Motion Control Systems

The controller of a robot sends a signal to one of the actuator causing it to accelerate toward the
next location. Even if a feedback signal is used to measure the movement and stop the motion
as soon as the joint reaches the desired destination, the joint may overshoot and go beyond the
desired value, requiring that a negative signal be sent to the actuator to return it and perhaps
continue this back-and-forth motion until the position is achieved accurately. At worst, with an
unstable system, the oscillations may become larger, not smaller, and eventually destroy the
system. This happens because the linkage and the actuator have inertia, and they may not stop
immediately when the signal is turned off. Obviously, it should be possible to decrease the signal
(current, voltage and so on) to the actuator and slow it down as it approaches the destination in
order to avoid overshoot.

Basic Components and Terminology

A control system is used to change (control) the behavior of a device, machine, or process (called
a plant). The plant may be an air conditioning system, a chemical process. In each case, the plant
creates an effect (output) such as the changing temperature of the room or the iron, the flow
rate in the chemical process, or the motion of the robot arm. To perform its function, the control
system uses sensors such as a thermostat, a flow-meter, or a potentiometer and encoder to
measure the output of the plant. The controller receives the output signal and, based on its
design, controls the plant and its output in relation to the desired input the desired temperature
in a room or on an iron, the rate of flow, or the final destination of the robot linkage and its speed.
There is a difference between the nature of different controllers. In the air conditioning system
or the iron, the controller regulates the output. In the robot, it tracks the motions and controls
their specification. This is called a servo-controller system.
An open-loop controller lacks a feedback signal so it is not aware of the output. A robot’s
processor calculates the destination’s joint variables and sends the information to the controller.
If the robot controller were open-loop, it might send a signal to a joint motor proportional to how
far it needs to go, but not be aware of whether or not the joint moves at the desired rate.
However, with a closed loop controller, which includes feedback, the controller will receive a
signal from the joint indicating its response to the control signal. If the motion is not as desired,
the controller increases or decreases the control signal to force the arm to behave as desired.
The error signal is the driving signal for the controller. For a stable system, the feedback signal
must be subtracted from the input reference signal. Otherwise, if they are added, the resulting
signal becomes larger as the output increases, further increasing the output until the system
‘‘blows up.’’ To better understand the relationship between the different elements of a control
system, let’s first consider the behaviour of a plant or the system’s dynamics.
A block diagram assists us in finding the relationship between different elements of the system
such as the plant the signals, the controller, the feedback loop, and others. The block diagram
also shows how the signals flow between different elements and how they are used. Later, we
use block diagrams to represent systems and derive mathematical relationships that govern
them.
System Dynamics

A plant’s behaviour is a function of its physical characteristics and external influences and how
they are related to each other. For example, when an iron is connected to a power source, it
starts to heat up. The rate at which the iron heats up is a function of factors such as voltage, the
resistance of the heating element, and how the element is attached to the body as well as a
function of the heat capacity of the iron and the materials used. As the iron’s temperature
increases, the heat it dissipates also increases until eventually a balance may be achieved.
Therefore, the behaviour of the iron is a function of its heat capacity, the input power, rate of
heat dissipation, and materials used. If a voltage is applied to a motor, the angular acceleration
of the rotor is a function of the voltage and the inertia of the rotor. However, if the same motor
is attached to an arm, the moment of inertia of the arm also affects the angular acceleration of
the rotor. The system’s dynamics is generally represented by differential equations and must be
known before a control scheme can be designed for the system.
Characteristics of First-Order Transfer Functions-

Time constant is an indication of how fast the system responds to the step function. 2. At a=1/t
is a pole. The location of the pole in a real-imaginary plane relative to the imaginary axis (y-axis)
specifies whether or not the system is stable and how fast it responds. A longer t indicates a
longer time to reach the final value. 3. Rise time is the time required between 10% and 90% of
the final value and can be found by substituting 10% and 90% in eq. as Tr = 2.2t 4. Settling time
is the time from 0% to 98% rise and is Ts = 4t. 5. Slope indicates how fast the rise time is. As t
increases, the slope decreases. Since rise time cannot be zero, the slope cannot be infinite.
Characteristics of Second-Order Transfer Functions

