International Journal of Control and Automation
Vol.7, No.2 (2014), pp.309-326
http://dx.doi.org/10.14257/ijca.2014.7.2.28
The Nonlinear Modeling and Control movement of a Human
Forearm for Prosthetic Applications
Gamal A. Elnashar
L.
Automatic Control Centre, School of Engineering-Egyptian Armed Forces,
15 Yassien Raghb St., Nasr City, Cairo, Egypt
m Onl
ad in
eb eV
y er
th si
is on
fil O
ei n
s I l y.
LL
EG
Abstract
A
enjygamal@gmail.com
The model and control of a human forearm is analyzed. In this work, we study the problem
of human hand control carrying a mass. The equation of motion and the natural frequency of
the forearm for small angular displacement are derived. We develop new methods that use
vector fields in the controller construction for a set of nonlinear dynamical systems. The
paper deals with compensate of non-linear system which has a similar idea as the method
mentioned in linear system. The nonlinear control design procedure, as in the case of linear
systems, involves three steps. The first step is the devise of a state-feedback control law, the
second step involves the design of a state estimator, and the third step merges the first two
steps to obtain a collective controller–estimator compensator. We have managed to design a
control law for the non-linear representation of a system, in such a way that the
representation of a closed loop system is affine, controllable, and observable and a closed
loop system is asymptotically stable. Throughout any motion, the forearm can be considered
a one-link robot manipulator which could be exploited to benefit people with disabilities
(missing extremities).
Keywords: Forearm; Nonlinear Control; Modeling; Observer, Vector Field; Lie
Derivatives
1. Introduction
Bo
ok
The problems associated with the design of artificial arm replacements are far more
challenging than those associated with the design of robotic arms or terminal devices. The
design of artificial arms is a multidisciplinary effort. The design team needs an understanding
of the mechanics of mechanisms, such as gears, levers, and points of mechanical advantage,
and electromechanical design, such as switches, dc motors, and electronics [1]. When the
someone wants to move the arm, the brain sends signals that first bond the chest muscles,
which send an electrical signal to the prosthetic arm, instructing it to reposition. The
procedure requires no more aware effort than it would for a person who has a ordinary arm.
Typically, a person with a prosthetic arm can make only a few motions, often so slowly that
many people use the arms only for limited activities. There is a separate motor for each
movement. By far the most common actuator for electrically powered prostheses is the
permanent magnet dc electric motor with some form of transmission [2]. In proportional
control, the amount/intensity of a controlled output variable is directly related (proportional)
to the amount of the input signal. For example, the output speed of a dc motor is proportional
to the amount of voltage applied to its terminals. This is why dc motors are said to be speed
controlled. This is also the reason why most of today’s commercially available prosthetic
ISSN: 2005-4297 IJCA
Copyright ⓒ 2014 SERSC
International Journal of Control and Automation
Vol.7, No.2 (2014)
ok
m Onl
ad in
eb eV
y er
th si
is on
fil O
ei n
s I l y.
LL
EG
A
L.
components are speed controlled—it is simple [3]. Output speed is proportional to the amount
of input signal. Proportional control is used where a graded response to a graded input is
required. In position control the position of the prosthetic joint is proportional to the input
amount/intensity. The input amount/intensity might be the position of another physiological
joint or a force level. If the position of another joint is used as the input then the system is
known as a position actuated, position servomechanism. If the amount of force applied by
some body part is the input, then the system is a force actuated, position servomechanism [4].
With position control the amputee’s ability to perceive and control prosthesis position is
directly determined by his or her ability to perceive and control the input signal. A major
disadvantage of position control is that, unlike velocity control, it must maintain an input
signal to hold an output level other than zero. This means that power must be continuously
supplied to the component to maintain a commanded position other than zero. This is one of
the reasons why speed or velocity control is the dominant mode of control in externallypowered prosthetics today, despite the fact that it has been shown that position control for
positioning of the terminal device in space is superior to velocity control. In control theory,
a state observer is a system that provides an estimate of the internal state of a given real
system, from measurements of the input and output of the real system [6]. It is typically
computer-implemented, and provides the basis of many practical applications. Knowing the
system state is necessary to solve many control theory problems; for example, stabilizing a
system using state feedback. In most practical cases, the physical state of the system cannot
be determined by direct observation. Instead, indirect effects of the internal state are observed
by way of the system outputs. If a system is observable, it is possible to fully reconstruct the
system state from its output measurements using the state observer. In this day of digital
circuits and microprocessor based controllers pulse width modulation is the preferred method
of supplying a graded (proportional) control signal to a component. A PWM stream only
requires a single digital output line and a counter on the microprocessor to be implemented,
whereas a conventional analog signal (linear dc voltage level) requires a full digital-to-analog
(D/A) converter. PWM techniques are used extensively in switched-mode power supply
design and audio amplifiers and as such, there is a large array of resources available to the
designer to choose from [8].
