아주대학교 기계공학과

Mechatronic System Design

Lecture 4
FALL, 2024

Motion Control
Motion control

▪ Control the trajectory of a machine

▪ Position control
▪ Velocity control (e.g. scanning)
▪ Path planning
▪ Disturbance rejection (vibrations from
environment, imperfections of the guidance
system, …)

1
History

2
History

▪ James Watt (1736 – 1819)

- centrifugal governor (1788)
- the first powered feedback system
- proportional control

From Wikipedia
History

▪ Elmer Ambrose Sperry (1901 – 1980)

- Proportional-integral-derivative (PID) controller (1911)
- Gyroscopic compass

Sperry's stabilizing gyroscope installed in the USS Henderson

History

▪ Nicholas Minorsky (1885 – 1970)

- Application of PID control to automatic steering of ships (1922)

USS New Mexico

History

▪ Harry Nyquist (1889 – 1976)

- Stability of closed-loop systems
History

▪ John G. Ziegler(1909-1997) and Nathaniel B. Nichol

s(1914-1997)
- Ziegler-Nichols tuning rules for PID
History

▪ Walter R. Evans (1920 – 1999)

- Guidance and control of aircraft
- Root locus : Determine stability of a system
History

▪ Rudolf E. Kálmán (1930 - 2016)

▪ Kalman filter :
- Uses a dynamics model, measured control input
and process measurement to estimate the process
output
- guidance, navigation, and control of vehicles,
particularly aircraft, spacecraft and dynamically
positioned ships
Introduction

10
Introduction to Control
Measured System System
Reference error input output
Controller System
+ _

Measured Output
Sensor

▪ Controller: a device, which monitors and physically alters the operating conditions of a
given dynamical system
➢ Transient response
- To be present in the short period of time immediately after the system is turned on
- Some major structural changes
- If the system is asymptotically stable, the transient response disappears
- Performance in point of time (HDD, Elevator)

➢ Steady-state response
- Approach to a desired output
- To be present after the transient response
11
- Performance in point of accuracy (Elevator, HDD, Satellite tracing antenna)
Walk around the control loop

12
Plant

13
Impulse Response of a Dynamic System

A function of the pole locations in the Laplace plane

Feedforward Control

▪ requires no sensor
▪ predictable (if reference is known)
(faster than pure feedback)
▪ cannot introduce instability
▪ no feeding back of sensor noise

▪ A plant has to be stable

▪ no compensation of model uncertainties
▪ can only compensate for disturbances that
are known (measured)
15
Feedforward control: Example

▪ Piezoelectric scanning unit

Magnitude
Frequency (Hz)

Phase(deg)
Measured
Fitted

Frequency (Hz)
Transfer function

The problem is a low damped

first eigenmode!

16
Model based feedforward control

▪ Transfer function

▪ Compensate dynamics by feedforward controller (pole‐zero

cancellation):

▪ Such a controller is unfortunately not realizable (infinite gain

at infinite frequencies!!)
17
To increase the damping

▪ Add a well damped ζ=1 second-order transfer function:

Notch filter

▪ And with an additional first order low pass filter against

noise:

Low pass filter

▪ Which combines into:

18
Feedfoward Compensated System

Because of ‘Zeros’

19
Second method: input shaping FF filter

Extra) Adaptive feedforward 20

Feedback control

▪ can stabilize unstable systems

▪ can suppress disturbances
▪ can handle uncertainties

▪ sensor required
▪ slower than FF (error has to occur first)
▪ (sensor) noise is also fed back (amplified)
▪ can introduce instability (also of open‐loop
stable systems) 21
Sensitivity
Function

22
Sensitivity Functions of Feedback Control

✓ Response to disturbances
Complementary
sensitivity function

✓ Output disturbance rejection

Sensitivity function

23
Sensitivity Functions of Feedback Control

Complementary sensitivity function Sensitivity function

Response to reference Output disturbance rejection

24
Stability and Robustness

▪ When is a (closed‐loop) system stable?

✓ A system is asymptotically stable if and only if all poles
are in the left half of the complex plane.

▪ When is a system robust?

✓Robustness of a feedback controlled system means that
the closed‐loop system can accept (some) model
uncertainties (parameter variations).

