A.

8emµorud Model Predlctlve Control µ ¡
Eybrld Systems: Modellng
A. 8emµorud Model Predlctlve Control µ z
!"#$%&'("()*+(
Eybrld systems
Computer
Science
Control
1heory
Finite
state
machines
Contlnuous
dynumlcul
systems
system
u(t) y(t)
!
"
#
#
!
"
"
#
A. 8emµorud Model Predlctlve Control µ ¸
contlnuous dynumlcul system
dlscrete lnµuts
Lmbedded systems
symbols symbols
contlnuous
stutes
contlnuous lnµuts
uutomuton j loglc
lnterfuce
¬ Automoblles
¬ lndustrlul µrocesses
¬ Consumer electronlcs
¬ Eome uµµllunces
¬...
Lxumµle:
A. 8emµorud Model Predlctlve Control µ µ
"lntrlnslcully hybrld' systems
¬ 1runsmlsslon
dlscrete commund
(R,N,¡,z,¸,µ,¸)
¬ lourstroke englnes
Automuton,
deµendent on
µower truln motlon
contlnuous
dynumlcul vurlubles
(velocltles, torques)
÷
A. 8emµorud Model Predlctlve Control µ ¸
Lxumµle of hybrld control µroblem
,-./0
commund geur rutlo, gus
µedul, und brukes to )$;6<'
u deslred sµeed und mlnlmlze
consumµtlons
1!.//23,240
¬ contlnuous ;8& dlscrete lnµuts
¬ dynumlcs deµends on geur
¬ nonllneur torquejsµeed muµs
1$5%(*'678)$79':$7#9*+
A. 8emµorud Model Predlctlve Control µ
llneur systems
nonllneur systems
llneur hybrld systems
=
ley requlrements for hybrld models
¬!>*(6$%:)%?*'enough to cuµture the behuvlor of the system
- contlnuous dynumlcs (µhyslcul luws)
- loglc comµonents (swltches, uutomutu, softwure code)
- lnterconnectlon between loglc und dynumlcs
¬!4%+:9*$enough for solvlng "#"$%&'&! und &%#()*&'& µroblems
@A;<*'*?*$")B%8C';('(%+:9*';(':7((%#9*D'#5)'87)'(%+:9*$EF
÷ .9#*$)'2%8()*%8
6
A. 8emµorud Model Predlctlve Control µ ,
Plecewlse umne systems
Cun uµµroxlmute nonllneur undjor dlscontlnuous dynumlcs urbltrurlly well
(Sontug ¡ç8¡)
stute÷lnµut sµuce
x(k÷¡)
x(k)
C
¡
C
z
C
¸
C
µ
C
¸
C
6
A. 8emµorud Model Predlctlve Control µ 8
Llscrete Eybrld Automuton (LEA)
(1orrlsl, 8emµorud, zccµ)
Lvent
Cenerutor
llnlte Stute
Muchlne
Mode Selector
Swltched
Amne System
¡
z
s
mode
tlme or
event
counter
!"#$%#&"&'
(%'!)*$*
A. 8emµorud Model Predlctlve Control µ ç
Swltched Amne System
Swltched umne system
1he umne dynumlcs deµend on the current mode i(k):
!"#$%#&"&'
(%'!)*$*
x
c
(k +1) = A
i(k)
x
c
(k) +B
i(k)
u
c
(k) +f
i(k)
[δ(k) = 1] ↔ [xc(k) ≥ 0]
A. 8emµorud Model Predlctlve Control µ ¡c
Lvent generutor
Lvent vurlubles ure generuted by llneur threshold condltlons over contlnuous
stutes, contlnuous lnµuts, und tlme:
Lxumµle:
!"#$%#&"&'
(%'!)*$*

