Marcus Völp
Constructing and Verifying Cyber Physical Systems
Effect of Poles and Zeros / PID Control
Summer 2015 Constructing and Verifying Cyber Physical Systems - Marcus Völp 109
Overview
Math
FeedbackControl
RTOS
Verification
Physics
Introduction
Mathematical Foundations (Differential Equations and Laplace Transformation)
Control and Feedback
Transfer Functions and State Space Models
Poles, Zeros / PID Control
Stability, Root Locust Method, Digital Control
Mixed-Criticality Scheduling and Real-Time Operating Systems (RTOS)
Program Verification
Differential Dynamic Logic and KeYmaera X
Differential Invariants
Summer 2015 Constructing and Verifying Cyber Physical Systems - Marcus Völp 110
Overview
Control
Regulating Control
Open Loop Control
Tracking Control
Closed Loop Control
Noise
Observability
Sensor Fusion
Block Diagrams / Simulink
Robustness
Stability / Final Value Theorem
K. Åström, R. Murray“Feedback Systems –An Introduction for Scientists and Engineers”Princeton University Press
G. Franklin, J.D. Powell, A. Emami-Naeini“Feedback Control of Dynamic Systems”7th ed. Pearson
Control Equations
Summer 2015 Constructing and Verifying Cyber Physical Systems - Marcus Völp 111
Overview
Observability
Sensor Fusion
Robustness
Stability
K. Åström, R. Murray“Feedback Systems –An Introduction for Scientists and Engineers”Princeton University Press
G. Franklin, J.D. Powell, A. Emami-Naeini“Feedback Control of Dynamic Systems”7th ed. Pearson
PID Control
Rise Time / Overshoot / Settling
Effect of Poles and Zeros
Summer 2015 Constructing and Verifying Cyber Physical Systems - Marcus Völp 112
Effect of Poles and Zeros
𝐻 𝑠 =𝑎𝑛𝑠
𝑛 + 𝑎𝑛−1𝑠𝑛−1 + …+ 𝑎1𝑠 + 𝑎0
𝑎𝑚𝑠𝑚 + 𝑏𝑚−1𝑠𝑚−1 + …+ 𝑏1𝑠 + 𝑏0
= 𝐾 𝑖=1
𝑛 𝑠 + 𝑧𝑖
𝑗=1𝑚 𝑠 + 𝑝𝑗
Gain
Transfer function is Laplace transform of impulse responsewe can directly compute the impulse response from the
