Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with...
Transcript of Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with...
![Page 1: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/1.jpg)
Feedback Fundamentals
Karl Johan Åström
Automatic Control LTHLund University
January 13, 2020
![Page 2: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/2.jpg)
Feedback Fundamentals
1. Introduction
2. Controllers with Two Degrees of Freedom
3. The Gangs of Four and Seven
4. The Sensitivity Functions
5. Fundamental Limitations
6. Summary
Theme: A closer look at feedback
![Page 3: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/3.jpg)
The Magic of Feedback
Good properties:
◮ Attenuate effects of disturbances - process control, automotive
◮ Make good systems from bad components - feedback amplifier
◮ Follow command signals - robotics, automotive
◮ Stabilize and shape behavior - flight control
Bad properties:
◮ Feedback may cause instability
◮ Feedback feeds measurement noise into the system
Arthur C. Clarke: Any sufficiently advanced technology is indistinguishablefrom magic
![Page 4: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/4.jpg)
The Magic of Integral Action
Consider a system (linear or nonlinear) controlled by a controller havingintegral action
u(t) =
∫ t
0e(τ)dτ + ....
If the closed loop is stable then the equilibrium must be such that the erroris zero!Proof: Assume that the error converges to a nonzero value leads to acontratdction!
![Page 5: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/5.jpg)
Important Issues
◮ Some important considerations
Load disturbancesMeasurement noiseProcess variationsUncertainties in modelingCommand signal following
◮ Evaluation, testing and specifications of control systems
◮ Quantification of performance and robustness
◮ Measurement and testing of performance and robustness
◮ Limitations what can and cannot be achieved by feedback
![Page 6: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/6.jpg)
A Basic Feedback Loop
F C P
Controller Process
−1
Σ Σ Σr e u
d
x
n
yv
Ingredients:
◮ Controller: feedback C, feedforward F
◮ Load disturbance d : Drives the system from desired state - e.g.slope of road
◮ Process: transfer function P
◮ Measurement noise n : Corrupts information about x
◮ Process variable x should follow reference r
![Page 7: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/7.jpg)
Quiz
Look at the block diagram
F C P
−1
ΣΣΣr e u
d
x
n
y
Find transfer function from r to u!
The Audience is Thinking ...
![Page 8: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/8.jpg)
Quiz
Look at the block diagram
F C P
−1
ΣΣΣr e u
d
x
n
y
Find transfer function from r to u!
The Audience is Thinking ...
Simple rule: blocks between singnals and 1 + PC in denominator!!
Gur =CF
1 + PC
![Page 9: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/9.jpg)
A More General Setting
Load disturbances and measurement need not enter at the process inputand measurement noise not at the output. A more general situation is.
C
Pyu
zn, d
w = (d, n, ysp), z = (e, v), find C to make z small!These problems can be dealt with in the same way but we will stick to thesimpler case. Always useful to understand disturbances, who they are andwhere they enter the system.
![Page 10: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/10.jpg)
Typical Requirements
A controller should
A: Reduce effects of load disturbances
B: Do not inject too much measurement noise into the system
C: Make the closed loop insensitive to variations in the process
D: Make output follow command signals
Performance is expressed by
◮ Response to command signals
◮ Attenuation of load disturbances
Robustness is expressed by sensitivity to
◮ Load disturbances
◮ Model uncertainty
![Page 11: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/11.jpg)
Feedback Fundamentals
1. Introduction
2. Controllers with Two Degrees of Freedom
3. The Gangs of Four and Seven
4. The Sensitivity Functions
5. Fundamental Limitations
6. Summary
Theme: A closer look at feedback
![Page 12: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/12.jpg)
Controller with Two Degrees of Freedom
F C P
Controller Process
−1
Σ Σ Σr e u
d
x
n
yv
◮ Load disturbance d : Drives the system from desired state
◮ Measurement noise n : Corrupts information about x
The controller has two degrees of freedom 2DOF because the signaltransmissions from reference r to control u and from measurement y tocontrol u are different. Horowitz 1963.