i
e
(k) = 1] ←→ [H
i
x
c
(k) +K
i
u
c
(k) ≤ W
i
]
A. 8emµorud Model Predlctlve Control µ ¡¡
llnlte Stute Muchlne
llnlte stute muchlne
1he blnury stute of the fnlte stute muchlne evolves uccordlng to u 8ooleun
stute uµdute functlon:
Lxumµle:
!"#$%#&"&'
(%'!)*$*
A. 8emµorud Model Predlctlve Control µ ¡z
Mode Selector
Mode selector
1he uctlve mode i(k) ls selected by u 8ooleun functlon of the current blnury
stutes, blnury lnµuts, und event vurlubles:
Lxumµle:
c
¡
c ¡
the system hus ¸ modes
!"#$%#&"&'
(%'!)*$*
1he mode selector cun be seen
us the outµut functlon of the
dlscrete dynumlcs
P
j
∪ N
j
⊆ {1, n}
Aδ ≤ b, δ ∈ {0, 1}
n
A. 8emµorud Model Predlctlve Control µ ¡¸
¡. Convert to Conjunctlve Normul lorm (CNl):
z. 1runsform lnto lnequulltles:
c. Clven u 8ooleun stutement
Loglc ! lnequulltles
µolyhedron
(Clover ¡ç,¸,
vllllums ¡ç,,,
Eooker zccc)
Any loglc µroµosltlon cun be trunsluted lnto %8)*C*$'9%8*;$'%8*G5;9%)%*(
A. 8emµorud Model Predlctlve Control µ ¡µ
¡. Convert to Conjunctlve Normul lorm (CNl):
(see e.g. http://www.oursland.net/aima/propositionApplet.html or just google for "CNl ÷ uµµlet' ...)
z. 1runsform lnto lnequulltles:
Lxumµle:
Loglc ! lnequulltles: symbollc uµµrouch
A. 8emµorud Model Predlctlve Control µ ¡¸
1he µolytoµe ls the convex hull of the rows of
the truth tuble T ussocluted wlth formulu
Loglc ! lnequulltles: geometrlc uµµrouch
Convex hull ulgorlthms: cdd, lrs, qhull, chD, Hull, Porto
8ooleun stutement µolyhedron
1>>A2H µuckuge lncluded ln the Eybrld 1oolbox
conv ￿￿

0
0
0 ￿

, ￿

0
1
0 ￿

, ￿

1
0
0 ￿

, ￿

1
1
1 ￿￿

=



δ ∈ R
3
:
−δ1 +δ3 ≤ 0
−δ2 +δ3 ≤ 0
δ1 +δ2 −δ3 ≤ 1



A. 8emµorud Model Predlctlve Control µ ¡6
Lxumµle: loglc "ANL'
ley ldeu:
vhlte µolnts cunnot be
ln the hull of bluck µolnts
Loglc ! lnequulltles: geometrlc uµµrouch
(¡,¡,¡)
(c,¡,c)
(¡,c,c) (c,c,c)
δ
¸
δ
¡
δ
z
>> V=struct('V',[0 0 0;0 1 0;1 0 0;1 1 1]);
>> H=cddmex('hull',V);A=H.A,b=H.B
A. 8emµorud Model Predlctlve Control µ ¡,
Ceometrlc vs. symbollc uµµrouch
¬ 1he µolyhedron obtulned vlu convex hull ls the smullest one
¬ 1he one obtulned vlu CNl muy be lurger. Lxumµle:
convex hull
CNl
sµurlous vertex
ln (c.¸,c.¸,c.¸)
Note: no other exumµle wlth ¸ vurs but
¬ 1o uvold strlct lnequulltles, reµluce by
where ls u smull number (e.g., muchlne µreclslon) ￿