poles and zeroes of the transfer function call impulse response the natural response of the system
Modes
Summer 2015 Constructing and Verifying Cyber Physical Systems - Marcus Völp 113
Effect of Real-Valued Poles
𝐻 𝑠 =1
(𝑠 + 𝜎)
ℎ 𝑡 = 𝑒−𝜎𝑡1(𝑡)
1.0
1.0
1
𝑒
impulse response
step response
𝐻 𝑠 =2𝑠 + 1
𝑠2 + 3𝑠 + 2
= 2𝑠 +
12
𝑠 + 1 𝑠 + 2
partial fraction expansion
xx o
Im(s)
Re(s)1-1-2
Summer 2015 Constructing and Verifying Cyber Physical Systems - Marcus Völp 114
Partial Fraction Expansion
real-valued function where 𝑄 𝑥 = 𝑥 − 𝑥0 … 𝑥 − 𝑥𝑚 , degree 𝑃 𝑥 < 𝑚
• for each 𝑟𝑖-fold real pole 𝑥 − 𝑥𝑖𝑟𝑖, add the following polynomial to 𝐴 𝑥 , using fresh constants:
𝑅 𝑥 =𝑃(𝑥)
𝑄(𝑥)
𝑎𝑖1𝑥 − 𝑥𝑖
+𝑎𝑖2
𝑥 − 𝑥𝑖2+ …+
𝑎𝑖𝑟𝑖𝑥 − 𝑥𝑖
𝑟𝑖
Suppose 𝑅 𝑥 = 𝐴 𝑥 where 𝐴 𝑥 is constructed as follows:
• because 𝑅(𝑥) is a real-valued function, both 𝑧𝑖 and 𝑧𝑖 appear as poles;for each 𝑠𝑖-fold complex pole 𝑥 − 𝑧𝑖
𝑠𝑖 , add the following to 𝐴 𝑥 :
𝑏𝑖1𝑥 + 𝑐𝑖1𝑥2 + 𝑝𝑖𝑥 + 𝑞𝑖
+𝑏𝑖2𝑥 + 𝑐𝑖2
𝑥2 + 𝑝𝑖𝑥 + 𝑞𝑖2+ …+
𝑏𝑖𝑠𝑖𝑥 + 𝑐𝑖𝑠𝑖𝑥2 + 𝑝𝑖𝑥 + 𝑞𝑖
𝑠𝑖
where 𝑥2+𝑝𝑖𝑥 + 𝑞𝑖 = 𝑥 − 𝑧𝑖 𝑥 − 𝑧𝑖
• solve by comparison of the coefficients
Summer 2015 Constructing and Verifying Cyber Physical Systems - Marcus Völp 115
Effect of Real-Valued Poles
𝐻 𝑠 =1
(𝑠 + 𝜎)
ℎ 𝑡 = 𝑒−𝜎𝑡1(𝑡)
1.0
1.0
1
𝑒
impulse response
step response
𝐻 𝑠 =2𝑠 + 1
𝑠2 + 3𝑠 + 2
= 2𝑠 +
12
𝑠 + 1 𝑠 + 2
𝐻(𝑠) = −1
𝑠 + 1+
3
𝑠 + 2
partial fraction expansion
𝐻 𝑠 = 2𝑠 +
12
𝑠 + 1 𝑠 + 2=
𝑎
𝑠 + 1+
𝑏
𝑠 + 2
2𝑠 + 1 = 𝑎𝑠 + 2𝑎 + 𝑏𝑠 + 𝑏
2𝑠 = 𝑎𝑠 + 𝑏𝑠 1 = 2𝑎 + 𝑏
𝑎 = −1, 𝑏 = 3
ℎ 𝑡 = −𝑒−𝑡 + 3𝑒−2𝑡 for 𝑡 ≥ 0
Summer 2015 Constructing and Verifying Cyber Physical Systems - Marcus Völp 116
Effect of Real-Valued Poles
𝐻 𝑠 =1
(𝑠 + 𝜎)
ℎ 𝑡 = 𝑒−𝜎𝑡1(𝑡)
1.0
1.0
1
𝑒
impulse response
step response
𝐻 𝑠 =2𝑠 + 1
𝑠2 + 3𝑠 + 2
= 2𝑠 +
12
𝑠 + 1 𝑠 + 2
𝐻(𝑠) = −1
𝑠 + 1+
3
𝑠 + 2
partial fraction expansion
ℎ 𝑡 = −𝑒−𝑡 + 3𝑒−2𝑡 𝑓𝑜𝑟 𝑡 ≥ 00 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
decays faster than 𝑒−𝑡
=> say: “Pole is faster”