![Page 13: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/13.jpg)
A Separation Principle for 2DOF Systems
Design the feedback C to achieve
◮ Low sensitivity to load disturbances d
◮ Low injection of measurement noise n
◮ High robustness to process uncertainty and process variations
Design the feedforward F to achieve
◮ Desired response to command signals r
![Page 14: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/14.jpg)
Many Versions of 2DOFBasic loop
Σr
F C P
−1
u y
Explicit representation of ideal response ym and control signal uff
ΣΣ
r
My
Mu
C P
−1
uff
ym ufb y
For linear systems all 2DOF configurations have the same properties. Forthe systems above we have CF = Mu + CMy
![Page 15: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/15.jpg)
Many Versions of 2DOF - Kalman Filter Architecture
TrajectoryGenerator
r
u
-x̂
xm
ProcessΣ ΣState
Feedback
KalmanFilter
ufb
uff
y
◮ A nice separation of the different functions
◮ The signals xm and uff can be generated from r in real time or fromstored tables
![Page 16: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/16.jpg)
Some Systems only Allow Error Feedback
Disk drive
Atomic Force Microscope
Only error can be measuredDesign for command disturbance attenuation, robustness
and command signal response can not be separated!
![Page 17: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/17.jpg)
Feedback Fundamentals
1. Introduction
2. Controllers with Two Degrees of Freedom
3. The Sensitivity Functions
4. Fundamental Limitations
5. Summary
Theme: A closer look at feedback
![Page 18: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/18.jpg)
System with Error Feedback
C P
−1
Σ Σ Σr e u
d
x
n
y
Controller Process
Three inputs r, d and n, four interesting signals e, u, x and y!
Gxr =PC
1 + PCGxd =
P
1 + PCGxn = −
PC
1 + PC
Gyr =PC
1 + PCGyd =
P
1 + PCGyn =
1
1 + PC
Gur =C
1 + PCGud = −
PC
1 + PCGun = −
C
1 + PC
Ger =1
1 + PCGed = −
P
1 + PCGen = −
1
1 + PC
![Page 19: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/19.jpg)
The Gang of Four GOF - controller with error feedback
Gxr =PC
1 + PC, Gxd =
P
1 + PCGxn = −
PC
1 + PC,
Gyr =PC
1 + PC, Gyd =
P
1 + PC, Gyn =
1
1 + PC,
Gur =C
1 + PC, Gud = −
PC
1 + PC, Gun = −
C
1 + PC,
Ger =1
1 + PC, Ged = −
P
1 + PC, Gen = −
1
1 + PC,
Only four transfer functions!!! (Sensitivity functions - the Gang of Four)!
S =1
1 + PC, T =
PC
1 + PC= 1− S, PS =
P
1 + PC, CS =
C
1 + PC
![Page 20: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/20.jpg)
Interpretation of The Gang of Four
Response of output y to load disturbance d is characterized by
Gyd =P
1 + PC= PS =
T
C
Response of control signal u to measurement noise n is characterized by
Gun = −C
1 + PC= −CS = −
T
P
Responses of y and u to reference signal r are characterized by
Gyr =PC
1 + PC= T , Gur =
C
1 + PC= CS
Robustness to process variations is characterized by
S =1
1 + PC, T =
PC
1 + PC
![Page 21: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/21.jpg)
Visualizing the GOF - Time Responses
0 1 2 3 40
0.5
1
0 1 2 3 40
1
2
3
4
0 1 2 3 40
0.1
0.2
0 1 2 3 4
0
0.5
1S
T PS
CS
![Page 22: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/22.jpg)
Another Way to Show Time Responses
0 10 20 30 40 50 600
0.5
1
1.5
0 10 20 30 40 50 60−0.5
0
0.5
1
1.5
2
yu
Step in r Step in d Measurement noise
![Page 23: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/23.jpg)
Visualizing the GOF - Frequency Responses
10−2
100
102
10−2
100
10−2
100
102
10−4
10−2
100
10−2
100
102
100
102
10−2
100
102
10−2
100
Normalized frequency ω/aNormalized frequency ω/a
pT(i
ω)p
pS(i
ω)p
pPS(i
ω)p
pCS(i
ω)p
ω 0 = aω 0 = 2aω 0 = 4a
Gain curves for the GOF is a good way to get a quick overview of afeedback system. Curves represent three different controller are designedfor a nano-positioner
Discuss!