In this study, a model for a forearm performing a motion is presented, using a new
controller technique based on vector fields. Furthermore, we evaluated three position
controllers. The forearm bar of mass m 1 and length b is shown in Figure 1. A mass m 2 is
carried by the angular of the forearm of a human hand. During motion, the forearm can be
considered to rotate about the joint (pivot point O) with muscle forcers modeled in the form
Bo
of a force by triceps c1x and a force in biceps c 2 , where c1 and c 2 are constant and
x is the velocity with which triceps are stretched (or contracted ). We will derive the
equation of motion and natural frequency of the forearm of the forearm for small angular
displacement . The paper is organized as follows: section 1 describes an introduction about
prosthetic research and its control techniques. In Section 2 the mathematical model of a
human forearm is described i.e. equation of motion for the angular motion of the forearm
about the pivot point O is derived, and the motion of the robot arm by a DC motor via a gear
is resulting Section 3 develops a method for constructing state-feedback stabilizing controls
law for a class of dynamical systems where position control algorithms are treated. Sections 4
and 5 nonlinear state-feedback controller and asymptotic state estimator are derived
respectively for one-link manipulator model. Nonlinear combined controller-estimator
compensator is simulated in Section 6 and Finally, A short conclusion in Section 7
summarizes the study.
310
Copyright ⓒ 2014 SERSC
International Journal of Control and Automation
Vol.7, No.2 (2014)
2. Mathematical Modeling
Equation of motion for the angular motion of the forearm about the pivot point O [9]:
b
(1)
cos P F2a2 F1a1 0
2
Where P the angular displacement of the forearm is, I 0 is the mass of inertia of the forearm
I 0P m 2 gb cos P m1g
and the mass carried:
1
I 0 m 2b 2 b 2 m1
3
A
And the forces in the biceps and triceps muscles ( F2 and F1 are given by
L.
(2)
m Onl
ad in
eb eV
y er
th si
is on
fil O
ei n
s I l y.
LL
EG
F2 c 2P
F1 c1x c1a1P
(3)
(4)
Where the linear velocity of the triceps can be expressed as
x a1P
(5)
Using Equations (2)-(4), equation (1) can be rewritten as
1
I 0 P m 2 gb m1gb cos P c 2a2 P c1a12 P 0
2
(6)
Let the forearm undergo small angular displacement about the static equilibrium
position, , so that
(7)
Bo
ok
P
Figure 1. Forearm of a human hand carrying a mass
Copyright ⓒ 2014 SERSC
311
International Journal of Control and Automation
Vol.7, No.2 (2014)
Using Taylor’s series expansion of cos P about P , the static equilibrium position, can be
expressed as (for small value of )
cos P cos cos sin
Using
(8)
P and P , Equation (6) can be expresses as
1
I 0 P m 2 gb m1gb cos sin c 2a2 c1a12 0
2
L.
Or
m Onl
ad in
eb eV
y er
th si
is on
fil O
ei n
s I l y.
LL
EG
A
1
1
I 0 m 2 gb m1gb cos sin m 2 gb m1gb c 2a2 c 2a2 c1a12 0 (9)
2
2
Notice that the static equilibrium equation of the forearm at P , Eq.(6) is given by
1
m 2 gb m1gb cos c 2a2 0
2
(10)
In view of Equation (10), Equation (9) becomes
1 2
1
2
2
m 2b b m1 c1a1 c 2a2 sin gb m 2 m1 0
3
2
(11)
which denotes the equation of motion of the forearm.
