text
stringlengths
0
89.3k
2 runs every time the secondary process termi
nates
3 MkMk1
4 ifMk not full then
5 mindex of first empty memory slot
6 MmkMnew
7 else
8 determine a memory slot except m 0
9 mFMk
10 MmkMnew
52 Primary Process Nominal Trajectory Optimiza
tion
Since a selection of tightened state and input constaints
disturbance tubes and terminal sets is stored in mem
12oryM the primary process can use this information to
optimize the nominal trajectories However rather than
using a single memory entry Mm all entries are fused
using a convex combination similar to 2833 ensuring
all available information is optimally used The primary
process is formulated as
min
zvpλlfzN N1X
i0lzi vi 46a
sti 0 N1
z0xk 46b
zi1AziBvipi 46c
ziM1M
m0λmZm
i 46d
viM1M
m0λmVm
i 46e
zNM1M
m0λmαmZf 46f
ψd
iM1M
m0λmQdm
id 1 nD46g
λm0M1X
m0λm 1 46h
where λ λ0λM1 are the convex combination
variables the costs lf l fulfill Assumption 1 ψd
iis
defined as in 18 and Zm
iVm
iQdm
i αmare retrieved
from memory slot Mm The resulting MPC control law is
given by
κprimary xk v
0 47
where v
0is the first element of the optimal solution v
of 46
Since the tubes computed by the secondary process are
more general than those used in 28 we cannot use the
proof in 28 to show recursive feasibility of the primary
process Therefore we additionally store the previous
solution of 46 in memory slot M0as a fallback solution
Specifically we store the shifted convex combination of
tubes after the primary process terminates ie for all
d 1 nDandi 0 N1 we compute
Zprev
iM1M
m0λ
mX F iΦem
shift 48a
Vprev
iM1M
m0λ
mU F iΦνm
shift 48b
Qdprev
i M1M
m0λ
m
σm
i2W F iΨdm
shift
48cwhere the shifted system responses are computed as
Φem
shift shift ΦemΦem
N 49a
Φνm
shift shift ΦνmΦνm
N 49b
with the shift operator defined in Section 21 and
ΦNdenoting the last block row of ΦemandΦνm The
shifted auxiliary variable Ψdm
shiftis computed with 18
using the shifted system responses and the shifted dis
turbance filter ie
Σm
shift shift ΣmΞm
1σm
NIn 50a
Ξm
shiftΞm 50b
where Ξm
1 Ξm