![Page 24: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/24.jpg)
A Warning!
Remember to always look at all responses!The step response below looks fine
0 1 2 3 4 50
0.5
1
t
y
Response of y to step in r
BUT ...
![Page 25: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/25.jpg)
All Four Responses
0 1 2 3 4 50
0.5
1
0 1 2 3 4 5−1
−0.5
0
0.5
1
0 1 2 3 4 50
20
40
60
80
0 1 2 3 4 5−1
−0.5
0
tt
yy
uu
Response of y to step in r Response of y to step in d
Response of u to step in r Response of u to step in d
The system is unstable!What is going on?
![Page 26: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/26.jpg)
The Gang of Four
Process: P(s) =1
s− 1, Controller: C(s) =
s− 1
s
Loop transfer function: L = PC =1
s− 1$
s− 1
s=
1
s
Notice cancellation of the factor s− 1! The Gang of Four
PC
1 + PC=
1
s + 1
P
1 + PC=
s
(s + 1)(s− 1)C
1 + PC=
s− 1
s + 1
1
1 + PC=
1
s + 1
Response of y to step in load disturbance d
Gyd(s) =P
1 + PC=
s
(s + 1)(s− 1)
This transfer function represents an unstable system
![Page 27: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/27.jpg)
2DOF System General – The Gang of Seven
F C P
Controller Process
−1
Σ Σ Σr e u
d
x
n
yv
The Gang of Four and transfer functions from reference r to e, x, y and u
Ger =F
1 + PCGxr =
PCF
1 + PCGyr =
PCF
1 + PCGur =
CF
1 + PC
![Page 28: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/28.jpg)
Some Observations
◮ To fully understand a system it is necessary to look at all transferfunctions
◮ A system based on error feedback is characterized by four transferfunctions The Gang of Four
◮ The system with a controller having two degrees of freedom ischaracterized by seven transfer function The Gang of Seven
◮ It may be strongly misleading to only show properties of a fewsystems for example the response of the output to command signals.A common omission in many papers and books.
◮ The properties of the different transfer functions can be illustrated bytheir transient or frequency responses.
![Page 29: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/29.jpg)
Feedback Fundamentals
1. Introduction
2. Controllers with Two Degrees of Freedom
3. The Gangs of Four and Seven
4. The Sensitivity Functions
5. Fundamental Limitations
6. Summary
Theme: A closer look at feedback
![Page 30: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/30.jpg)
The Sensitivity Functions
◮ Sensitivity function S =1
1 + PC=
1
1 + L
◮ Complementary sensitivity function T = 1− S =PC
1 + PC=
L
1 + L
◮ Input sensitivity function Gxl =P
1 + PC= PS
◮ Output sensitivity function Gun =C
1 + PC= CS
are called sensitivity functions. They have interesting properties and usefulphysical interpretations. We have
◮ The functions S and T only depend on the loop transfer function L
◮ S + T = 1
◮ Typically S(0) small and S(∞) = 1 and consequently T(0) = 1 andT(∞) small
![Page 31: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/31.jpg)
Poles Zeros and Sensitivity Functions
The sensitivity functions depend only on the loop transfer function L = PC
S =1
1 + LT =
L
1 + L
Notice that
◮ The sensitivity function S is zero and the complementary sensitivityfunction is one at the poles of L
◮ The sensitivity function S is one and complementary sensitivityfunction T is zero at the zeros of L
![Page 32: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/32.jpg)
Disturbance Attenuation
r = 0
d n
yol
P ΣΣ
Process
u ycl
d n
−C
PΣ Σ
Process
Output without control Y = Yol(s) = N(s) + P(s)D(s)Output with feedback control
Ycl =1
1 + PC
(
N + PD)
=1
1 + PCYol = SYol
The effect of feedback is thus like sending the open loop output through asystem with the transfer function S = 1/(1 + PC). Disturbances withfrequencies such that pS(iω)p < 1 are reduced by feedback, disturbanceswith frequencies such that pS(iω)p > 1 are amplified by feedback.