The undamped natural frequency of the forearm can be expressed as:
1
c 2a2 sin gb m 2 m1
2
n
1
b 2 m 2 m1
3
(12)
Bo
ok
The design of fully functioning artificial arms with physiological speeds-of- response and
strength (or better) that can be controlled almost without thought is the goal of upper
extremity prosthetics research. Unfortunately, current prosthetic components and interface
techniques are still a long way from realizing this goal [1]. By far the most common actuator
for electrically powered prostheses is the permanent magnet dc electric motor with some form
of transmission. While there is much research into other electrically powered actuator
technologies, such as shape memory alloys and electro active polymers, none is to the point
where it can compete against the dc electric motor.
312
Copyright ⓒ 2014 SERSC
A
Figure 3. Schematic of an armature
controlled DC-Motor
m Onl
ad in
eb eV
y er
th si
is on
fil O
ei n
s I l y.
LL
EG
Figure 2. A manipulator of length b1
and mass m1 controlled by a DC
motor via a gear.
L.
International Journal of Control and Automation
Vol.7, No.2 (2014)
In terms of artificial arm, consider a model of one-link robot manipulator shown in Figure
2. Many different types of drive mechanisms have been devised to allow wrist and forearm
drive motors and gearboxes to be mounted close to the first and second axis of rotation, thus
minimizing the extended mass of the arm. The motion of the robot arm is controlled by a DC
motor via a gear [10]. The DC motor is armature-controlled and its schematic is shown in
Figure 3.The torque delivered by the motor is: T m K m i a , where k m is the motor-torque
constant, and i a is the armature current. Let N denote the gear ratio. Then we have
P radius of motor gear Number of teeth motor gear 1
radius of arm gear
Number of teeth arm gear
N
m
The work done by gears are proportional to their number of teeth and the work done by the
gears must be equal. Let T P denote the torque applied to the robot arm. Then,
T PP T mm . Thus, the torque applied the rod is T P NT m NK m i a . We use Newton’s
second law to write the equation modeling the arm dynamics,
(13)
ok
1 2
1
2
m 2b b m1 P m 2 gb m1gb cos P NK m i a
3
2
Bo
Where g=9.8 m/sec2 is the gravitational constant. Applying Kirchhoff’s voltage law to the
armature circuit yields
La
di a
d P
Rai a k b N
u
dt
dt
(14)
Where k P is the back emf constant. We can now construct a third-order state-space model of
the one-link robot. For this we choose the following state variables:
x 1 P , x 2
d P
P , x 3 i a ,
dt
Copyright ⓒ 2014 SERSC
313
International Journal of Control and Automation
Vol.7, No.2 (2014)
Then, the model in state-space format is
A
L.
(15)
m Onl
ad in
eb eV
y er
th si
is on
fil O
ei n
s I l y.
LL
EG
x2
1
x 1 m 2 gb m1 gb
0
Nk m
2
cos x 1
x 0 u ,
x 2
1 2
1 2 3
2
2
m b b m1
m 2b b m1 1
x 3 2
3
3
La
kbN
R
x2 x3
La
La
Reasonable parameters of the robot are: m2=5kg, m1=2kg, b=30cm, N=10, km=0.1Nm/A,
kb=0.1 V sec/rad, Ra=1Ω, La=100mH. Then the robot model takes the form:
x 1
x2
0
x 2 34.59 cos x 1 1.96x 3 0 u
x 3 10x 2 10x 3
10
(16)
We assume that the output y, is
y x1
(17)
Uncontrolled one-link robot
3
2
1
-1
Bo
ok
x1 (rad)
0
-2
-3
-4
-5
-6
0
1
2
3
4
5
6
7
8
Time (sec)
Figure 4. Plots of the one-link manipulator’s, versus time for three different
initial angles (0)
314
Copyright ⓒ 2014 SERSC
International Journal of Control and Automation
Vol.7, No.2 (2014)
Time histories of state trajectories of the uncontrolled nonlinear system of the model by
(16) and (17) [ x 1 (t ) (t ) ] versus time is shown in Figure 4. The manipulator is driven
by u 0 , initial conditions x 1 (0) 1, 1 and 2 and x 2 (0) 0 and x 3 (0) 0 . It is clear
that drastic change response due to the initial conditions and also the system has more than
one equilibrium points.
3. Nonlinear Control
L.
One of the objective of this paper is to devise a method for constructing state-feedback
stabilizing control law for a class of dynamical nonlinear systems modeled by
m Onl
ad in
eb eV
y er
th si
is on
fil O
ei n
s I l y.