▪ Both properties are investigated with the Bode and

Nyquist plot.

25
Loop-Shaping Design

▪ Shape the frequency response of the open-loop system such that:

✓ high gain (>>1) at low frequency
✓ low gain (<<1) at high frequency (Gain Margin)
✓ less than 180 degree phase lag in crossover region (unity‐gain phase
margin).

26
Stability

▪ Stability check using open-loop TF

✓ Gain margin GM >> 0
✓ Phase margin PM >> 0

27
PID Feedback
Control

28
Virtual stiffness by feedback in a CD player

▪ 200 μm radial vibrations at 25 Hz

▪ Mass lens: 10 ·10−3 kg
▪ Max radial error: 0.2 μm

29
Virtual Stiffness by Feedback Control

Virtual stiffness

▪ Measure position
▪ Actuate with force proportional and opposite to the deviation (feedback!)

▪ Gives virtual spring stiffness

30
PID Controller for Mass-Spring System
𝑚𝑘
𝑄= = 50
𝑐

Phase(deg) Magnitude(abs)
𝑚
0.25𝑘𝑔
𝑓0 = 𝑘/𝑚 = 20 𝐻𝑧

1
𝐺 𝑠 =
𝑚𝑠 2 + 𝑐𝑠 + 𝑘

Frequency(Hz)

Control Objectives
▪ Bandwidth of 100 Hz (unity‐gain cross‐over frequency)
✓ So far beyond the natural frequency on the mass line
PID controller
▪ Asymptotically stable, all poles in the left half plane
▪ Sufficiently damped (is requirement!)
✓ Specification, max Q = 1.3 (+3dB)
31
▪ Zero steady‐state error (0 Hz)
PID Controller

http://www.mathworks.com/matlabcentral/fileexchange/screenshots/22857/original.jpg

▪ PID controller has Proportional, Integral, and Derivative gains

P (Proportional) gain

I (Integral) gain

D (Derivative) gain

32
With P-controller

1. Preliminary guideline of ▪ Transfer function of P-controller

1
PID-control =
G ( = c )
r
e u y ▪ Open-loop transfer function
Cfb(s) G(s)
+ _

y
M(s)=1

CD‐player example:
f=800 Hz
m = 10 g = 0.01 kg
1 1
G (s) = 2
=
ms 0.01s 2

Goal
Control of lens position
: Unit gain cross-over freq. f=800 Hz 33
With P-controller (cont.)

▪ Closed-loop transfer function

e u y
r
Cfb(s) G(s)
+ _

y
M(s)=1

34
With PD-control (Damping effect)

e u y
r
Cfb(s) G(s)
+ _

y
M(s)=1

To provide 2. Second PID guideline

the desired damping effect
𝑘𝑝 𝑘𝑝
𝑘𝑑 = 𝑘𝑝 𝜏𝑑 = =
𝜔𝑑 𝟎. 𝟑𝟑𝝎𝒄
35
D-action to be terminated at 3 times of 𝒇𝒄

3. First PID guideline: correction for the additional gain of 3 dB caused by D

control (2nd guideline)
k p = kt = G −1 ( = c ) = mc 2
1
kp L pd ( = c ) = (kt + kd jc )
kd = (−mc 2 )
0.33c
kt 1
= ( kt + jc )
1 0.33c (−mc 2 )
L pd ( s ) = (k p + kd s ) 1 kt
ms 2 = (1 + jc )
0.33c (−mc 2 )
1 (mc 2 )
= (1 + jc )
0.33c (−mc 2 )
= (1 + 3 j )  3.372
Kp=0.33*kt=0.33/Gplant@fc

Lpd ( s ) = (0.33 + 1 j )  1.0572

D-action to be terminated at 3 times of 𝒇𝒄

4. Third PID guideline (tamed PD control)

d 1 +  d s
C pdt ( s ) = kt Lead-lag compensator
c 1 +  t s
Open-loop TF
 d kt 1 +  d s
Lpdt ( s ) =
c ms 2 1 +  t s
There are many resonating mode-
shapes at higher frequencies.
To attenuate these and prevent
instability 37
I-control (super-spring action)

