Вы находитесь на странице: 1из 57

Model Reference Adaptive

Control (MRAC)


The Model-Reference Adaptive system (MRAS) was originally

proposed to solve a problem in which the performance
specifications are given in terms of a reference model.
This model tells how the process output ideally should respond to
the command signal.
The Adaptive Controller has two loops. The inner loop consists of
the process and an ordinary feedback controller.
The outer loop adjusts the controller parameters in such a way that
the error, which is the difference between the process output y and
model output ym is small.
The MRAS was originally introduced for flight control.
In this case, the reference model describes the desired response of
the aircraft to joystick motions.
Model reference adaptive systems were originally derived for
deterministic continuous systems.

The mechanism for adjusting the

parameters in a model reference adaptive
system can be obtained in two ways:
Using a gradient method and
Appling a stability theory

MIT Rule

MIT Rule for adaptation of feed forward gain

Let the process be described by the transfer function kG(s),

where G(s) is known and k is an unknown parameter.
The design problem is to find a feedforward controller that gives a system
with transfer function Gm(s) = k0 G(s)
where k0 is a given constant.
With the feed forward controller u = uc ,
where u is the control signal and uc the command signal,
the transfer function from the command signal to output becomes

MIT Rule based MRAS for first

order system

Determination of adaptation gain


Design of MRAS using Lyapunov

The drawback of MIT rule based MRAS
design is that there is no guarantee that
the resulting closed loop system will be
To overcome this difficulty, the Lyapunov
theory based MRAS can be designed,
which ensures that the resulting closed
loop system is stable.

Lyapunov theory based MRAS for

first order system

Relation between MRAS and STR

MRAS and STR were regarded as two
quite different approaches to adaptive
But later it was proved that they are
actually very closely related.
In particular the direct self tuning regulator
with cancellation of process zeros can be
interpreted as a MRAS.

Gain scheduling

Gain scheduling is an adaptive control strategy, where

the gain of the system is determined and based on its
value the controller parameters are changed.
In many cases, it is possible to find measurable variables
that correlate well with changes in process dynamics.
These variables can then be used to change the
controller parameters.
This approach is called gain scheduling because the
scheme was originally used to measure the gain and
then change, that is, schedule the controller to
compensate for changes in the process gain.

The system can be viewed as having two loops.

There is an inner loop composed of the process and the
Outer loop contains components that adjust the controller
parameters on the basis of the operating conditions.
Gain scheduling can be regarded as mapping from process
parameters to controller parameters.
It can be implemented as a function or a table lookup.
The concept of gain scheduling originated in connection with
the development of flight control systems.
In this application, the Mach number and the altitude are
measured by air data sensors and used as scheduling
variables. This was used, for instance, in the X-15.
In process control the production rate can be often chosen as
a scheduling variable, since time constants and time delays
are often inversely proportional to production rate.
Gain scheduling is thus a very useful technique for reducing
the effects of parameter variations.

Parameters can be changed quickly in response to
changes in plant dynamics
This strategy is very easy to apply if the plant dynamics
depends in a well known
fashion on a relatively few easily measurable variables
It is an open-loop adaptation scheme, with no real
learning or intelligence
The design required for its implementation is enormous.

Design of gain scheduling

The key issue in the design of gain scheduling
controllers is the determination of variables that
can be used as scheduling variables.
One criterion for selection of the scheduling
variable is that these auxiliary variables must
reflect the operating conditions of the plant.
Ideally there should be simple expressions for
how the controller parameters relate to the
scheduling variables.
It is thus necessary to have a good insight into
the dynamics of the process if gain scheduling is
to be used.

Design of gain scheduling controllers can be carried out by

one of the following techniques:
Design of gain scheduling controllers can be carried out
by one of the following techniques:
Gain scheduling based on measurement of auxiliary
Time scaling based on the production rate and
Nonlinear transformations
It should be noted here that by linearizing of nonlinear
actuators, we get a very improved performance, but this
should not be regarded as gain scheduling because,
gain scheduling should consist of a measurement of
variable related to the operating condition of the process.