impulse step
Summer 2015 Constructing and Verifying Cyber Physical Systems - Marcus Völp 117
Effect of Compex Poles
𝐻 𝑠 =𝜔𝑛
2
𝑠2 + 2𝜉𝜔𝑛𝑠 + 𝜔𝑛2
Because H(s) is real valued, complex poles always occur as conjugate pairs 𝑝𝑖 and 𝑝𝑖
Characterize complex poles in terms of their real and imaginary parts: 𝑠 = −𝜎 ± 𝑗𝜔𝑑
denominator: 𝑏(𝑠) = (𝑠 + 𝜎 − 𝑗𝜔𝑑) (𝑠 + 𝜎 + 𝑗𝜔𝑑) in 𝐻 𝑠 =𝑎(𝑠)
𝑏(𝑠)
= 𝑠 + 𝜎 2 + 𝜔𝑑2
= 𝑠2 + 2𝜎𝑠 + 𝜎2 + 𝜔𝑑2
Write result when finding transfer function of second order differential equation as:
𝜎 = 𝜉𝜔𝑛 𝜔𝑑 = 𝜔𝑛 1 − 𝜉2
damping ratio undamped natural frequency
Summer 2015 Constructing and Verifying Cyber Physical Systems - Marcus Völp 118
Effect of Compex Poles
𝐻 𝑠 =𝜔𝑛
2
𝑠2 + 2𝜉𝜔𝑛𝑠 + 𝜔𝑛2
(𝑠 + 𝜎 − 𝑗𝜔𝑑) (𝑠 + 𝜎 + 𝑗𝜔𝑑)
𝜎 = 𝜉𝜔𝑛 𝜔𝑑 = 𝜔𝑛 1 − 𝜉2
x
x
𝜔𝑛
𝜃 = sin−1 𝜉
𝜎
𝜔𝑑
=𝜔𝑛
2
(𝑠 + 𝜉𝜔𝑛)2+𝜔𝑛
2(1 − 𝜉2)
ℎ 𝑡 =𝜔𝑛
1 − 𝜉2𝑒−𝜎𝑡 sin𝜔𝑑𝑡 1(𝑡)
𝑒−𝜎𝑡
−𝑒−𝜎𝑡
Summer 2015 Constructing and Verifying Cyber Physical Systems - Marcus Völp 119
Effect of Zeros
𝐻 𝑠 =𝑠 + 𝑎
𝑏(𝑠)where a, is not a pole of 𝑏(𝑠)
plant𝑢(𝑡)
𝐻 𝑠ℒ 𝑈(𝑠) ∙
𝑢 𝑡 = 𝑒−𝑎𝑡 ℒ 𝑒−𝑎𝑡 =1
𝑠+𝑎𝑌 𝑠 =
1
(𝑠 + 𝑎)
𝑠 + 𝑎
𝑏(𝑠)=
1
𝑏(𝑠)
Zeros cancel frequencies; they are transmission blocking(unless there is also a pole)
Summer 2015 Constructing and Verifying Cyber Physical Systems - Marcus Völp 120
Effect of Real-Valued Zeros
𝐻1 𝑠 =2
(𝑠 + 1)(𝑠 + 2)
=2
(𝑠 + 1)−
2
(𝑠 + 2)
𝐻1 𝑠 =2(𝑠 + 1.1)
1.1(𝑠 + 1)(𝑠 + 2)
=2
1.1
0.1
(𝑠 + 1)+
0.9
(𝑠 + 2)=
0.18
(𝑠 + 1)+
1.64
(𝑠 + 2)
xx o-1-2
Zeros at pole location cancel the mode defined by the pole.Zeros near a pole damp the effect.
impulse
step
impulse
step
Summer 2015 Constructing and Verifying Cyber Physical Systems - Marcus Völp 121
Effect of Complex Zeros
𝐻 𝑠 =
1𝑎𝜉
𝑠 + 1
𝑠2 + 2𝜉𝑠 + 1Pair of complex poles at 𝜔𝑛 = 1, 𝜃 = sin−1 𝜉
e.g., 𝜉 = 0.5 and 𝜃 = 30°
x
x
-1
𝐻 𝑠 =1
𝑎𝜉
𝑠 + 𝑎𝜉
𝑠2 + 2𝜉𝑠 + 1
o
𝑎 = 1
𝑎 = 2𝑎 = 4
Nearby zeros cause a faster rise at the cost of higher overshoot.