![Page 33: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/33.jpg)
Assessment of Disturbance Reduction - Bode
We haveYcl(s)
Ycl(s)= S(s) =
1
1 + P(s)C(s)
Feedback attenuates disturbances of frequencies ω such that pS(iω)p <.It amplifies disturbances of frequencies such that pS(iω)p > 1
10−2
10−1
100
101
10−2
10−1
100
101
Sensitivity Function
ω
pS(i
ω)p
![Page 34: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/34.jpg)
Assessment of Disturbance Reduction - Nyquist
Ycl
Yol
=1
1 + PC= S
Geometric interpretation: Distur-bances with frequencies insidethe circle are amplified by feed-back. Disturbances with frequen-cies outside the circle are re-duced. Disturbancds with fre-quencies inside the circle are am-plified. Worst amplifiaticon forfrequencies closest to the criticalpointDisturbances with frequenciesless than ω s are reduced byfeedback, those with higherfrequencies afre amplified.
−1ωms
ω s
![Page 35: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/35.jpg)
Properties of the Sensitivity Function
◮ Can the sensitivity be small for all frequencies?
No we have S(∞) = 1!
◮ Can we have pS(iω)p ≤ 1?
If the Nyquist curve of L = PC is in the first and third quadrant!Passive systems!
◮ Bode’s integral, pk RHP poles of L(s), zk RHP zeros of L(s)
∫∞
0log pS(iω)pdω = π
∑
Re pk −π
2lim
s→∞sL(s)
◮ Fast poles (and slow zeros) in the RHP are bad!
◮ Useful to let the loop transfer function go to zero rapidly for highfrequencies (high-frequency roll-off) because the last term vanishes
◮ The "water-bed effect". Push the curve down at one frequency and itpops up at another! Design is a compromise!
![Page 36: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/36.jpg)
The Water Bed Effect - Bode’s Integral
0 1 2 3−3
−2
−1
0
1
Frequency ω [rad/s] (linear scale)
logpS(i
ω)p
10
0.1
1.0
Serious Design s.g
Log M
agnit
ude
Frequency0.0 0.5 1.0 1.5 2.0
∫∞
0log pS(iω)pdω = π
∑
Re pk −π
2lim
s→∞sL(s)
The sensitivity can be decreased at one frequency at the cost of increasingit at another frequency.
Feedback design is a trade-off
![Page 37: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/37.jpg)
Robustness
Effect of small process changes dP on closed loop responseT = PC/(1 + PC)
dT
dP=
C
(1 + PC)2=
ST
P,
dT
T= S
dP
P
Effect of large process canges:how much ∆P can the processchange without making the closedloop unstable?
pC∆Pp < p1 + PCp
orp∆Pp
pPp<
1
pT p
∆P must be stable
Nyquist plot
A
B
n
x
y
1 + L
C∆P
![Page 38: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/38.jpg)
Another View of Robustness
A feedback system where the process has multiplicative uncertainty, i.e.P +∆P = P(1 + δ ), where δ = ∆P/P is the relative error, can berepresented with the following block diagrams
P
−C
Σ
δ δ
−PC
1 + PC
The small gain theorem gives the stability condition
pδ p =p∆Pp
pPp<
∣
∣
∣
1 + PC
PC
∣
∣
∣ =1
pT p
Same result as obtained before!
