Datasets:
tensorkelechi
/
arxiv-cstext

Dataset card Data Studio Files
xet
Community
1
text
stringlengths
0
89.3k
we can use the following lemma from 28 to implement
them as linear constraints
Lemma 7 Lemma 1 in 28 Consider scalars α
0β0 matrices ARnnΓRnn polytopic
setsXxHxxhxwith HxRnxnand
YyHyyhywithHuRnyn and a compact
convex set Z Rn then the following conditions are
sufficient and necessary for
αAX βY ΓZ A1
to hold
ΛRnynxΛ0 A2a
ΛHxαHyA A2b
ΛhxβhyhZΓH
y A2c
where hZΓH
y h
hZΓH1
yhZΓHny
yi
denotes the stacked support functions individually eval
uated at the rows of Hy
PROOF The proof can be found in 28 Lemma 1 2
The disturbance inclusion constraints 32k 32l can
both be rewritten in the form A3 by applying 31
Theorem 21 and thus can be implemented as linear
constraints using Lemma 8
Lemma 8 Consider a scalar β0 a vector aRn
matrices AiRnnΓRnn polytopic sets Xix
HxixhxiwithHxiRnxnandYyHyy
hywithHuRnyn and a compact convex set Z Rnthen the following conditions are sufficient and necessary
for
a N1M
i0AiXiβY ΓZ A3
to hold
ΛiRnynxΛi0 i 0 N1A4a
ΛiHxiHyAi i 0 N1A4b
N1X
i0ΛihxiβhyhZΓH
yHya A4c
where hZΓH
y h
hZΓH1
yhZΓHny
yi
denotes the stacked support functions individually eval
uated at the rows of Hy
PROOF Using standard properties of polytopic sets
we define the scaled polytope βYyHyyβhy
Then using Theorem 23 in 31 we get
βY ΓZyHyyβhyhZΓH
y
Therefore A3 is a polytope containment problem
which can be restated as A4 using 30 Prop 1 2
If the set Yin Lemma 8 is restricted to a hyperrect
angle then it can be specialized to βbeing a diago
nal matrix ie β diag β1 βny This is due to
the fact that for hyperrectangle Yand diagonal ma
trixβthe representation βYyHyyˆβhywith
ˆβ diag β1In βnyIn avoids inverting βand thus
still allows optimization over β We can use this to re
lax the structural constraint Σ i1iσInto Σ i1i
diag σi1 σin as discussed in Remarks 1 2 Addi
tionally if Wis a hyperrectangle we recover the distur
bance tube parameterization in 25 as a special case
A2 Implementation of 46
Since the sets ZiVi andQd
iin 46 are all computed
via the Pontryagin difference of a constant polytopic
and a variable polytopic set see 43 44 they share
the same shape 31 Theorem 23 For example the sets
Zm
izHxzhm
zall have the shape of state con
straint Xbut differ in their sizes hm
zfor different memory
entries Using this observation constraints 46d 46g
can be implemented using the following lemma
Lemma 9 Consider scalars am0and polytopic sets
XmxHxxhm
xwith HxRnxnform
0 M1 then the following set description holds
M1M
m0amXmxHxxM1X
m0amhm
x A5
19PROOF Using standard properties of polytopic sets
we get the scaled polytope amXmxHxxamhm
x
Then using the definition of the Minkowski sum we get
a1X1a2X2x1x2Hxx1a1h1
x Hxx2a2h2
x
Substituting xx1x2 we get the halfspace represen
tation of Xa1X1a2X2as
XxHxxHxx1x2a1h1
xa2h2
x
Applying this recursively for all m 0 M1 then
yields A5 2
B Experimental Details