Non-Linear Control Systems: Recommended Books

Non-Linear Control
Systems
Dr. Iftikhar Ahmad
Rana
Recommended Books
Nonlinear Systems
by Hassan Khalil
Applied Nonlinear Control
by Jean-Jacques E. Slotine
&Weiping Li
http://web.mit.edu/nsl/www/videos/lectures.html
https://www.youtube.com/watch?v=ONshHl6J__I
Introduction
Deals with analysis and design nonlinear
control systems i.e. of control systems having
at least one nonlinear component.
It is assumed that a nonlinear closed-loop
system is given, and we wish to determine
the characteristics of systems behavior.
In the design, we are given a nonlinear plant
to be controlled and some specifications of
closed-loop system behavior, and our task is
to construct a controller so that the closedloop system meets the desired characteristics
Why Nonlinear Control

Linear control is a mature subject
with a variety of powerful methods
Why researchers from aircraft and
spacecraft control, robotics, process
control, and biomedical Engg. have
recently shown interest in
development and application of
nonlinear control methodologies.
Many reasons for this are:
Improving in existing control systems

Linear control methods
Rely on key assumption of small range
operation for linear model to be valid.
For large range poor performance/unstable
(because nonlinearities not compensated)
In Robot motion: Nonlinear forces (Coriolis
and Centripetal vary as square of speed)
been neglected. Controllers accuracy
quickly degrades as speed increase.
http://ww2010.atmos.uiuc.edu/(Gh)/guide
s/mtr/fw/crls.rxml
So to achieve pre-specified accuracy in robot tasks

(pick-and-place, arc welding, laser cutting)
the speed of motion, thus productivity, has to be
kept low.
In Nonlinear Controllers:
May handle large range operations directly

a large workspace), highly accurate for a very large
range of robot speed,
Simple NLC (computed torque controller) can fully
compensate the nonlinear forces in robot motion
Analysis of Hard Nonlinearities

Supposition in LC: System model is
linearizable
But discontinuous nature of some NLties
(Coulomb friction, saturation, backlash,
hysteresis) does not allow lin. approx called
HNLties. Their effects cant be derived from
linear methods- they cause undesirable
behavior like instabilities,
spurious(bastard) limit cycles
Their effect must be predicted and
compensated, by using NL methods.
Dealing with model

uncertainties
Supposed: model parameters are well-known

But many control probs involve uncertainties, may be
due to slow time varying parameters (ambient air
pressure during flying aircraft) or abrupt change
of parameter (in inertial parameter of robot due
to grasping new object)
LC based on inaccurate model parameters may
exhibit performance degradation, even instability
Nonlinearities can be intentionally introduced in
controller part so that model uncertainties may
be tolerated
E.g. Robust and adaptive controllers for this
purpose
Design simplicity
Good NLC designs may be simpler (than their linear
counterparts) as they are deeply rooted in the physics
of the plant.
E.g. Swinging Pendulum, attached to a hing in
vertical plane
Starting from some arbitrary initial angle, it oscillates,
then stops along vertical. Its beh. could be analyzed
by linearizing the system, but physically its
stability has very little to do with the e-values of
some linearized system matrix:
Actually, total mech energy of the system is
progressively dissipated by various friction
forces (e.g at hing) so that the pendulum comes
to rest at a position of min energy
Cost and performance

optimality
LCs may require high quality
actuators and sensors to produce
linear behavior in the specified
operation range
But NLC may permit the use of less
expensive components with NL
characteristics
As the performance optimality, we
can cite bang-bang type
controllers which can produce fast
Nonlinear Control
Area of automatic control
Learning techniques of its analysis
can enhance the ability of control
engg to deal with practical control
probs effectively
Provides sharper understanding of
real world viz inherently NL.
In past computational difficulty
associated with NLC analysis, now
computer tech.
NLC for large range operations

.attracted, and powerful
microprocessors has made
implementation of NLCs simple
matter
Demand of high-speed highaccuracy robots, high performance
aircrafts
Importance of NLC theory, clear
from ever increasing no. of papers,
reports on NLC research &
Nonlinear System Behavior

