.1270 𝑠 3.7037 𝑠 4.4444 𝑠 2.6455 𝑠 3.8095 𝑠
𝑝 Max
peak
1 1 1.58 1.57 1 1
𝑡 𝑠
4.9735 𝑠 5.0794 𝑠 4.7619 𝑠 4.8677 𝑠 4.6561 𝑠 4.8677 𝑠
𝑟 Max
peak
2.43 2.44 2.06 2.46 1.49 1.69
𝑡 𝑠
3.5979 𝑠 3.4921 𝑠 3.0688 𝑠 3.9153 𝑠 2.1164 𝑠 3.1746 𝑠
𝜙 Max
peak
1.27 1.25 1.06 1.05 1.12 1.12
Robuststability
−1
+1.5𝑖
437.0835 496.8456 345.5263 390.9862 393.0204 402.7999
𝑠(𝜆𝑖) −1
−1.5𝑖
437.0835 496.8456 345.5263 390.9862 393.0204 402.7999
−1.5 3.6939 737.9024 958.4084 698.5479 1464.9 1180.7
−2 774.7516 201.6491 875.6205 698. 3.5979 3.5979
𝑀1
0.0025 0.0021 0.0031 0.0026 0.0042 0.0025
𝑀2(× 10−4) 4.9464 4.8731 8.0732 5.7187 2.1604 3.6129
𝑀3
0.0023 0.0020 0.0016 0.0021 0.0010 0.0013
Robustperformance
−1
+1.5𝑖
0.0751 0.0381 0.0248 0.0296 0.0281 0.0276
𝑟𝑖(𝜆𝑖) −1
+1.5𝑖
0.0751 0.0381 0.0248 0.0296 0.0281 0.0276
−1.5 6.6667
× 10−5
0.0759 0.0695 0.0613 0.2103 0.0855
−2 0.0595 0.0095 0.0115 0.0011 01757 0.0144

61
Table: 2 show the comparative study.
Diagonal form Controllable form Observable form
Right Left Right Left Right Left
‖𝐾‖2 Left Right Left
Time specifications Right Right Right
Robust
stability
𝑠(𝜆𝑖) Right Right Left
𝑀1
Right Left Left𝑀2
𝑀3
Robust
performance
𝑟𝑖(𝜆𝑖)
Left Left Left
7. CONCLUSION
The aim of the present work in this paper is to apply block pole placement technique to a multi-
input multi-output system with an application to a lateral motion of a Boeing 747, in order to
determine the best feedback gain matrix. It has been shown that this technique is due to the fact
that the characteristic matrix polynomial of the system has block poles as roots. These block poles
can be chosen to be right or left as specified before. Their forms are not unique, but we restricted
our study to the case of the canonical forms (diagonal, controllable and observable).
The choice of feedback gain matrix is done by the comparison between the two forms of the
block poles in terms of best response characteristics and system robustness. In our case we can
say that the choice between left and right solvents is made according to the form of these
solvents. We can conclude that the choice of the feedback gain matrix is made according to the
application itself from a desirable system’s response.
REFERENCES
[1] Chia-Chi Tsui, “Robust Control System Design: Advanced State Space Techniques”, Second Edition,
Marcel Dekker, 2004.
[2] Malika Yaici, KamelHariche, “On Eigenstructure Assignment Using Block Poles Placement”
European Journal Of Control, May 2014.
[3] Malika Yaici, KamelHariche, “On Solvents Of Matrix Polynomials” International Journal Of
Modeling And Optimization, Vol. 4, No. 4, August 2014.
[4] Katsuhiko Ogata, “Modern Control Engineering”, Third Edition, Prentice Hall, 1997.
[5] M.V. Cook, “Flight Dynamics Principles”, Second Edition, Butterworth-Heinemann, 2007.
[6] William L. Brogan, “Modern Control Theory”, Third Edition, Prentice Hall, 1990.
[7] K. Hariche And E. D. Denman” On Solvents And Lagrange Interpolating Λ -Matrices” Applied
Mathematics And Computation 25321-332 (1988).
[8] E. Periera, “On Solvents Of Matrix Polynomials” Appl. Numer. Math., Vol. 47, Pp. 197-208, 2003.
[9] E. Periera, “Block Eigenvalues And Solution Of Differential Matrix Equation” Mathematical Notes,
Miskolc, Vol. 4, No.1 (2003), Pp. 45-51.
[10] F. R. Gantmacher, Theory Of Matrices, New York: Chelsea, 1960.