LL
EG
y h x
A
x f (x ) G (x )u ,
(18)
where f : R n R n ,G : R n R nxm and the output map h : R n R p
We thus first discuss a method for reducing a nonlinear system model into an equivalent
form that is a generalization of the controller form known from linear system theory. In our
subsequent discussion, we will be using three types of Lie derivatives [11, 12]. They are as
follows:
1. Derivative of a vector field with respect to a vector field, also known as the Lie bracket.
Given the vector-valued functions f : R n R n and g : R n R n , where f and g are
C vector fields, their Lie bracket is defined as
f , g
f
g
g
f
x
x
(19)
2. Derivative of a function with respect to a vector field. Let let h : R n R be a C
function on R n . Let Dh hT , where h is the gradient (a column vector) of h with
respect to x . Then, the Lie derivative of the function h with respect to the vector field f,
denoted Lf h or Lf (h ) , is defined as
Dh f
h
h
h
f1
f 1
fn
x 1
x 1
x n
(20)
Bo
ok
Lf h Lf h h , f
3. Derivative of Dh with respect to the vector field. The Lie derivative of Dh with respect
to the vector field f , denoted Lf (Dh ) , is defined as
f
h
DLf h LTf h
Lf Dh
f Dh
x
x
T
(21)
Our goal now is to construct a C∞ state variable transformation z T (x ), T 0 0
for which there is a C inverse x T
the new coordinates has the form
Copyright ⓒ 2014 SERSC
1
(z ) , such that system model x f (x ) g (x )u in
315
International Journal of Control and Automation
Vol.7, No.2 (2014)
z2
z 1
z3
z2
zn
z n 1
z f n (z , z ,
n
1
2
0
0
u
0
, z n ) 1
(22)
L.
A transformation z T (x ) such that T (0) 0 and for which there is a C inverse
T
, Lnf 2T1 , Lnf 1T
m Onl
ad in
eb eV
y er
th si
is on
fil O
ei n
s I l y.
LL
EG
T T1 , Lf T1 ,
A
x = T−1(z) is called a diffeomorphism[13]. The transformation z T (x ) has the form
(23)
The above means that the problem of constructing the desired transformation z T (x ) is
reduced to finding its first componentT 1 . The remaining components of T can be successively
computed using T 1 .
The row vector
T1 (x )
is the last row of the inverse of the
x
controllability matrix, provided the controllability matrix is invertible. We denote the last row
of the inverse of the controllability matrix by q(x). Then, the problem we have to solve is to
find T1 R n R such that
T1 (x )
q (x ), T1 (0) 0
x
where the controllability matrix of the nonlinear system can be expressed in terms of Lie
brackets as:
Q ad 0f , g
ad f , g
1
ad
f , g
n 1
(24)
The controllability matrix of the system model in Equations (16, 17) is
Q ad f , g
ad f , g
1
0
19.6
0
ad f , g 0 19.6 196
10 100 804
2
(25)
ok
0
The above controllability is of full rank on R3. The last row of its inverse is
Bo
q 0.051 0 0 , Hence T1 0.05x 1
Having obtainedT 1 , we construct the desired transformation Z T (x ) , where
T1
T (x ) L f T 1
L f L f T
316
0.051x 1
0.051x 2
1.77 cos x 1 0.1x 3
(26)
Copyright ⓒ 2014 SERSC
International Journal of Control and Automation
Vol.7, No.2 (2014)
Note that the inverse transformation x T
1
(z ) exist and has the form
19.6z 1
T (z )
19.6z 2
10z 3 17.7 cos(19.6z 1 )
(27)
0
0
0.05
T
0
0.05 0
Further more
x
1.77 sin x 1
0
0.1
(28)
A
L.
1
z
m Onl
ad in
eb eV
y er
th si
is on
fil O
ei n
s I l y.