Physical systems are inherently NL
which are represented by NL DEqs.
But if operating range of control
system is small and involved
nonlinearities are smooth, then the
control systems may reasonably
approximated by a linearized system
whose dynamics is defined by a set
of linear DEqs.
Nonlinearities
Inherent/natural: come with systems
hardware and motion, e.g. centripetal
forces in rotational motion, Coulomb
friction bet contacting surfaces. They have
undesirable effects and control systems
have to be properly compensated for
them.
Intentional/artificial NLties: e.g.
Adaptive control laws and bang-bang
optimal control laws
Continuous and discontinuous

NLties
Discontinuous NLties cant be locally
approximated by linear funs, they are
called Hard NLties
Linear Systems
LC theory predominantly deals with
the study of LTI control system of
form , being state vector, system
matrix
A linear system has a unique eq pt, if
is sing.
Eq pt is stable if all e-values of have
ve real parts, regardless of ICs.
General solution can be solved
analytically
In the presence of external input

u(t)

i.e.
The system response has properties:
It satisfies principle of superposition
Inelectrical engineering, in alinear circuit, the input (an applied
time-varying voltage signal) is related to the output (a current or
voltage anywhere in the circuit) by a linear transformation. Thus, a
superposition (i.e., sum) of input signals will yield the
superposition of the responses.
Asmp stability of implies bounded-input

bounded-output stability in presence of
A sinusoidal input leads to sinusoidal
output of same freq.
NL system behavior
of NL system is much more complex.
Behavior
Due to lack of linearity and associated

superposition property, NL response to
external inputs is quite different
Example: A simplified model of the motion of
an underwater vehicle can be written
is vehicle velocity, is control input (thrust
provided by a propeller). The Nlty
corresponds to a typical square law drag.
If we apply unit step input in thrust u from motors for 5 sec (and then we
remove) to see what happens (as we normally do with the linear systems), 5 sec
later by ve unit step input
The velocity increases to the extent of becoming constant where the square drag
& the thrust exactly balance each other
After 5 sec when you remove the thrust (u=0), you are left with the drag which
decreases the velocity
Initially v=0
1
Velocity v
Thrust u
t
0
5 sec 10 sec
5 sec 10 sec
We see that the system settles much faster in response to +ve unit step
than it does
to the subsequent ve unit step, very sluggish behavior
i.e. apparent damping coefficient is larger at high speeds than at
low speeds.
Now if we use larger step input of amplitude u=10

10
10
Velocity v
Thrust u
2
0
5 sec
10 sec
5 sec
10 sec
ng speed in response to the first step is not 10 times that obtained in

to the first unit step in previous experiment as it would be in a linear s
This can again be seen as follows
efully understanding and effectively controlling this

nlinear behavior is important if the vehicle is to move
large dynamical range and change speeds continuall
s typical of industrial remotely operated underwater vehicle.
Some Common NL System

Behaviors
1-Multiple
Equilibrium Points
NL systems frequently have more than one Eq.
point
(where the system can stay forever without
moving)
Example: Consider the first order system
Its linearization is
And its solution is
can be written
as
Integrating
Gives the actual response of given

system, which is
This response is plotted for various Ics.

System has two eq. points
Its qualitative behavior strongly
depends on its Ics.
Response of Linearized system
Response of Nonlinearized system
Motion Stability
For
linearized system, the motion
always converges to eq. point for any IC.

But for actual NL system, the motion
starting from will converge to eq. point
And those starting from will converge to
infinity (actually in finite time called
finite escape time)
This means that stability of NL system
depends on ICs
In the presence of bounded external

input
The
stability may also be dependent on the
input value.
Consider the bilinear system
If , then state
If , then state
2-Limit Cycles
NL
systems can display oscillations of fixed
amplitude and fixed period without external
excitation called Limit Cycles (self-excited
oscillations)
Example: Van der Pol Equation