![Page 39: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/39.jpg)
Robustness and Sensitivity
Gain and phase margins
gm ≥Ms
Ms − 1, φm ≥ 2 arcsin
1
2Ms
Constraints on both gain and phase marginscan be replaced by one constraint on maxi-mum sensitivity Ms.
x
y
1/Ms ωms
ω s
−1
◮ Ms = 2 guarantees gm ≥ 2 and φm ≥ 30○
◮ Ms = 1.6 guarantees gm ≥ 2.7 and φm ≥ 36○
◮ Ms = 1.4 guarantees gm ≥ 3.5 and φm ≥ 42○
◮ Ms = 1 guarantees gm =∞ and φm ≥ 60○
![Page 40: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/40.jpg)
When are Two Systems Close?
0 1 2 3 4 50
50
100Open loop
t
y
0 0.1 0.2 0.3 0.4 0.5−1
0
1
2
3Closed loop
t
y
0 0.5 1 1.5 20
100
200
300
400
500Open loop
t
y
0 0.02 0.04 0.06 0.08 0.10
0.5
1
Closed loop
ty
Comparing open loop step responses can be misleading!Open loop frequency responses are slightly better
Much better to compare closed loop responses
![Page 41: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/41.jpg)
Vinnicombe’s Metric
Intuition
![Page 42: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/42.jpg)
Summary of the Sensitivity Functions
S =1
1 + L, T =
L
1 + L, Ms = max pS(iω)p, Mt = max pT(iω)p
The value 1/Ms is the shortest distance from the Nyquist curve of the looptransfer function L(iω) to the critical point −1.
S =� log T
� log P=
Ycl(s)
Yol(s)
How much can the process be changed with stable ∆P without making the closedloop system unstable?
p∆Pp
pPp<
1
pT p
Bode’s integral the water bed effect.
∫∞
0
log pS(iω)pdω = π∑
Re pk −π
2lim
s→∞sL(s)
![Page 43: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/43.jpg)
Requirements and Sensitivity Functions
Disturbances
◮ Effect of feedback: ycl = Syol
◮ Load disturbances: Gyd = PS
◮ Measurement noise: Gun = −CS
Process uncertainty
◮ Small variations: δT/T = SδP/P
◮ Large variations: p∆Pp/pPp ≤ 1/pT p, stable ∆P
◮ Gain and phase and sensitivity margins: gm, φm, sm = 1Ms
Command signal following
◮ Error feedback: Gyr = T , Gur = CS
◮ 2DOF: Gyr = TF , Gur = CSF
![Page 44: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/44.jpg)
Testing Requirements
Introduce test points in the control system!Use test signals in design phase and on the real system!
r e u yv
d n
x
Controller Process
ΣΣ
Σ
ΣΣF(s) C(s) P(s)
−1
i1
i2
o11 o12
o21o22
![Page 45: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/45.jpg)
Feedback Fundamentals
1. Introduction
2. Controllers with Two Degrees of Freedom
3. The Gangs of Four and Seven
4. The Sensitivity Functions
5. Fundamental Limitations
6. Summary
Theme: A closer look at feedback
![Page 46: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/46.jpg)
Fundamental Limitations
Large signal behavior
◮ Limits on actuation: size and rate
◮ Limits due to equipment safety
Small signal behavior
◮ Sensor noise
◮ Resolution of AD and DA converters
◮ Friction
Dynamics
◮ Nonminimum phase dynamics
Right half plane zerosRight half plane poles (instabilities)Time delays
![Page 47: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/47.jpg)
Voice coil drive for a hard disk drive
Jd2φ
dt2= T = kt I
md2x
dt2= F = kt I
r = 0.05 m
J = 5$ 10−6kg m
2
m = 2$ 10−3kg
kt = 2 N/A
Imax = 0.5A
Vmax = 5 V
Maximum acceleration amax =kt Imax
m= 500 m/s2 (50g) Maximum
velocity vmax =Vmax
kt
= 2.5 m/s
![Page 48: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/48.jpg)
Minimum Time Transitions
0 2 4 6 80
0.005
0.01
0 2 4 6 80
1
2
3
0 2 4 6 8
−0.5
0
0.5
0 5 10 150
0.01
0.02
0.03
0 5 10 150
1
2
3
0 5 10 15
−0.5
0
0.5P
ositi
on[m
]
Pos
ition
[m]
Vel
ocity
[m/s
]
Vel
ocity
[m/s
]
Cur
rent
[A]
Cur
rent
[A]
Time [ms]Time [ms]
![Page 49: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/49.jpg)
Limitations due to NMP Dynamics
Process dynamics can impose severe limitations on what can be achieved.