Im(s)
Re(s)
Summer 2015 Constructing and Verifying Cyber Physical Systems - Marcus Völp 122
Effect of Complex Zeros
𝐻 𝑠 =
1𝑎𝜉
𝑠 + 1
𝑠2 + 2𝜉𝑠 + 1
=1
𝑠2 + 2𝜉𝑠 + 1+
1
𝑎𝜉
𝑠
𝑠2 + 2𝜉𝑠 + 1Recall: ℒ
𝜕𝑦
𝜕𝑡= 𝑠𝐻 𝑠 , where ℒ 𝑦 = 𝐻 𝑠
= 𝐻0(𝑠) +1
𝑎𝜉𝑠𝐻0(𝑠)
𝑦(𝑡) = 𝑦0(𝑡) +1
𝑎𝜉
𝜕𝑦0𝜕𝑡
𝑦(𝑡)
𝑦0(𝑡)
1
𝑎𝜉
𝜕𝑦0𝜕𝑡
Summer 2015 Constructing and Verifying Cyber Physical Systems - Marcus Völp 123
Time Domain Specification
0,9
0,1
settling time: 𝑡𝑠time after which signal remains within epsilon distance around reference
peak time: 𝑡𝑝 - time of maximum overshoot
overshoot: 𝑀𝑝 - maximum over reference signal
rise time: 𝑡𝑟 - time to reach vicinity of settling point
Second order system
rise time 𝜔𝑛 ≥1.8
𝑡𝑟
overshoot 𝑀𝑝 = 𝑒
−𝜋𝜉
1−𝜉2
=> 𝜉 ≥ 𝜉(𝑀𝑝)
settling time 𝜎 ≥4.6
𝑡𝑠(decay below 1%)
𝐻 𝑠 =𝜔𝑛
2
𝑠2 + 2𝜉𝜔𝑛𝑠 + 𝜔𝑛2
𝜎 = 𝜉𝜔𝑛 𝜔𝑑 = 𝜔𝑛 1 − 𝜉2
Summer 2015 Constructing and Verifying Cyber Physical Systems - Marcus Völp 124
Time Domain Specification
Second order system
rise time 𝜔𝑛 ≥1.8
𝑡𝑟
overshoot 𝑀𝑝 = 𝑒
−𝜋𝜉
1−𝜉2
=> 𝜉 ≥ 𝜉(𝑀𝑝)
settling time 𝜎 ≥4.6
𝑡𝑠(decay below 1%)
𝐻 𝑠 =𝜔𝑛
2
𝑠2 + 2𝜉𝜔𝑛𝑠 + 𝜔𝑛2
𝜎 = 𝜉𝜔𝑛 𝜔𝑑 = 𝜔𝑛 1 − 𝜉2
Im(s)
Re(s)
Im(s)
Re(s)
Im(s)
Re(s)x
x
x
sin−1 𝜉
𝜔𝑛
𝜎
rise time overshoot settling timeIm(s)
Re(s)
x
x
x
x
So far:
• by adding zeros,
• we can block / damp certain frequencies
• we can eliminate / damp modes if the zeros are near poles
• we can influence rise time at cost of overshoot
• by adding roots,
• we can add sinusoids and exponentials to the response
• we can influence overshoot, settling time and rise time
Summer 2015 Constructing and Verifying Cyber Physical Systems - Marcus Völp 125
Control
Control: add roots and zeros to obtain desired plant behavior.
Summer 2015 Constructing and Verifying Cyber Physical Systems - Marcus Völp 126
Exercise: Tracking and Pole Placement
Compute the coefficients 𝑐𝑖 , 𝑑𝑖 such that the closed loop has the characteristic equation 𝑠 + 6 𝑠 + 3 𝑠2 + 3𝑠 + 9 = 0.