The 2nd order NL D.Eq.
, , and are +ve constants. Mass spring damper
system with a position dependent damping coefficient
For
large values of , the damping co-eff. is +ve
which removes energy from the system, so the
system motion is cgt,
For small values of , the damping co-eff. is -ve
which adds energy to the system, so the system
motion has dgt tendency,
Shows sustained oscillations independent of Ics.
This limit cycle is sustained by releasing energy
into and absorbing energy from the environment,
through the damping term.
Sustained oscillations can also be seen in marginally

stable linear systems (such as mass spring system
without damping or in response to sinusoidal inputs)
Difference
In NL systems, the amplitude of self-sustained excitation is

indep of Ics while oscillations of marginally stable linear
systems has its amplitude determinted by its Ics
Marginally stable linear systems are very sensitive to
changes in system parameters (slight change lead to
stable convergence or to instability) while limit cycles are
not easily affected by parameter changes.
LIMIT CYCLES can be found in many areas of

engineering
Aircraft wing fluttering: Limit cycle

caused by interaction of aerodynamic forces
and structural vibrations, frequently
encountered, sometimes dangerous.
Hopping motion of a legged robot, is
another instance of limit cycle,
In electrical circuits: they occur in
laboratory electronic oscillators
y are Desirable/Undesirable, an engg has to know how to eliminate them whe

esirable.
versly how to generate or amplify them when desirable
quires an understanding of properties of limit cycles and a familiarity with the
s for manupulating them
3- Bifurcation
Stability of eq. points or the no. of eq. points can change on

changing the parameters of NL dynamical system
Values of these parameters at which qualitative nature of
systems motion change are critical/bifurcation values.
Phenomenon called bifurcation theory.
E.g. smoke rising from cigarette first accelerates upwards
(as is lighter than ambient air) but beyond some critical
velocity breaks into swirls.
Example: Undamped Diffusion Equation

Damped Diffusion Equation is
In plot eq. points as function of , we see as varies from
+ve to ve,
One eq. point splits into 3, () representing qualitative
change in dynamics, so is critical bifurcation value
called PITCHFORK
Stable 0
Stable
Unstable
Stable
Another kind of bifurcation involves the emergence of

limit cycles
In
this case, a pair of complex conjugate evalues cross the left-half plane into the
right-half plane
And response of the unstable system
diverges to a limit cycle. Plot for state
trajectories ( and ) as is varied
Bifurcation value
Called
Hopf Bifurcation
Stable eq. point
Limit cycle
4-CHAOS
For stable linear systems, small
difference in Ics can cause only small
difference in output.
NL systems show chaos, i.e. system
output is extremely sensitive to ICs,
so that output of the systems
becomes unpredictable (even if exact
model or extremely accurate
computer)
Steady state behavior is highly
showing chaotic behavior:

Example
Consider a simple NL system which may represent lightlydamped, sinusiodally forced mechanical structure
undergoing large elastic deflections
Ics taken for plot are
(thick line)
(thin line)
Due to presence of strong Nlty , the two responses are radically
different after some time.
3
2
1
0
1
2
3
10
15
2
0
2
5
3
0
3
5
4
0
4
5
5
0
Time/Sec

Non-Linear Control Systems: Recommended Books

Uploaded by

Copyright:

Available Formats

Non-Linear Control Systems: Recommended Books

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Non-Linear Control Systems: Recommended Books

Uploaded by

Copyright:

Available Formats

Non-Linear Control

Why Nonlinear Control

Improving in existing control systems

So to achieve pre-specified accuracy in robot tasks

May handle large range operations directly

Analysis of Hard Nonlinearities

Dealing with model

Supposed: model parameters are well-known

Cost and performance

NLC for large range operations

Nonlinear System Behavior

Continuous and discontinuous

In the presence of external input

Asmp stability of implies bounded-input

Due to lack of linearity and associated

Now if we use larger step input of amplitude u=10

ng speed in response to the first step is not 10 times that obtained in

This can again be seen as follows

efully understanding and effectively controlling this

Some Common NL System

Gives the actual response of given

This response is plotted for various Ics.

Response of Linearized system

Response of Nonlinearized system

always converges to eq. point for any IC.

In the presence of bounded external

Consider the bilinear system

Example: Van der Pol Equation

Sustained oscillations can also be seen in marginally

In NL systems, the amplitude of self-sustained excitation is

LIMIT CYCLES can be found in many areas of

Aircraft wing fluttering: Limit cycle

y are Desirable/Undesirable, an engg has to know how to eliminate them whe

Stability of eq. points or the no. of eq. points can change on

Example: Undamped Diffusion Equation

Another kind of bifurcation involves the emergence of

Stable eq. point

showing chaotic behavior:

You might also like