LL
EG
Applying the transformation Z T (x ) to the model of the one-link robot manipulator
yields
T
T
f (x )
g (x )
u
x
x
x T 1 ( z )
x T 1 ( z )
z2
0
z
z3
0 u
34z 2 sin19.6z 1 19.6z 2 10z 3 17.7 cos(19.6z 1 ) 1
(29)
4. Nonlinear State-feedback Control
Once the plant model is transformed into the controller form, we can construct a statefeedback controller in the new coordinates and then, using the inverse transformation,
represent the controller in the original coordinates [14]. While constructing the controller in
the new coordinates, a part of the controller is used to cancel nonlinearities, thus resulting in a
linear system in the new coordinates. Then, we proceed to construct the other part of the
controller. This part can be designed using linear control methods because the feedback
linearized system is linear. The controller form of the one-link manipulator model is given by
(29). It is easy to design a stabilizing state-feedback controller in the new coordinates. It takes
the form
(30)
Bo
ok
u f 3 (z 1 , z 2 , z 3 ) k 1z 1 k 2z 2 k 3z 3 ,
Where f 3 (z 1 , z 2 , z 3 ) 34z 2 sin19.6z 1 19.6z 2 10z 3 17.7 cos(19.6z 1 )
Suppose that the desired closed-loop poles of the feedback linearized one-link manipulator
are to be located at 2 j 3.46,8 which verify a damping factor 0.5 and natural
frequency n 4 rad / sec . Then, the linear feedback gains k i i 1, 2,3 that shift the
poles to these desired locations are k 1 128, k 2 48 , k 3 12 .
Applying (30) with the above values of the linear feedback gains to the model gives
Copyright ⓒ 2014 SERSC
317
International Journal of Control and Automation
Vol.7, No.2 (2014)
1
0
0
z 0
0
1 z
128 48 12
(31)
Next, applying the inverse transformation yields the controller in the original coordinates,
u 6.528x 1 1.428x 2 0.2x 3 21.24cos x 1 1.81x 2 sin x 1
(32)
In Figure 5, three plots of the manipulator’s link angle, x 1 , versus time are presented.
0 0 . The control law applied to the
T
L.
The initial conditions have the form x (0) (0)
m Onl
ad in
eb eV
y er
th si
is on
fil O
ei n
s I l y.
LL
EG
A
manipulator is given by (32).
One-link robot driven by the linearizing state-feedback controller
2
x1 (rad)
1.5
1
0.5
0
-0.5
0
1
2
3
4
5
Time (sec)
Figure 5. Plots of the one-link manipulator’s link angle, x 1 , versus time for
Bo
ok
three different initial angles (0) . The manipulator is driven by the statefeedback control law (32)
5. Nonlinear Observer Form
In this section we discuss the problem of transforming a class of nonlinear system models
into an observer form, which is then used to construct state estimators for these systems [1517]. We consider a nonlinear system model of the form of equation (18). We desire to find a
state variable transformation, written as x T (z ) , such that the model in the z coordinates is
z Az ( y ) f (z ),
318
y 0 0
0 1 z cz
(33)
Copyright ⓒ 2014 SERSC
International Journal of Control and Automation
Vol.7, No.2 (2014)
It is easy to verify that the state z of the dynamical system
z (A LC )z ( y ) Ly
(34)
Will asymptotically converge to the state z of (33) if the matrix A−Lc is asymptotically
stable.
This is because the dynamics of the error, e z z are described by
e (A LC )e ,
e (t 0 ) z (t 0 ) z (t 0 )
L.
(35)
T T
z
f (z )
z
z
m Onl
ad in
eb eV
y er
th si
is on
fil O
ei n
s I l y.
LL
EG
x
A
Taking the derivative of x T (z ) with respect to time yields
where we represent the Jacobian matrix of T(z), denoted
T
T
DT
z
z 1
T
, as
z
T
z n
(36)
(37)
With the help of Lie-derivative notation introduced above,
T
T
f ,
z k 1 z k
1 T
ad f ,
z k
,
k 1, 2,, n 1
(38)
We express all the columns of the Jacobian matrix (37) of T (z ) in terms of the starting
vector
T
, that is,
z 1
ok
T 0 T 1 T
ad f ,
ad f ,
z
z 1
z 1
n 1 T
ad f ,
z 1
Bo
To obtain an expression for the starting vector
(39)
T
, we use the output equation
z 1
y h (x ) z n
(40)
We use the chain rule when taking the partial derivative of (40) with respect to z to get
h (x ) T
0 0
x z
Copyright ⓒ 2014 SERSC
0 1
(41)
319
International Journal of Control and Automation
Vol.7, No.2 (2014)
We utilize derivative of Dh with respect to the vector field to express the first component
on the left-hand side of the above as
h (x ) T
T
L0f Dh
0
x z 1
z 1
(42)
By repeated application of the Leibniz formula and (39), we express (41) equivalently as
Lnf 2 Dh Lnf 1 Dh
T
T
0 0
z 1
0 1
T
(43)
L.