62
[11] I. Gohberg, P. Lancaster, L. Rodman, Spectral Analysis Of Matrix Polynomials: I. Canonical Forms
And Divisors, Linear Algebra Appl., 20, Pp. 1-44, 1978.
[12] I. Gohberg, P. Lancaster, L. Rodman, Spectral Analysis Of Matrix Polynomials: II. The Resolvent
Form And Spectral Divisors, Linear Algebra Appl. 21, 1978.
[13] I. Gohberg, P. Lancaster, L. Rodman, Matrix Polynomials, Academic Press, 1982.
[14] P.Lancaster, Lambda-Matrices And Vibrating Systems, New York, Pergamon Press, 1966.
[15] P. Lancaster, M. Timenetski, “The Theory Of Matrices”, 2nd Ed., Academic Press, 1985.
[16] Leang S. Shieh, Y. T Tsay ”Transformation Of Solvents And Spectral Factors Of Matrix Polynomials
And Their Applications” Int J. Control. Vol. 34, Pp. 813-823, 1981.
[17] William S. Levine, “Control System Advanced Methods” CRC Press ,2011.
[18] Magdi S. Mahmoud And Yuanqing Xia, Applied Control Systems Design,Springer-Verlag London
Limited 2012.
[19] BelkacemBekhiti, A. Dahimene, B. Nail, K. Hariche, And A. Hamadouche, “On Block Roots Of
Matrix Polynomials Based MIMO Control System Design”, 4th International Conference On
Electrical Engineering ICEE Boumerdes , 2015.
[20] B. C. Moore, "On The Flexibility Offered By State Feedback In Multivariable Systems Beyond
Closed Loop Eigenvalue Assignment", IEEE Trans. Autom. Control, Vol AC-21, No.5,Pp. 689-692,
1976.
[21] J. Lu, H. D. Chiang And J. S. Thorp, "Eigenstructure Assignment By Decentralized Feedback
Control", IEEE Trans. Autom. Control, Vol. AC-38, No.4, Pp. 587-594, 1993.
[22] T. Clarke, S. J. Griffin And J. Ensor, "Output Feedback Eigenstructure Assignment Using A New
Reduced Orthogonality Condition", Int. J. Control, Vol. 76, No.4, Pp. 390-402, 2002.
[23] O. Bachelier, J. BoscheAnd D. Mehdi, On Pole Placement Via Eigenstructure Assignment Approach,
IEEE Trans. Autom. Control, Vol. AC-51, No.9, Pp. 1554-1558, 2006.
[24] A. Pomfret And T. Clarke, "An Extension To Output-Feedback Eigenstructure Assignment:
Structuring Controllers By Exploiting Unused Design Freedom", Int. J. Control, Vol.82, No.2, Pp.
207-216, 2009.
[25] A. N. Andry, E. Y. Shapiro And J. C. Chung, "Eigenstructure Assignment For Linear Systems", IEEE
Trans. Aerosp. Electron.Syst., Vol. 19, No.5, Pp. 711-729, 1983.
[26] K. M. Sobel And E.J. Lallman, "Eigenstructure Assignment For The Control Of Highly Augmented
Aircraft", J. Of Guidance, Control, And Dynamics, Vol. 12, No.3, Pp. 18-324, 1989.
[27] J. -F. Magni, "MultimodelEigenstructure Assignment In Flight-Control Design", Aerosp. Sci.
Technol., Vol.3, Pp. 141-151, 1999.

THE LEFT AND RIGHT BLOCK POLE PLACEMENT COMPARISON STUDY: APPLICATION TO FLIGHT DYNAMICS

More Related Content

What's hot (16)

Similar to THE LEFT AND RIGHT BLOCK POLE PLACEMENT COMPARISON STUDY: APPLICATION TO FLIGHT DYNAMICS (20)

More from ieijjournal1 (17)

Recently uploaded (20)

THE LEFT AND RIGHT BLOCK POLE PLACEMENT COMPARISON STUDY: APPLICATION TO FLIGHT DYNAMICS