(characteristic equation is the denominator of the closed-loop transfer function set to zero)
Controller𝑐2𝑠
2 + 𝑐1𝑠 + 𝑐0𝑠(𝑠 + 𝑑1)
Plant1
𝑠2 + 3𝑠 + 9
𝑅(𝑠) 𝑌(𝑠)Σ+
+
𝑊(𝑠)
disturbance
reference signal output
𝑈(𝑠)Σ
Σ
+
-
++
𝑉(𝑠)
sensor noise
𝐸(𝑠)
𝑌(𝑠)
𝑅(𝑠)= 𝑇(𝑠) = 𝐺𝐷𝑆 =
𝐺𝐷
1 + 𝐺𝐷
1 + 𝐺𝐷 = 0
1 +𝑐2𝑠
2 + 𝑐1𝑠 + 𝑐0𝑠(𝑠 + 𝑑1)
1
𝑠2 + 3𝑠 + 9= 0
Summer 2015 Constructing and Verifying Cyber Physical Systems - Marcus Völp 127
Exercise: Tracking and Pole Placement
𝑠 + 6 𝑠 + 3 𝑠2 + 3𝑠 + 9 = 0
Controller𝑐2𝑠
2 + 𝑐1𝑠 + 𝑐0𝑠(𝑠 + 𝑑1)
Plant1
𝑠2 + 3𝑠 + 9
𝑅(𝑠) 𝑌(𝑠)Σ+
+
𝑊(𝑠)
disturbance
reference signal output
𝑈(𝑠)Σ
Σ
+
-
++
𝑉(𝑠)
sensor noise
𝐸(𝑠)
1 +𝑐2𝑠
2 + 𝑐1𝑠 + 𝑐0𝑠(𝑠 + 𝑑1)
1
𝑠2 + 3𝑠 + 9= 0
𝑠 𝑠 + 𝑑1 𝑠2 + 3𝑠 + 9 + 𝑐2𝑠2 + 𝑐1𝑠 + 𝑐0 = 0
𝑠4 + 12𝑠3 + 54𝑠2 + 135 𝑠 + 162 = 0
𝑠4 + (𝑑1 +3)𝑠3 + 𝑐2 + 3𝑑1 + 9 𝑠2 + 9𝑑1 + 𝑐1 𝑠 + 𝑐0 = 0
𝑑1 + 3 = 12
𝑐2 + 3𝑑1 + 9 = 54
9𝑑1 + 𝑐1 = 135
𝑐0 = 162
𝑑1 = 9, 𝑐2 = 18, 𝑐1 = 54, 𝑐0 = 162
Summer 2015 Constructing and Verifying Cyber Physical Systems - Marcus Völp 128
PID Control
Controller𝐷𝑐𝑙𝑜𝑠𝑒𝑑(𝑠)
Plant𝐺(𝑠)
𝑅(𝑠) 𝑌(𝑠)Σ+
+
𝑊(𝑠)
disturbance
reference signal output
𝑈(𝑠)Σ
Σ
+
-
++
𝑉(𝑠)
sensor noise
𝑌 = 𝑇𝑅 = 𝐺𝐷𝑆𝑅 = 𝐺𝐷1
1 + 𝐺𝐷R
𝐸 = 𝑅 − 𝑌 = 𝑆𝑅 =1
1 + 𝐺𝐷𝑅
P: feedback proportional to error 𝑢 𝑡 = 𝑘𝑝𝑒(𝑡)
𝐷𝑐𝑙𝑜𝑠𝑒𝑑 𝑠 =𝑈(𝑠)
𝐸(𝑠)= 𝑘𝑝
purely algebraic; no dynamics proportional gain
𝐸(𝑠)
Summer 2015 Constructing and Verifying Cyber Physical Systems - Marcus Völp 129
PID Control
Controller𝑘𝑝
Plant𝐺(𝑠)