L0f Dh L1f Dh
A
The first term of the left hand side is called the observability matrix of the nonlinear
T
is equal to the last
z 1
m Onl
ad in
eb eV
y er
th si
is on
fil O
ei n
s I l y.
LL
EG
system model (18). It follows from (43) that the starting vector
column of the inverse of the observability matrix. The observability matrix for the system is
L0f Dh
0
1
0
1
0
0 1
L f Dh
L2f Dh 34.59sin x 1 0 1.96
(44)
The starting vector T z 1 is the last column of the inverse of the above observability
matrix,
T
T
0 0 0.51
z 1
(45)
We will now find the Jacobian matrix of the desired transformation:
(46)
ok
0
1
0
T 0 T 1 T 2 T
ad f ,
1 10
ad f ,
ad f ,
0
z
z 1
z 1
z 1
0.51 5 40
Bo
Therefore, we can take
0
1
0
19.6 9.8 1.96
x T (z ) 0
1 10 z and z T 1 (x ) 10
1
0 x
0
0
0.51 5 40
1
(47)
The system in the new coordinates has the form
1
1
T
T
z
u Az ( y ,u )
f (x )
g (x )
z
z
x T ( z )
x T ( z )
y cz
320
Copyright ⓒ 2014 SERSC
International Journal of Control and Automation
Vol.7, No.2 (2014)
Performing simple manipulations yields
0
z 1
0
y 0
0 0
339.1cos y 19.6u
0 0 z 34.59 cos y 19.6 y
1 0
10 y
0 1 z z 3
(48)
m Onl
ad in
eb eV
y er
th si
is on
fil O
ei n
s I l y.
LL
EG
z A LC z Ly y ,u
A
L.
Once the nonlinear system model (18) is in the observer form, we can proceed with the
construction of an asymptotic state estimator using the technique from the linear systems. The
state estimator dynamics in the new coordinates are
(49)
where the estimator gain vector L is chosen so that the matrix (A−Lc) has its eigenvalues in
the desired locations in the open left-hand complex plane. It is easy to verify that the state
estimation error, z z satisfies the linear differential equation
d
z z A Lc z z
dt
(50)
Applying the inverse transformation to (49), we obtain the state estimator dynamics in the
original coordinates. We illustrate the above method of constructing a state estimator using
the model of the one-link manipulator from the previous model. We first construct a state
estimator for the one-link manipulator model in the observer form given by (48). Suppose that
we wish the eigenvalues of the state estimator matrix, A − Lc, to be located at {−9,−10,−11}.
Then, the estimator gain vector is L= [ 990 299 30 ]T .With the above choice of the state
estimator design parameters, its dynamics in the new coordinates become
0 990
990
339.1cos y 19.6u
0 299 z 299 y 34.59 cos y 19.6 y ,
30
1 30
10 y
0 1 z z 3
(51)
ok
0
z 1
0
y 0
Bo
Using transformations (47), we represent the state estimator dynamics in the original
coordinates,
10 y
1
0
20
30
x 79.4 0.196 1.96 x 1 y 34.59 cos y 80.4 y ,
92
210
302 y 10u
9.01 9.8
Copyright ⓒ 2014 SERSC
(52)
321
International Journal of Control and Automation
Vol.7, No.2 (2014)
Performing simple manipulations, we represent the above state estimator as
1
0
0
0
20
0
x 0
0
1.96 x 34.59 cos y 79.4 y y 0 u ,
0 9.1 9.8
92
10
0
y 1 0 0 x x 1
(53)
A
6. Nonlinear Combined Controller–estimator Compensator
L.
y 1 0 0 x x 1
m Onl
ad in
eb eV
y er
th si
is on
fil O
ei n
s I l y.
LL
EG
The designed state feedback controller is singularity free, and guarantees asymptotic
tracking of smooth reference trajectories for the speed of the motor under time -varying
load torque and rotor resistance uncertainty, for any initial condition [20]. Having
constructed a state estimator, we use the state estimates, rather than the states
themselves, in the state-feedback control law implementation. The question now is how
the incorporation of the state estimator affects the stability of the closed -loop system
with the combined estimator– controller compensator in the loop. Recall that in the case
of linear systems, we were able to show that the closed-loop system was asymptotically
stable. In the present case, one has to be careful when analyzing the stability of the
closed-loop system with the combined estimator–controller compensator in the loop.