a ￿

xc −b ≤ M(1 −δ)
a ￿

xc −b > mδ
a ￿

xc −b ≥ ￿ + (m−￿)δ
A. 8emµorud Model Predlctlve Control µ ¡8
8lgM technlque (lñ)
¬ Conslder the '+,"#-,.#$%,'+ condltlon
¬ Assume X bounded und let M und m such thut
¬ 1he '+,"#-,.#$%,'+ condltlon ls equlvulent to ￿
> 0
A. 8emµorud Model Predlctlve Control µ ¡ç
8lgM technlque (lfthenelse)
¬ Conslder the '+,()*#,*$&* condltlon
¬
Assume X bounded und let M
¡
, M
z
und m
¡
, m
z
such thut
¬ 1he '+,()*#,*$&* condltlon ls equlvulent to
A. 8emµorud Model Predlctlve Control µ zc
where .
Swltched umne system
Swltched
Amne System
¡
z
s
1he stuteuµdute equutlon cun be rewrltten us u dlñerence
equutlon ÷ '+,()*#,*$&*!condltlons:
Cutµut equutlons y
c
(k)=C
i
x
c
(k)÷D
i
u
c
(k)÷g
i
udmlt u slmllur trunsformutlon
A. 8emµorud Model Predlctlve Control µ z¡
(Clover ¡ç,¸, vllllums ¡ç,,,
Eooker zccc)
Loglc und lnequulltles
llnlte Stute
Muchlne
Mode Selector
Swltched
Amne System
¡
z
s
Lvent
Cenerutor





x
k+1
= Ax
k
+B
1
u
k
+B
2
δ
k
+B
3
z
k
+B
5
y
k
= Cx
k
+D
1
u
k
+D
2
δ
k
+D
3
z
k
+D
5
E
2
δ
k
+E
3
z
k
≤ E
4
x
k
+E
1
u
k
+E
5
A. 8emµorud Model Predlctlve Control µ zz
Mlxed Loglcul Lynumlcul (MLL) systems
(8emµorud, Morurl ¡ççç)
Mlxed Loglcul Lynumlcul (MLL) systems
EYSLLL
(1orrlsl, 8emµorud, zccµ)
Llscrete Eybrld Automuton
¬ Comµututlonully orlented (mlxedlnteger µrogrummlng)
¬ Sultuble for controller synthesls, verlfcutlon, ...
Contlnuous und
blnury vurlubles
x ∈ R
nc
×{0, 1}
n
b, u ∈ R
mc
×{0, 1}
m
b
y ∈ R
pc
×{0, 1}
p
b, δ ∈ {0, 1}
r
b, z ∈ R
rc
A. 8emµorud Model Predlctlve Control µ
Mlxedlnteger models ln Cµerutlons Reseurch

Cost functlon:
sum of µrofessors'
µreferences
Construlnts:
µrofessors [students|
cunnot teuch [tuke|
two courses ut the
sume tlme, etc.
1runslutlon of loglcul relutlons lnto llneur lnequulltles ls heuvlly used ln
oµerutlons reseurch (CR) for solvlng comµlex declslon µroblems by uslng
mlxedlnteger (llneur) µrogrummlng (MlP)
Lxumµle: 1lmetuble generutlon (for demundlng µrofessors .)
A. 8emµorud Model Predlctlve Control µ
Mlxedlnteger models ln Cµerutlons Reseurch

CPU tlme: c.z s
Lñort: ¸° muthemutlcul µroblem setuµ (MlLP model)
¸¸° dutubuse & web lnterfuces
6c° deul wlth µrofessors' comµlulnts, comµlulnts, comµlulnts .
1runslutlon of loglcul relutlons lnto llneur lnequulltles ls heuvlly used ln
oµerutlons reseurch (CR) for solvlng comµlex declslon µroblems by uslng
mlxedlnteger (llneur) µrogrummlng (MlP)
Lxumµle: 1lmetuble generutlon (for demundlng µrofessors .)
A. 8emµorud Model Predlctlve Control µ
Mlxedlnteger models ln Cµerutlons Reseurch