Gain scheduling based on measurement of

auxiliary variables

From the above expression, it is clear that it is sufficient to make the gain
proportional to the cross section of the tank.
So it can be seen that we have established a relation between the gain (auxiliary
variable) and the area of the tank (variable relating operating condition of process).
The above example illustrates that it is sufficient to measure one or two variables
in the process and use them as scheduling variables.
But often it is not easy to determine the controller parameters as a function of
measured variables.
The design of controller must then be redone for different working points of the
process. Some care must also be taken if the measured signals are
noisy. They should be filtered properly before they are used as scheduling

Time scaling based on the production rate

Consider concentration control for a fluid that flows
through a pipe, with no mixing, and through a tank, with
perfect mixing. A schematic diagram of the process is

Notice that the sampled data model has only one parameter, a, that does
not depend on q.
A constant gain controller can easily be designed for the sampled data
The gain scheduling is realized simply by having a controller with
constant parameters, in which the sampling rate is inversely proportional
to the flow rate.
This will give the same response independent of the flow, in looking at
the sampling instants, but the transients will be scaled in time.
To implement the gain scheduling controller for the case where flow is
varying, it is necessary to measure not only the concentration but also
the flow.
Errors in flow measurement will result in jitter in the sampling period. To
avoid this, it is necessary to filter the flow measurement.

Nonlinear Transformations
It is of great interest to find transformations such that the transformed
system is linear and independent of the operating conditions.
In case of concentration control problem, time scaling was used to
make the model independent of flow.
i.e ts = (Vd/q)t was used to make the model independent of flow.
All processes associated with material flows like rolling mills, band
transporters, flows in pipes, etc. have this property.

The design procedure making use of nonlinear

transformation is as described below:
The system is first transformed into a fixed linear system.
The transformation is usually nonlinear and depends on the
states of the process.
A controller is then designed for the transformed model,
and the control signals of the model are retransformed into
the original control signals.
The result is a special type of nonlinear controller, which
can be interpreted as a gain scheduling controller.

The nonlinear transformation u = g1(x,v) and

z = g2(x) makes the relation between v and z
A state feedback controller from z is then
computed that gives v.
The control signal v is then transformed into the
original signal u.
Feedback linearization requires good knowledge
about the nonlinearities of the process.
The method of nonlinear transformations is
described with the following examples.

Example: Nonlinear transformation

of a Pendulum

Example: Nonlinear transformation

of a second order system

Auto-Tuning of PID regulators

The commonly used techniques for auto
tuning of PID regulators are:
Open loop response method (transient
response method)
Closed loop response method
Tuning by use of external equipment
Use of expert systems

Transient response methods

Methods based on Relay Feedback: (Ultimate cycle


The main drawback of the transient response

method is that it is sensitive to disturbance
because it relies on open loop experiments.
The relay based methods avoid this difficulty
because the required experiments are
performed in closed loop.
The key idea here is the fact that many
processes have limit cycle oscillations under
relay feedback.

Here the idea is to determine the critical gain and critical period first
and then determine the controller parameters.
The critical period (Tu) is the period of oscillations when relay
feedback is applied.
The gain at which the oscillations just begin is termed as critical
period (Ku).
Ziegler and Nichols have devised a simple method for determining
the parameters of the controller based on the values of critical gain
and critical period

The controller parameters are given by following

Systems with better

damping can be obtained by relay auto tuner

When tuning is demanded, the switch is set to T, which

means that relay feedback is activated
and the PID regulator is disconnected. When a stable
limit cycle is established, the PID
parameters are computed, and the PID controller is then
connected to the process. Naturally the
method will not work for all systems. First, there will not
be unique limit cycle oscillations for an
arbitrary transfer function. Second, PID control is not
appropriate for all processes. This type of
tuning works well for a large class of systems
encountered in process control.