◮ An important part of recognizing the difficult problems
◮ Time delays and RHP zeros limit the achievable bandwidth
◮ Poles in the RHP requires high bandwidth
◮ Systems with poles and zeros in the right half plane can be verydifficult or even impossible to control robustly. Think about the bicyclewith rear wheel steering!
Remedies:
◮ Add sensors and actuators (changes and removes zeros) or redesignthe process
![Page 50: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/50.jpg)
The First IEEE Bode Lecture 1989
Video by IEEEhttp://www.ieeecss-oll.org/video/respect-unstable
Published in IEEE Control Systems Magazine August 2003
![Page 51: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/51.jpg)
Summary of Dynamics Limitations
◮ A RHP zero z limits the achievable gain crossover frequency
ωgc ≤ z
√
Ms − 1
Ms + 1
◮ A RHP pole p requires a high gain crossover frequency
ωgc ≥ p
√
Ms + 1
Ms − 1
◮ A RHP pole-zero pair gives a lower bound to the maximumsensitivities
Ms ≥∣
∣
∣
p + z
p− z
∣
∣
∣, Mt ≥
∣
∣
∣
p + z
p− z
∣
∣
∣
◮ A RHP pole p and a time delay τ gives a lower bound to themaximum sensitivities
Ms ≥ epτ , Mt ≥ e
pτ
![Page 52: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/52.jpg)
Bode Plots – Nonminimum Phase Systems
10−1
100
101
10−2
10−1
100
101
102
10−1
100
101
−300
−200
−100
0
Gai
npG(i
ω)p
ω
Pha
se∠
G(ω
)
Blue curves Bode gain and phase curves of G = GmpGap
Notice that pGap(iω)p = 1Red dashed phase of minimum phase component Gmp
![Page 53: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/53.jpg)
Assessment Plots – Achievable Gain Crossover Frequencies
Process transfer function P(s) = 11+sT
e−s
10−1
100
101
−270
−180
−90
0
ωgc
∠P(i
ω),∠
Pnm
p(i
ω)
I
P
D
Lag dominated T=10
10−1
100
101
−270
−180
−90
0
PI
PID
ωgc
I
P
D
Balanced T=1
10−1
100
101
−270
−180
−90
0
ωgc
∠P(i
ω),∠
Pnm
p(i
ω)
I
P
D
Delay dominated T=0.1
◮ Blue line: phase curve for process transfer function G = GmpGap
◮ Red dashed line: phase curve for allpass factor of process transferfunction (Gap = e−s).
◮ Red full vertical line: phaselag of allpass component of processtransfer function Gap(s) is 60○
◮ Designs for 45○ phase margin for pure I: phase −45○, P: phase−135○ and D: phase −225○ controllers
![Page 54: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/54.jpg)
Consequences for Design
◮ Poles are intrinsic properties of the system
◮ Changing poles requires redesign of the system
◮ Zeros depend on how inputs and outputs are coupled to the states
◮ Zeros and time delays can be changed by moving or adding sensors(and actuators)
A nice property of the FOTD P(s) =K
1 + sTe−sL and SOTD
P(s) =K
(1 + sT1)(1 + sT2)e−sL are that they capture the time delay
which limits the achievable performance. Time delay is influenced bysensor positions!