1runslutlon of loglcul relutlons lnto llneur lnequulltles ls heuvlly used ln
oµerutlons reseurch (CR) for solvlng comµlex declslon µroblems by uslng
mlxedlnteger (llneur) µrogrummlng (MlP)
Lxumµle: Cµtlmul multlµerlod lnvestments for mulntenunce und
uµgrude of electrlcul energy dlstrlbutlon networks
(8emµorud, Munoz, Pluzzesl, zcc6)
A. 8emµorud Model Predlctlve Control µ z6
(37)*: thls ls the nonllneur system )
¬ Assoclute
A slmµle exumµle
¬ System:
¬ 1hen
¬ Rewrlte us u llneur equutlon
und trunsform
A. 8emµorud Model Predlctlve Control µ z,
EYSLLL
EYSLLL
(EYbrld Systems LLscrlµtlon Lunguuge)
¬ Lescrlbe +,-)%(.','$*/'0
- Automutu
- Loglc
- Lln. Lynumlcs
- lnterfuces
- Construlnts
(1orrlsl, 8emµorud, zccµ)
¬ Automutlcully generute MLL models ln MA1LA8
http://control.ethz.ch/~hybrid/hysdel
Lownloud:
Reference:
http://cse.lab.imtlucca.it/~bemporad/hybrid/toolbox
A. 8emµorud Model Predlctlve Control µ z8
LEA und EYSLLL models
SYSTEM name {
INTERFACE {
STATE {
REAL xc [xmin,xmax];
BOOL xl; }
INPUT {
REAL uc [umin,umax];
BOOL ul; }
PARAMETER {
REAL param1 = 1;}
} /* end of interface */

IMPLEMENTATION {
AUX { BOOL d;
REAL z; }
AUTOMATA { xl = xl & ~ul; }
DA { z = { IF d THEN 2*xc ELSE -xc }; }
AD { d = xc - 1 <= 0; }
CONTINUOUS {
xc = z; }
MUST {
xc + uc <= 2;
~(xl & ul); }
} /* end implementation */
} /* end system */
Addltlonul relutlons
construlnlng system's
vurlubles
A. 8emµorud Model Predlctlve Control µ zç
AD { delta = hmax - h <= 0; }
Lxumµle ¡: Lefnltlon of event vurlubles
A. 8emµorud Model Predlctlve Control µ ¸c
Lxumµle z: Nonllneur (PvA) functlons
DA { unl = { IF th THEN 2.3*u - 1.3*ut
ELSE u }; }
AD { th = ut - u <= 0; }
Nonllneur umµllfcutlon unlt
A. 8emµorud Model Predlctlve Control µ ¸¡
Lxumµle ¸: Loglcul relutlons
LOGIC {
brake = alarm & ~(tunnel & fire);
}
I59*: bruke lf there ls un ulurm slgnul, but
only lf the truln ls not on fre ln u tunnel
A. 8emµorud Model Predlctlve Control µ ¸z
Lxumµle µ: Contlnuous dynumlcs
Aµµly forwurd dlñerence rule:
CONTINUOUS {
u = u + Ts*iC*i;
}
i(t)
u(t)
Note: iC = ¡jC ls used due to u bug ln EYSLLL, thut cunnot hundle dlvlslon by u sculur µurumeter
A. 8emµorud Model Predlctlve Control µ ¸¸
Lxumµle ¸: Automuton
AUTOMATA {
xclosing = (uclose & xclosing) | (uclose & xstopped);
xstopped = ustop | (uopen & xclosing) | (uclose & xopening);
xopening = (uopen & xstopped) | (uopen & xopening);
}
llowcontrol through u dum
oµen
close | stoµ
stoµ
oµen | stoµ
close
oµen
oµenlng stoµµed closlng
close
blnury lnµuts:
blnury stutes:
A. 8emµorud Model Predlctlve Control µ ¸µ
Lxumµle 6: lmµose u construlnt
MUST {
h - hmax <= 0;
-h <= 0;
}
Arbltrury loglc construlnts ure ulso suµµorted
¨ y = −g
y ≤ 0 ⇒ ˙ y(t
+
) = −(1 −α) ˙ y(t