1. The peak time is the time to the maximum response value. 2. The rise time is the time it takes
to go from 10% of the response to 90%. 3. Unlike for a first-order system, no time constant
is defined for a second-order transfer function. 4. Settling time is reached when the
response does not vary more than (+-)2%

Characteristic Equation: Pole/Zero Mapping

When the denominator of the transfer function is set to zero, the resulting equation is called the
characteristic equation. The roots of the characteristic equation are called poles, whereas the
roots of the numerator of the transfer function are called zeros. Pole/zero mapping is the
graphical representation of the locations of the poles and zeros in a real imaginary plane.

# A tele robot (such as a surgical robot or a space repair-robot that is meant to follow the
operator’s motions accurately), the steady-state output should be the same as the input. In that
case, the robot’s motions will be the same as the operator’s input joystick one to one mapping
with no steady state error signal. On the other hand, a large robot with large motions for
example, the space shuttle robot that handles satellites must make large motions for small
motions of the joystick, and therefore, must have a large gain, even though no steady state error
signal is desired.
Root Locus Method
The root locus is the collection of the loci of the roots of the characteristic equation plotted on a
real imaginary plane as parameters vary. Root locus is a powerful tool for both analysis of the
system whether or not it is stable, system sensitivity, whether or not it is underdamped, critically
damped, or over damped, and so on as well as system design (determining the location of roots
or the magnitude of the gains for specific system behaviour).
If poles and zero are all to the left of the imaginary axis, this system can never be unstable. It
makes root locus a very powerful and useful technique.
Start and End of the Root Locus - The start of root locus is where pole is zero (poles of the open
loop transfer function). Each portion of the root locus ends at a zero or at 1. Therefore, by plotting
the open-loop transfer function zeros, the ends of root loci can be marked off. Each portion starts
at a pole and ends at a zero or 1. If all the roots are numbered sequentially from right to left, the
root locus exists to the left of the odd-numbered roots only.
# Proportional Controllers
There is proportional gain in the feed forward loop. When gain varies, the locations of the poles
and zeros of the system changes on a root locus. Therefore, the system’s behaviour is dependent
on the value of gain. We may desire a system with an overshoot less than a certain percent, an
overdamped system, or a system whose rise time is less than a certain value. This process, called
pole placement, allows the designer to select poles and calculate the value of gain that yields
the particular pole locations.

Proportional controllers are the simplest controllers and are very common. We only need to
change the amplification value of a controller that already exists without having to add anything
to the system. However, it is not always possible to find appropriate pole locations with
proportional controllers that yield satisfactory results.

Proportional plus Integral Controllers - Integral controllers provide a means for eliminating
steady state error in a system. This is because an integrator adds an additional s to the
denominator of the transfer function, therefore raising its type. For type-0 systems, the steady
state error for a step function is a finite value and for type-1 systems, it is zero. The steady state
error with a ramp input for a type-0 system is infinite, but finite for a type-1 system. Each
integrator within the system raises its type, reducing the error boundary.
If we are designing a control system for a robot. It should be clear that (1) the response of the
robot actuators to a step function (to go from one location to another) should not overshoot, (2)
it should rise to the value of the input signal as quickly as possible, and (3) it should not have any
steady-state error. Obviously, if the response has an error, all our estimations of acceleration,
velocity, and positional accuracy will be wrong. Therefore, we need to create a controller that
delivers all these requirements simultaneously. When a proportional controller was used, even
allowing an overshoot resulted in a significant steady-state error. When we placed the poles on
the real-axis, making it critically damped and eliminating the overshoot, the steady-state error
was further increased. In order to have faster response, a high gain is needed, but that creates
overshoot and error. When overshoot is reduced, error increases further. Consequently, a
proportional controller alone cannot simultaneously provide for fast response, no overshoot, and
zero steady-state error. However, a system with both proportional and integral elements will
improve system response. The integral gain should be small compared to the proportional gain.