The plant’s nonlinearities may cause difficulties in stabilizing the closed -loop system
using the combined controller–estimator compensator [21]. However, s for which the
Lipschitz condition is satisfied—one should be able to find stability conditions in terms
of the on linearity’s Lipschitz constant and the location of the estimator’s poles rather
easily. We now illustrate the implementation of the combined estimator –controller
compensator on the one-link manipulator model., the controller is
u 6.528x 1 1.428x 2 0.2x 3 21.24cos x 1 1.81x 2 sin x 1
(54)
Bo
ok
In Figure 6, plots of the one-link manipulator output for three different initial conditions
are shown. The manipulator is driven by the combined controller-estimator compensator
consisting of the state estimator (53) and the control law (54). The initial conditions of the
estimator were set to zero while the initial conditions of the plant were the same as when we
implemented the state-feedback control alone. An observer designed using Lie algebraic
methods is valid in any region where a state transformation to (51) can be found. A state
observer typically combines system input/output with a mathematical model to predict the
behavior of that system. Estimated x 1 (t ) versus time with compensator in the loop and zero
initial conditions is shown in Figure 7. Comparing plots of Figures 5, 6 and 7, the results
show that, state observer is much alike in dealing with nonlinear system. This technique is
often successful for solving real-world control problems.
322
Copyright ⓒ 2014 SERSC
International Journal of Control and Automation
Vol.7, No.2 (2014)
x1 versus time with the compensator in the loop
2
1.5
x1 (rad)
1
A
L.
0.5
m Onl
ad in
eb eV
y er
th si
is on
fil O
ei n
s I l y.
LL
EG
0
-0.5
0
1
2
3
4
5
Time (sec)
Figure 6. Plots of the one-link manipulator’s link angle, x 1 , versus time for
three different initial angles (0) . The manipulator is driven by the combined
controller–estimator compensator
Estimated x1 versus time with the compensator in the loop
2
Estimated x1 (rad)
1.5
1
0.5
ok
0
Bo
-0.5
0
1
2
3
4
5
Time (sec)
Figure 7. Plots of the one-link manipulator’s link angle, x 1 , versus time for
three different initial angles (0) . The manipulator is driven by the combined
controller–estimator compensator
7. Conclusions
An analysis and design of fully functioning artificial arms with speeds-of response and
strength is conducted. The equation of motion and natural frequency of a human forearm is
derived. Synthesis of control law for non-linear systems based on vector fields has been
successfully solved. The method is exact and does not require any system linearization. We
have managed to design a control law for the non-linear representation of a system, which is
Copyright ⓒ 2014 SERSC
323
International Journal of Control and Automation
Vol.7, No.2 (2014)
m Onl
ad in
eb eV
y er
th si
is on
fil O
ei n
s I l y.
LL
EG
References
A
L.
controllable and observable, in such a way that the representation of a closed loop system is
affine, controllable, observable and asymptotically stable. This fact gives us more flexibility
in the choice of the desired behavior of a closed loop system than a linear one. We presented
an introduction to qualitative theory of a nonlinear control system, with the main emphasis on
controllability and observabilty properties of such systems. We introduced the differential
geometric language of vector fields and Lie bracket. We explored Lie-algebraic techniques
for nonlinear observer design. The advantage of these techniques, were that they attempt to
exploit our knowledge of linear observer design by reducing a nonlinear observer problem to
one that can be handled by linear techniques. We illustrated our considerations with forearm
nonlinear system. The combination of the nonlinear observer and the nonlinear controller
stabilizes the system and guarantees exponential convergence of the tracking error to zero.
Bo
ok
[1] R. F. f. Weir, “Design of artificial arms and hands for prosthetic applications”, McGraw-Hill Companies,
(2004).
[2] R. F. f. Weir, C. W. Heckathorne and D. S. Childress, “Cineplasty as a Control Input for Externally Powered
Prosthetic Components”, Journal of Rehabilitation Research and Development, vol. 38, no. 4, (2001)
July/August, pp. 357–363.
[3] W. H. Yeadon and A. W. Yeadon, “Handbook of Small Electric Motors”, McGraw-Hill, New York, (2001).
[4] G. F. Franklin, J. D. Powell and A. Emami-Naeini, “Feedback Control of Dynamic Systems (5th ed.)”,
Pearson Prentice Hall, Upper Saddle River, NJ, (2006).
