Scribd
Upload
68 views

Nonlinear Control

1) Exact state linearization is possible if the system has a relative degree equal to the number of states, allowing transformation to a linear controllable canonical form under feedback. 2) If the relative degree is less than the number of states, an alternative output function satisfying a system of PDEs may allow exact linearization by increasing the relative degree to the number of states. 3) However, finding such an alternative output function is not always possible and its solution may be impossible to calculate explicitly, limiting the applicability of exact linearization.

Uploaded by

Santiago Garrido Bullón
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
68 views

Nonlinear Control

1) Exact state linearization is possible if the system has a relative degree equal to the number of states, allowing transformation to a linear controllable canonical form under feedback. 2) If the relative degree is less than the number of states, an alternative output function satisfying a system of PDEs may allow exact linearization by increasing the relative degree to the number of states. 3) However, finding such an alternative output function is not always possible and its solution may be impossible to calculate explicitly, limiting the applicability of exact linearization.

Uploaded by

Santiago Garrido Bullón
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 28

NONLINEAR CONTROL

Ingeniería de Control III


Examples of Exact Linearization
• Example: Exact Linearization of a Mechanical System
• Consider the system of Figure. Let x1 be the position of the cart
that moves with the velocity x2.
• We assume that the spring force is described by a nonlinear
function (x1) of the position and that the damper force is a
nonlinear function (x2) of the velocity.
• The system is then described by the equations.
• This is a nonlinear system. The special structure means, however, that
it is simple to find a feedback that compensates for the nonlinearities.

• If r is a reference signal, we can use

giving the closed loop system

• This is a linear relation between reference signal and position. By


choosing the coefficients a1 and a2 we can get any desired pole
placement.

• The example we gave is very special. We shall look at a case where it


is not as obvious how to compensate for the nonlinearity.
Example: Speed Control of Aircraft
• We will consider a highly simplified model of speed control for
an aircraft. Let the speed be x1 and the thrust of the engine mx2
where m is the mass of the aircraft.
• Let the aerodynamic drag be given by the nonlinear function
mD(x1).
• The relationship between force and acceleration for the aircraft
then gives

• Let the desired thrust be the input u to the system. We assume


that the engine responds with a certain time constant, giving
• A state space description of the system is thus

• Now, it is not possible to compensate directly for the


nonlinearity D(x1), by choosing u, since those quantities are not
in the same equation.
Relative Degree
• We must first decide what nonlinearities to consider. It turns out
to be useful to assume the system to be in control affine form

where f, g and h are infinitely differentiable functions.


• The state x is assumed to be a vector with n components, while
u and y are scalars.
• We allow essentially arbitrary nonlinearities in the state but
assume a simple affine dependence in u.
Relative Degree
• Obviously y does not depend explicitly on u, since u is not
explicitly present in h.

• If u is changed instantaneously there will be no immediate


change in y; the change comes gradually via x.

• To check the corresponding behavior of y we can differentiate


the output equation.
Relative Degree
• We then get

where

which can he written

• Here we have suppressed the x-dependence in the notation.


• We see that y depends directly on u, at least for some values of
x, if and only if
Relative Degree
• We will say that the system has relative degree 1 if

• Now assume that Differentiating once more we get

• The system is said to have relative degree 2 if

• In that case
y depends directly on u.
• Otherwise we can continue differentiating.
Relative Degree
• It is useful to introduce some notation for the repeated
differentiations.
• Introduce the partial differential operators

• L f is called the Lie-derivative in the direction of f.


• Letting v denote the relative degree, we can rewrite our
calculations as follows
Relative Degree
• Definition The relative degree v is the least positive integer
such that

• Definition A system has strong relative degree v if the relative


degree is v and furthermore
Input-Output Linearization
• Consider a system with strong relative degree v.
• From the definition and computations of the previous section
we have

• We see an interesting possibility: introduce the feedback

• This is possible since

• The resulting relationship is from reference signal to


output.
Input-Output Linearization
• There is unfortunately a disadvantage in this methodology,
which becomes clear with a change of variables.
• Let us introduce the new variable
Input-Output Linearization
• Remarks
• Even if the choice of the last components of z is essentially
arbitrary, it must of course be done so that  becomes invertible,
i.e. it must be possible to compute x for a given z.

• It is possible to show that this can be done, at least locally in


the state space.
Input-Output Linearization
Input-Output Linearization
• The linearizing feedback is
Input-Output Linearization
• We see that the whole system has not been linearized. There is
possibly still some nonlinear dynamics affecting the state
variables

• This dynamics is not visible in the output and is called the zero
dynamics of the system.
• If one performs the calculations above for a linear system, it is
possible to show that the poles of the zero dynamics equals the
zeros of the transfer function from u to y.
• It is natural to view instability of the zero dynamics as a
generalization of the concept of non-minimum phase to
nonlinear systems.
Exact State Linearization
• We note that the potential problems with the zero dynamics disappear if v = n. This is
such an important special case that it deserves a theorem.
Exact state linearization
• This form of feedback is called exact state linearization or just exact
linearization.
• Remark 1. A natural variation is to use the feedback,

where the ai are chosen to give a good pole placement.

• Remark 2. It is not obvious that one should aim at a linear dynamics.

• A more general feedback would be


replacing the nonlinear function  e by another arbitrary function  .
One could thus try to replace "bad" nonlinearities by "good"
nonlinearities.

• Remark 3. This description is often regarded as a controller


canonical form for nonlinear systems.
• Now suppose we have a system with relative degree v < n. We can
argue as follows: since we have to use a feedback from all state
variables, we can regard some other function of the state as output.

• This alternative output can then be chosen so that the relative degree
is n. Denote the new output

• We see that the function c has to satisfy

• This is a system of partial differential equations for the function c.

• One can show that sometimes lacks a solution. Even when a solution
exists, it is not certain that it can be calculated explicitly. Here we find
a definite limit for the use of exact linearization.

You might also like