Proportional-plus-Derivative Controllers- Sometimes it is impossible to meet the design

requirements with proportional or proportional plus integral controllers. In these cases, the
dynamic behaviour of the system must be altered in order to achieve the design requirements.
This may be achieved by proportional-plus derivative (PD) controller.
Proportional-Integral-Derivative Controller (PID)- The addition of a derivative component to a
controller changes its behaviour, allowing the placement of the roots in desirable locations and
thereby achieving design requirements. However, a proportional and derivative controller may
not result in zero steady-state error. Many systems, including robots, may require zero steady-
state error in addition to other requirements. Therefore, it will be necessary to add an integrator
to the system as well. However, care must be taken to ensure that the addition of the integrator
does not otherwise change the behaviour of the system. In order to maintain the behaviour of
the system and the general shape of the root locus, we may place one of the zeros of Equation
near the origin and therefore cancel the dynamic effect of the integrator pole at the origin (called
zero-pole cancellation).With this, although the system behaviour remains almost unchanged, its
steady-state error will go to zero because the system type is raised. Remember that zeros must
be real and distinct.
# Ideal integral and derivative controllers are used to change the response of a plant according
to the required design specifications such as the settling time, speed of the response, percent
overshoot, and steady-state error elimination. However, they are both active systems and
require power. In addition, a derivative controller has a wide bandwidth, therefore although it
can differentiate high frequencies in the system, it can also create problems when noise is
present. Alternately, a lead compensator or a lag compensator may be used. In each case, the
circuits for lead and lag compensators are passive, basically consisting of resistors, capacitors,
and inductors. Lead and lag compensation is usually performed along with frequency domain
analysis of system.
Lag Compensators - A lag compensator consists of a zero placed near a pole close to the origin.
The addition of the pole near the origin (and not exactly at the origin which makes it a pure
integrator) acts similar to an integrator, but over time the system loses its accuracy as the steady-
state error increases. Therefore, lag compensators are assumed to be leaky. The addition of the
zero near the pole keeps the root locus about the same.
Lead Compensators - A lead compensator consists of a zero near the origin that acts similar to a
derivative controller, plus a pole near it. A lead compensator causes little change in the overall
shape of the root locus, but provides for passive derivative compensation with limited
bandwidth. A lead compensator has a limited bandwidth so they are better for high frequency
noise reduction.

The analysis and the design techniques associated with root locus are based on the time (or
Laplace) domain. However, many systems function with inputs that vary continuously;
therefore, it is better to analyse them in the frequency domain.
The transfer function that describes the relationship between the input and output of a system
is only applicable when the system is initially relaxed (no initial conditions; otherwise, the Laplace
transform cannot be applied) and only relates the inputs and output, but not the internal signals
within the system. An alternative to this method of representation is state-space, where different
signals within the system may be linked together to create a set of first-order linear equations
that are easy to solve and provide information about internal signals. State-space is two
dimensional, linear, time invariant equation, where the second-order part of the equation is
converted into first-order by the introduction of the velocity state variable.
Digital Control
Digital control is used in systems where microprocessors are used for controlling the system and
in which signals are sampled. Many of the techniques used in analog control are also used in
digital control systems, including root locus, lead-lag, proportional, integral, and derivative
control, Bode diagrams, and others. However, one essential difference is that digital systems are
discrete, not continuous. Principally, a digital system can first be designed in the s-plane as an
analog system, and subsequently through digital filtering, be converted to digital domain (called
z-plane), or it can be designed in the digital domain. So long as the sampling rate in the system is
relatively high, both techniques are acceptable.