To increase the loop gain at low freq. and achieve a zero steady-state error
4. Fourth guideline of PID-control
Integration action should start at extremely low frequencies and stop at a
frequency equal to one third of the starting frequency of D control (1/10
of 𝜔𝑐)

𝜔𝑖 = 0.1 ∙ 𝜔𝑐 ≈ 60𝑟𝑎𝑑/𝑠
𝑘𝑖 = 𝑘𝑝 ∙ 𝜔𝑖 ≈ 1.8e6

1 +  d s ki
C pidt ( s ) = k p +
1+ts s
38
Bode of PID-control

10times

𝟓𝟓°

𝒇𝒊 𝒇𝒅 𝒇𝒄 𝒇𝒕

1 +  d s ki
C pidt ( s ) = k p + 39
1+ts s
Open Loop Transfer Function (GC)

Magnitude(abs)
Phase(deg)

Frequency(Hz)

PM = ? GM = ? 40
S vs. T
Magnitude(abs)

Disturbance rejection Waterbed effect

Owing to I-control

𝑇 𝑠
Resonance frequency 𝑆(𝑠)

Frequency(Hz)

41
Increased sensitivity for disturbances

“Waterbed effect” : Bode sensitivity integral equals zero

for system with 2 more poles than zeros

42
General Guidelines

G1. P gain kp = 0.33*kt = 0.33/Gplant@fc

G2. D gain kd = kp /ωd = kp / 0.33ωc

G3. Tamed D (filtered D) ωt = 3ωc

1+d s
C pdt ( s ) = k p where  d = 1 / d = 3 / c ,  t = 0.33 / c
1+ts
G4. I gain ki = kp * ωi , ωi = 0.33ωd = 0.1ωc
1 +  d s ki
C pidt ( s ) = k p +
1+ts s
PM = 45 deg. , Max. peak of T(s) and S(s) ~ +3dB
Good starting point for further optimization 43
PID Control of Resonating Eigenmodes (Type III)

X 2 (s) k2
=
F ( s ) m1m2 s 4 + (m1k2 + m2 k1 + m2 k2 ) s 2 + k1k 2

X 2 (s) c2 s + k2
=
F ( s ) m1m2 s 4 + [m1c2 + m2 (c1 + c2 )]s 3 + [m1k2 + m2 (k1 + k2 ) + c1c2 ]s 2 + (c1k 2 + c2 k1 )s + k1k 244
PID Control of Resonating Eigenmodes (Type III)

L(s)=Cpid(s)G(s)

L(s)
Magnitude[abs]

2차 고유 모드 Q≈20

wc=100 Hz
phase[deg]

Frequency [Hz] 45
PID Control of Resonating Eigenmodes (Type III)

46
PID Control of Resonating Eigenmodes (Type III)

Solution
1. Lower loop-gain : 2차 고유모드의 진폭이 1 이하가 되도록! → bandwidth 20 Hz
2. Use a notch filter : 주파수 튜닝이 어려움, High Q
3. 센서 마운팅에 감쇠 추가 : 강성 저하
4. Additional low pass filter

47
Notch Filter

𝐾(𝑠 2 + 2𝛽𝑠 + 𝜔02 ) 𝜔0

𝐻 𝑠 = 2 𝛽𝑠 ≪ 𝐵 = 𝜔0/𝑄
𝑠 + 𝜔0 /𝑄 𝑠 + 𝜔02 2
𝜔0 ∶ 𝑁𝑜𝑡𝑐ℎ 𝑓𝑟𝑒𝑞𝑢𝑒𝑛𝑐𝑦
𝑄 ∶ 𝑞𝑢𝑎𝑙𝑖𝑡𝑦 𝑓𝑎𝑐𝑡𝑜𝑟

48
Dynamic effects (1)

▪ Actuator flexibility
✓ In the frequency range of interest (typically 𝟎~𝟏𝟎𝝎𝑩𝑾) the driven
system no longer behaves as on a rigid body due to compliance
between the motor and the load

𝒌
𝒙

49
Dynamic effects (2)

▪ Guiding flexibility
✓ Guiding system to suppress motion in an undesired direction
✓ The limited stiffness of the guiding system
✓ Relative positions of a sensor, an actuator, a center of gravity
Mode 1