[5] G. J. Bakker and H. Nijmeijer, “Comments on Control design for multivariable nonlinear time-varying
systems”, International Journal of Control, vol. 41, no. 6, (1985), pp. 1627–1629.
[6] N. S. Nise, “Control Systems Engineering (3rd ed.)”, Wiley, New York, (2000).
[7] K. Ogata, “System Dynamics (4th ed.)”, Pearson Prentice Hall, Upper Saddle River, NJ, (2004).
[8] C. Altafini, “Geometric Control Methods for Nonlinear Systems and Robotic Applications”, Doctoral Thesis,
Stockholm, (2001) May.
[9] S. S. Rao, “Mechanical vibrations”, 5th ed., Prentice Hall Upper Saddle River, NJ, (2011).
[10] A. Isidori, “Nonlinear Control Systems”, third edition, Springer-Verlag, London, (1995).
[11] R. Marino and P. Tomei, “Nonlinear Control Design: Geometric, Adaptive and Robust”, Prentice-Hall
Europe, London, (1995).
[12] R. Su, “On the linear equivalents of nonlinear systems”, Systems & Control Letters, vol. 2, no. 1, (1982) July,
pp. 48–52.
[13] L. R. Hunt, R. Su and G. Meyer, “Global transformations of nonlinear systems”, IEEE Transactions on
Automatic Control, vol. AC-28, no. 1, (1983) January, pp. 24–31.
[14] S. H. ˙ Zak and C. A. MacCarley, “State-feedback control of non-linear systems”, International Journal of
Control, vol. 43, no. 5, (1986), pp. 1497–1514.
[15] D. Bestle and M. Zeitz, “Canonical form observer design for non-linear time-variable systems”, International
Journal of Control, vol. 38, no. 2, (1983), pp. 419–431.
[16] M. Zeitz, “Comments on Comparative study of non-linear state-observation techniques”, International Journal
of Control, vol. 46, no. 5, (1987), pp. 1861–1863.
[17] B. L. Walcott, M. J. Corless and S. H. Zak, “Comparative study of non-linear state-observation techniques”,
International Journal of Control, vol. 45, no. 6, (1987), pp. 2109-2132.
[18] D. Bestle and M. Zeitz, “Canonical form observer design for non-linear time-variable systems”, International
Journal of Control, vol. 38, no. 2, (1983), pp. 419-431.
[19] A. Mesbaha, A. E. M. Huesmana, H. J. M. Kramerb and P. M. J. Van den Hofa, “A comparison of nonlinear
observers for output feedback model-based control of seeded batch crystallization processes”, Journal of
Process Control, vol. 21, (2011), pp. 652–666.
[20] A. Ebrahim and G. Murphy, “Adaptive backstepping control of an induction motor under time-varying load
torque and rotor resistance uncertainty”, Int. J. of Automation and Control, vol. 2, no. 4, (2008), pp. 401417.
[21] F. A. Pintara and N. Yoganandana, “Dynamic Bending Tolerance of the Human Forearm”, Traffic Injury
Prevention, vol.3, no. 1, (2002), pp. 43-48.
324
Copyright ⓒ 2014 SERSC
International Journal of Control and Automation
Vol.7, No.2 (2014)
Author
Bo
ok
m Onl
ad in
eb eV
y er
th si
is on
fil O
ei n
s I l y.
LL
EG
A
L.
Gamal A. Elnashar was born in Dakhliah-Egypt on February 17,
1965. In 1988 he received his B.Sc. degree from the department of
electrical engineering. His MSc. Degree in the field of automatic control
from Military Technical Collage (MTC)-Egypt in 1994. He received his
Ph. D. degree from the department of electrical and computer
engineering from the Catholic University of America- Washington D. C.
in 2000. He has been serving on the MTC research faculty since the year
2000 in the areas of identification, design, and control engineering
systems. He worked as a visiting scholar in Virginia Tech.-Blacksburg in
2008. He has authored several reports and papers on data acquisition
systems, sensors and automatic control-related short courses at MTC.
Copyright ⓒ 2014 SERSC
325
Bo
ok
m Onl
ad in
eb eV
y er
th si
is on
fil O
ei n
s I l y.
LL
EG
A
L.
International Journal of Control and Automation
Vol.7, No.2 (2014)
326
Copyright ⓒ 2014 SERSC