𝑅(𝑠) 𝑌(𝑠)Σ+
+
𝑊(𝑠)
disturbance
reference signal output
𝑈(𝑠)Σ
Σ
+
-
++
𝑉(𝑠)
sensor noise
𝑌(𝑠) =𝑘𝑝𝐺(𝑠)
1 + 𝑘𝑝𝐺(𝑠)R(s)
P: feedback proportional to error 𝑢 𝑡 = 𝑘𝑝𝑒(𝑡)
e.g. second order plant: 𝐺 𝑠 =𝐴
𝑠2+𝑎1𝑠+ 𝑎2
𝐸(𝑠)
characteristic equation: 1 + 𝑘𝑝𝐺 𝑠 = 0
𝑠2 + 𝑎1𝑠 + 𝑎2 + 𝑘𝑝𝐴 = 0
control natural frequencyno control over damping
𝑎1 = 1.4, 𝑎2 = 𝐴 = 1
𝑘𝑝 = 1.5
𝑘𝑝 = 6
improve tracking and rise time,but decrease damping (2nd order only)
higher order: increase damping for some poles; increase for others
Summer 2015 Constructing and Verifying Cyber Physical Systems - Marcus Völp 130
PID Control
Controller𝑘𝐼𝑠
Plant𝐺(𝑠)
𝑅(𝑠) 𝑌(𝑠)Σ+
+
𝑊(𝑠)
disturbance
reference signal output
𝑈(𝑠)Σ
Σ
+
-
++
𝑉(𝑠)
sensor noise
I: integral feedback 𝑢 𝑡 = 𝑘𝑖 𝑡0𝑡𝑒(𝑡)𝜕𝑡
𝐸(𝑠)
Infinite value of control with zero system errorConsider all past values
=> cancel constant disturbances zero error
𝐷𝑐𝑙𝑜𝑠𝑒𝑑 𝑠 =𝑈(𝑠)
𝐸(𝑠)=
𝑘𝐼𝑠
integral gain𝑎1 = 1.4, 𝑎2 = 𝐴 = 1, 𝑘𝐼 = 0.5
Summer 2015 Constructing and Verifying Cyber Physical Systems - Marcus Völp 131
PID Control
Plant𝐺(𝑠)
𝑅(𝑠) 𝑌(𝑠)Σ+
+
𝑊(𝑠)
disturbance
reference signal output
𝑈(𝑠)Σ
Σ
+
-
++
𝑉(𝑠)
sensor noise
𝐸(𝑠)
𝐸(𝑠)
𝑅(𝑠)=
1
1 +𝑘𝐼𝑠 𝐺(𝑠)
=𝑠
𝑠 + 𝑘𝐼𝐺(𝑠)
𝑈(𝑠)
𝑅(𝑠)=
𝑘𝐼𝑠
1 +𝑘𝐼𝑠 𝐺(𝑠)
=𝑘𝐼
𝑠 + 𝑘𝐼𝐺(𝑠)
𝑌(𝑠)
𝑅(𝑠)=
𝑘𝐼𝑠 𝐺(𝑠)
1 +𝑘𝐼𝑠 𝐺(𝑠)
=𝑘𝐼𝐺(𝑠)
𝑠 + 𝑘𝐼𝐺(𝑠)
Controller𝑘𝐼𝑠
error
control
response
𝑒 ∞ =0
0 + 𝑘𝐼𝐺(0)= 0
𝑢 ∞ =𝑘𝐼
0 + 𝑘𝐼𝐺(0)= 𝐺−1 0 = 1
𝑦 ∞ =𝑘𝐼𝐺(0)
0 + 𝑘𝐼𝐺(0)= 1
𝑟 𝑡 = 1 𝑡
𝑅 𝑠 =1
𝑠
final value thm.
