Datasets:
tensorkelechi
/
arxiv-cstext

Dataset card Data Studio Files Community
1
text
stringlengths
0
89.3k
1xk1AxkBκMPCxkp
034
Therefore dynamics 1 can be equivalently rewritten as
xk 1 Axk BκMPCxk p
0σ
1wk35Using this equivalence we show recursive feasibility
of 32 in the following theorem
Theorem 3 Optimization problem 32 is recursively
feasible for system 1under controller 33 ie the fea
sible set of 32is RPI
PROOF We prove the theorem using a standard ar
gument 3 constructing a feasible candidate sequence
based on the shifted previous solution which is possi
ble contrary to 25 because of the novel terminal con
straints We start the proof by stating constraint 32d
in elementwise fashion ie
Φe
ii1 Σ ii1σiIn i 1 N 36a
Φe
i1jAΦe
ijBΦν
ijΣi1j i 1 N1
j 0 i136b
With these and the computed auxilliary distur
bance wk given by 34 we define the scalarvalued
candidate as ˆ αα the vectorvalued candidates as
ˆziz
i1 Φe
i10wk i 0 N137a
ˆzNAKfz
N Γwk 37b
ˆviv
i1 Φν
i10wk i 0 N237c
ˆvN1Kfz
N Φν
N0wk 37d
ˆpip
i1 Σ
i20wk i 0 N237e
ˆpN1 Ξ
0wk 37f
and the matrixvalued candidates as
ˆΦe shift ΦeΦe
N 38a
ˆΦν shift ΦνΦν
N 38b
ˆΣ shift ΣΞ
1σ
NIn 38c
ˆΞΞ 38d
where the shift operator is defined in Section 21
ΦNdenotes the last block row of ΦeandΦν and
Ξ
1 Ξ
1 Ξ
N1 The matrixvalued candidates are
thus constructed by shifting the optimal solutions up
and left by one block row and block column respec
tively and by appending the last block row of ΦeΦν
and Ξ
1 Ξ
N1σ
NIn respectively Note that due
to the choice of these candidates we get ˆ σiσ
i1 i
0 N1ˆσNσ
N and ˆΓΓAΦe
N0BΦν
N0Ξ
0
by definition Next we show that candidates 37 ful
fill nominal dynamics 32c error dynamics 32d and
the system response constraint 32e Using 37a with
i 0 the state candidate sequence is initialized with
ˆz0z
1 Φe
10wk Az
0Bv
0p
0σ
1wk
Axk BκMPCxk ηk xk 1
8where we used Φe
10σ
1Inby 36a nominal dynam
ics 32c and 35 For the remainder of the state can
didate sequence ie i 1 N1 it holds that
ˆziz
i1 Φe
i10wk
32c
36bAz
iBv
ip
i
AΦe
i0BΦν
i0 Σ
i10