EE263 Autumn 2015 S. Boyd and S.

Lall

Example: Input Design

1
Examples (ideas only, no details)

I let’s consider a specific system

ẋ = Ax, y = Cx
16
with x(t) ∈ R , y(t) ∈ R (a ‘16-state single-output system’)

I model of a lightly damped mechanical system, but it doesn’t matter

2
Typical output

y
0

−1

−2

−3
0 50 100 150 200 250 300 350

t
3

1
y

−1

−2

−3
0 100 200 300 400 500 600 700 800 900 1000

I output waveform is very complicated; looks almost random and unpredictable

3
 t
Modal components 0.2

−0.2
0 50 100 150 200 250 300 350
1

−1
0 50 100 150 200 250 300 350
0.5

−0.5
0 50 100 150 200 250 300 350
2
we’ll see that such a solution can
0
be decomposed into much sim-
−2
pler (modal) components 1
0 50 100 150 200 250 300 350

−1
0 50 100 150 200 250 300 350
2
(idea probably familiar from
0
‘poles’)
−2
0 50 100 150 200 250 300 350
5

−5
0 50 100 150 200 250 300 350
0.2

−0.2
0 50 100 150 200 250 300 350

4
Input design

add two inputs, two outputs to system:

ẋ = Ax + Bu, y = Cx, x(0) = 0

16×2 2×16
where B ∈ R ,C∈R (same A as before)
problem: find appropriate u : R+ → R2 so that y(t) → ydes = (1, −2)

simple approach: consider static conditions (u, x, y constant):

ẋ = 0 = Ax + Bustatic , y = ydes = Cx

solve for u to get:

 

−1 −0.63
ustatic = −CA−1 B ydes =
0.36

5
Simple approach
let’s apply u = ustatic and just wait for things to settle:

0
−0.2

u1
−0.4
−0.6
−0.8
−1
−200 0 200 400 600 800 1000 1200 1400 1600 1800
t
0.4
0.3
u2
0.2
0.1
0
−0.1
−200 0 200 400 600 800 1000 1200 1400 1600 1800

2
t
1.5
y1

0.5

−200 0 200 400 600 800 1000 1200 1400 1600 1800
t
0

−1
y2

−2

−3

−4
−200 0 200 400 600 800 1000 1200 1400 1600 1800
t

. . . takes about 1500 sec for y(t) to converge to ydes

6
Faster convergence
using very clever input waveforms (EE263) we can do much better, e.g.

0.2
0

u1
−0.2
−0.4
−0.6

0 10 20 30 40 50 60
t
0.4
u2
0.2

−0.2
0 10 20 30 40 50 60
t
1

0.5
y1

−0.5
0 10 20 30 40 50 60
t
0
−0.5
y2

−1
−1.5
−2
−2.5
0 10 20 30 40 50 60
t

. . . here y converges exactly in 50 sec

7
Still faster
in fact by using larger inputs we do still better, e.g.
5

u1
0

−5
−5 0 5 10 15 20 25
t
1

u2 0.5
0
−0.5
−1
−1.5
−5 0 5 10 15 20 25

2
t
1
y1

−1

−5 0 5 10 15 20 25
t
0

−0.5
y2

−1

−1.5

−2
−5 0 5 10 15 20 25
t

. . . here we have (exact) convergence in 20 sec

8
in this course we’ll study

I how to synthesize or design such inputs

I the tradeoff between size of u and convergence time