2nd order 𝑎1 = 𝐴 = 0
Integral control is robustagainst plant variation
Summer 2015 Constructing and Verifying Cyber Physical Systems - Marcus Völp 132
PID Control
Plant𝐺(𝑠)
𝑅(𝑠) 𝑌(𝑠)Σ+
+
𝑊(𝑠)
disturbance
reference signal output
𝑈(𝑠)Σ
Σ
+
-
++
𝑉(𝑠)
sensor noise
𝐸(𝑠) Controller…+ 𝑘𝐷 𝑠
D: derivative feedback 𝑢 𝑡 = …+ 𝑘𝐷𝜕𝑒(𝑡)
𝜕𝑡
𝐷𝑐𝑙𝑜𝑠𝑒𝑑 𝑠 =𝑈(𝑠)
𝐸(𝑠)= ⋯+ 𝑘𝐷𝑠
Derivative supplies no information about desired end state=> use in combination with other controllers (e.g., P, PI)
Add derivative control for increased stability
Summer 2015 Constructing and Verifying Cyber Physical Systems - Marcus Völp 133
PID Control
Plant𝐺(𝑠)
𝑅(𝑠) 𝑌(𝑠)Σ+
+
𝑊(𝑠)
disturbance
reference signal output
𝑈(𝑠)Σ
Σ
+
-
++
𝑉(𝑠)
sensor noise
𝐸(𝑠) Controller
𝑘𝑝 +𝑘𝐼𝑠+ 𝑘𝐷𝑠
PID: derivative feedback 𝑢 𝑡 = 𝑘𝑝𝑒 𝑡 + 𝑘𝐼 𝑡0𝑡𝑒(𝜏)𝜕𝜏 + 𝑘𝐷
𝜕𝑒(𝑡)
𝜕𝑡
𝐷𝑐𝑙𝑜𝑠𝑒𝑑 𝑠 =𝑈(𝑠)
𝐸(𝑠)= 𝑘𝑝 +
𝑘𝐼𝑠+ 𝑘𝐷𝑠
𝐽𝑚 = 1.13 × 10−2 𝑁 ∙ 𝑚 ∙ 𝑠𝑒𝑐2/𝑟𝑎𝑑
DC Motor: Θ𝑚(𝑠)
𝑉𝑎(𝑠)=
𝐾
𝑠(𝜏𝑠+1)with 𝐾 =
𝐾𝑡
𝑏𝑅𝑎+𝐾𝑡𝐾𝑒, 𝜏 =
𝑅𝑎𝐽𝑚
𝑏𝑅𝑎+𝐾𝑡𝐾𝑒
𝑏 = 0.028 𝑁 ∙ 𝑚 ∙ 𝑠𝑒𝑐/𝑟𝑎𝑑
𝑅𝑎 = 0.45Ω
𝐾𝑡 = 0.067 𝑁 ∙ 𝑚/amp 𝐾𝑒 = 0.067 𝑉 ∙ 𝑠𝑒𝑐/rad
Summer 2015 Constructing and Verifying Cyber Physical Systems - Marcus Völp 134
PID Control
P
PI
PID
step input step disturbance𝑘𝑝 = 3
𝑘𝐼 = 9
𝑘𝐷 = 0.3
Summer 2015 Constructing and Verifying Cyber Physical Systems - Marcus Völp 135
Summary
Observability
Sensor Fusion
Robustness
Stability
K. Åström, R. Murray“Feedback Systems –An Introduction for Scientists and Engineers”Princeton University Press
G. Franklin, J.D. Powell, A. Emami-Naeini“Feedback Control of Dynamic Systems”7th ed. Pearson
PID Control
Rise Time / Overshoot / Settling
Effect of Poles and Zeros
Summer 2015 Constructing and Verifying Cyber Physical Systems - Marcus Völp 136
Leftovers
Stability• How to derive stability from location of zeros and poles in s-plane?• How to move zeros and poles by adding control?
Sensitivity• Feedback control systems are much less error prone
to plants changing their gain as open loop systems.
Summer 2015 Constructing and Verifying Cyber Physical Systems - Marcus Völp 137
Overview
Math
FeedbackControl
RTOS
Verification
Physics
Introduction
Mathematical Foundations (Differential Equations and Laplace Transformation)
Control and Feedback
Transfer Functions and State Space Models
Poles, Zeros / PID Control
Stability, Root Locust Method, Digital Control
Mixed-Criticality Scheduling and Real-Time Operating Systems (RTOS)
Coordinating Networked Cyber-Physical Systems
Program Verification
Differential Dynamic Logic and KeYmaera X
Differential Invariants
CPS