Mode 2

Mode 3

50
PID Controller
Gain Tuning

51
PID Controller

▪ Comparison of P, PI, PD controller with gains

Rise time Overshoot Settling time S-S error

𝐾𝑝 (=𝐾) Decrease Increase Small change Decrease

𝐾𝐼 (=𝐾/𝑇𝐼 ) Decrease Increase Increase Eliminate

𝐾𝐷 (=𝐾𝑇𝐷 ) Small change Decrease Decrease Small change

52
Ziegler-Nichols Tuning of PID

▪ Step response model: Process reaction curve

System
G(s)

Ideal system

System
G(s)

Real system
Dead time (time delay, lag) L = td

This tuning method is an experimental method 53

Ziegler-Nichols Tuning of PID

▪ Method 1
➢ The choice of controller parameters based on a decay ratio ~ 0.25

Controller 𝐾 𝑇𝐼 𝑇𝐷 𝑇𝑃

P 1/A 4L

PI 0.9/A L/0.3 5.7L

PID 1.2/A 2L L/2 3.4L

Period

A
𝑇𝑝 expected time constant for closed-loop system
Response

0.25 A

Time (s)
54
Ziegler-Nichols Tuning of PID

▪ Self-oscillation model

✓ Preparation and process of self-oscillation model

• P-control (closed-loop system!)
• Crank up gain to self-oscillation
• Reference-step starts oscillation
✓ Get parameters for Self-oscillation model
• Ultimate gain 𝐾𝑢
• Ultimate period 𝑇𝑢
55
Ziegler-Nichols Tuning of PID
▪ Method 2
➢The criteria for adjusting parameters based on Evaluating
the system at the limit of stability (Oscillation!!)

𝐾 𝑇𝐼 𝑇𝐷 𝑇𝑃

P 0.5𝐾𝑈 𝑇𝑢

PI 0.45𝐾𝑈 1/1.2𝑇𝑢 1.4𝑇𝑢

PID 0.6𝐾𝑈 0.5𝑇𝑢 0.125𝑇𝑢 0.85𝑇𝑢

56
General Gain Tuning of PID Controller

Zero 𝐾𝐼 (= 𝐾/𝑇𝐼 ) and 𝐾𝐷 (= 𝐾 ∗ 𝑇𝐷 ). Set 𝐾𝑝 low

Apply square wave command at about 10% of

desired loop bandwidth
Use large amplitude, but avoid saturation

Raise 𝐾𝑝 for lower overshoot than required

overshoot, no ringing

Reduce noise at source

Raise 𝐾𝐷 to eliminate most overshoot
or
Increase resolution
Yes or
Too noisy ?
Lower 𝐾𝐷
No or
Lower 𝐾𝑝
Raise 𝐾𝐼 for reducing steady state error
etc

Done
Summary

▪ Feedback control: pole placement, system

stabilization, robustness
▪ Feedforward control: pole-zero cancellation,
performance enhancement, simple and fast
▪ Combination of FF and FB : two degree of freedom
control
→Trade-off between performance and robustness

58
Term Project

1. Simulate PID control for a Mass-Spring System in Fig. 4.32.

(a) Set PID gains by the PID guidelines. Determine the proper
unit-gain cross-over frequency.
(b) Draw frequency response functions of PID controller,
open-loop transfer function, sensitivity, and
complementary sensitivity function.
(c) Simulate a step reference response, and sinusoidal
reference responses of f=1, 10, 100 Hz.
(d) Tune the PID gains and discuss the effect of each gains.
(e) Discuss the results for performance and stability.
59
Term Project

(f) Find PID gains which make the fastest(highest bandwidth)

stable closed-loop system(PM>25 deg., GM>0) with a first-
order low-pass filter. Simulate step responses.
(g) Instead of low-pass filter, use a notch-filter in the loop.
Discuss the results using OL TF, S(s),T(s) and etc.

Due by 2023-05-25

Note: Use Matlab (Simulink). Write a report as a journal paper

format. Matlab Simulink model or codes should be included
with comments. Upload the report to Ajou Bb.
60