![Page 55: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/55.jpg)
Stabilizing an Inverted Pendulum with Delay
Right half plane pole at
p =√
g/{
Requiring pτ < 0.33 gives τ√
g/{ < 0.33or
{ >gτ 2
0.33( 100τ 2
A neural lag of 0.07 s gives { > 0.5 mMake a demo!
A vision based system with sampling rate of 50 Hz (a time delay of 0.02 s)can robustly stabilize the pendulum if { > 0.04 m.
![Page 56: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/56.jpg)
Example - The X-29
Advanced experimental aircraft. Much design effort was done with manymethods and much cost. Specifications φm = 45○ could not be reached.
Nonminimum phase factor
Pnmp(s) =s− 26
s− 6
The zero-pole ratio isz
p=
13
3
Ms ≥13 + 3
13− 3= 1.6
φm ≤ 2 arcsin1
2Ms
= 36○
The simple calculation showsthat it is impossible to obtain a phase margin of 45○!
Millions of dollars could have been saved!!
![Page 57: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/57.jpg)
Bicycles with Rear Wheel Steering
Whitt and Wilson Bicycling Science 1982
Many people have seen theoretical advantages in
the fact that front-drive, rear-steered recumbent
bicycles would have simpler transmissions than
rear-driven recumbents and could have the cen-
ter of mass nearer the front wheel than the rear.
The U.S. Department of Transportation commis-
sioned the construction of a safe motorcycle with
this configuration. It turned out to be safe in an
unexpected way: No one could ride it.
The difficulties are caused by the dynamical properties of rear wheelsteering. One reason for learning control is to find such difficulties at anearly stage of the design.
![Page 58: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/58.jpg)
Richard Klein University of IllinoisUsing bicycles in control education for Mechanical Engineers
![Page 59: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/59.jpg)
Bicycle with Rear Wheel Steering
Transfer function regular bike
P(s) =am{V0
bJ
s +V0
a
s2 −mg{
J
Transfer function rear wheelsteering (change sign of V0)
P(s) =am{V0
bJ
−s +V0
a
s2 −mg{
J
RHP pole at√
mg{/J
RHP zero at −V0/a
Klein’s bikes
![Page 60: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/60.jpg)
The Lund Unridable Rear-Steered Bike
![Page 61: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/61.jpg)
The UCSB Ridable Rear-Steered Bike
![Page 62: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/62.jpg)
Feedback Fundamentals
1. Introduction
2. Controllers with Two Degrees of Freedom
3. The Gangs of Four and Seven
4. The Sensitivity Functions
5. Fundamental Limitations
6. Summary
Theme: A closer look at feedback
![Page 63: Feedback Fundamentals · 2020-01-13 · Feedback Fundamentals 1. Introduction 2. Controllers with Two Degrees of Freedom 3. The Gangs of Four and Seven 4. The Sensitivity Functions](https://reader033.fdocuments.in/reader033/viewer/2022042414/5f2eca5d65efe41b6220e447/html5/thumbnails/63.jpg)
Summary
◮ Error feedback and systems with two degrees of freedom 2DOF
2DOF allows separation of command signal response from the otherrequirements
◮ A system with error feedback is characterized by four transferfunctions (Gang of Four GOF) S, T , PS, CS
◮ A system with two degrees of freedom is characterized by seventransfer functions (Gang of Seven = GOF, FS, FT and FCS)
◮ Several transfer functions are required to understand a feedbacksystem. Analysis and specifications should cover all transferfunctions!
◮ There are fundamental limiations caused by nonlinear as well aslinear non-minimum phase dynamics