Datasets:
tensorkelechi
/
arxiv-cstext

Dataset card Data Studio Files Community
1
text
stringlengths
0
89.3k
reads as
Azi Bvi i1M
j0
AΦe
ij BΦν
ijW W
pi iM
j0Σi1jWAB D
and by collecting the singletons we get
Azi Bvipi i1M
j0AΦe
ij BΦν
ijW W
iM
j0Σi1jWFiΣAB D 17
Since the superset of above inclusion is a Minkowski sum
embedding it in an optimization problem is conservative
and nonconvex 30 Therefore we make use of the fol
lowing lemma to convert 17 into a form that can be
formulated linearly in the optimization variables
Lemma 1 LetDd d
Ad
Bdenote the dthvertex
ofDand define
ψd
i d
Azi d
Bvipi 18a
Ψd
ij d
AΦe
ij d
BΦν
ijΣi1j 18b
Then if the following inclusion holds for i 0 N1
andd 1 nD
ψd
i i1M
j0Ψd
ijW W Σi1iW 19
containment condition 17is satisfied
PROOF First note that due to the separation x
zeanduvν the same d
Aand d
Bare applied
to the nominal and error variables Therefore the same
vertex of Dis used in 18a 18b and we can prove the
lemma for any vertex d impyling that the proof holdsfor all d
Ad
B D due to convexity of D We start
by rewriting 17 using 18 as
ψd
ii1M
j0Ψd
ij Σi1jW W iM
j0Σi1jW20
where we used Ψd
ij Σi1j d
AΦe
ij d
BΦν
ij Next
we note that the following inclusion holds
Ψd
ij Σi1jW Ψd
ijW Σi1jW 21
which is a direct consequence of 30 Proposition 2 Fi
nally we show 20 holds by using 21 and commuta
tivity of the Minkowski sum ie
ψd
i i1M
j0Ψd
ij Σi1jW W
21
ψd
i i1M
j0Ψd
ijW W i1M
j0Σi1jW
19
Σi1iW i1M
j0Σi1jW
iM
j0Σi1jW
Hence we have shown that enforcing 19 is sufficient
for 17 to hold 2
Additionally we impose a diagonal structure on Σ ie
Σi1iσiInfori 0 N1 which is similarly done
in 25 Section 41 Using the correspondence to homo
thetic tubes 8 we note that this diagonal structure with
scalings σiis equivalent to the dilation factors αi After
applying Lemma 1 and imposing the diagonal structure
the inclusion 17 for i 0 N1 and d 1 nD
reads as
ψd
i i1M
j0Ψd
ijW W σi1W 22
which can be linearly embedded in an optimization prob