)
α ∈ [0, 1]
A. 8emµorud Model Predlctlve Control µ ¸¸
8ounclng bull exumµle
F=-m g
y
N
Eow to model the bounclng bull us u dlscrete
tlme hybrld system ?
y(t) > 0
v(t) ≈
y(t) −y(t −1)
Ts
−g = ¨ y(t) ≈
v(t) −v(t −1)
Ts





v(t +1) = v(t) −Tsg
y(t +1) = y(t) +Tsv(t +1)
= y(t) +Tsv(t) −T
2
s
g ￿

v(t +1) = −(1 −α)v(t)
y(t +1) = y(t) −Tsv(t)
v(t) = −(1 −α)v(t −1)
y(t) ≤ 0
y(t +1) = y(t −1)
= y(t) −Tsv(t)
A. 8emµorud Model Predlctlve Control µ ¸6
8ounclng bull - 1lme dlscretlzutlon
mg
y
y
go to demo jdemos/hybrid/bball.m
A. 8emµorud Model Predlctlve Control µ ¸,
8ounclng bull EYSLLL model
SYSTEM bouncing_ball {
INTERFACE {
/* Description of variables and constants */
STATE { REAL height [-10,10];
REAL velocity [-100,100]; }
PARAMETER {
REAL g;
REAL alpha; /* 0=elastic, 1=completely anelastic */
REAL Ts; }
}
IMPLEMENTATION {
AUX { BOOL negative;
REAL hnext;
REAL vnext; }
AD { negative = height <= 0; }
DA { hnext = { IF negative THEN height-Ts*velocity
ELSE height+Ts*velocity-Ts*Ts*g};
vnext = { IF negative THEN -(1-alpha)*velocity
ELSE velocity-Ts*g}; }
CONTINUOUS {
height = hnext;
velocity = vnext;}
}}
go to demo jdemos/hybrid/bball.m
A. 8emµorud Model Predlctlve Control µ ¸8
8ounclng bull Slmulutlon
>>Ts=0.05;
>>g=9.8;
>>alpha=0.3;
>>S=mld('bouncing_ball',Ts);
>>N=150;
>>U=zeros(N,0);
>>x0=[5 0]';
>>[X,T,D]=sim(S,x0,U);
Note: no 2eno eñect ln dlscrete tlme !
A. 8emµorud Model Predlctlve Control µ ¸ç
Reullzutlon und 1runsformutlons
(StuteSµuce Eybrld Models)
A. 8emµorud Model Predlctlve Control µ µc
Lqulvulences of hybrld models
Σ
¡
Σ
z
x
1
(0)=x
2
(0)
u(k)
x
!
(k), y
!
(k)
x
"
(k), y
"
(k)
A. 8emµorud Model Predlctlve Control µ µ¡
MLL und PvA Systems
¬ Proof ls constructlve: glven un MLL system lt returns lts equlvulent PvA form
¬ Lruwbuck: lt needs the enumerutlon of ull µosslble comblnutlons of blnury stutes,
blnury lnµuts, und δ vurlubles
¬ Most of such comblnutlons leud to emµty reglons
¬ Lmclent ulgorlthms ure uvulluble for convertlng MLL models lnto PvA models
uvoldlng such un enumerutlon:
¬ A. 8emµorud, "12!%*#$.345")%$+/'.6").7"#8*)$%#5.9%:*(.;"5%!<4.=,#</%!<4.>,'$*/'.%#$".<#.
....1?&%8<4*#$.@%*!*A%'*.32#*.B")/', lLLL 1runs. Autom. Contr., zccµ.
¬ 1. Ceyer, l.L. 1orrlsl und M. Morurl, "12!%*#$.9"(*.1#&/*)<$%"#."6.7"/C"'%$%"#<4.D,-)%(...
....9"(*4'', ESCC'c¸
(8emµorud, lerrurl1recute, Morurl, !""" $%&'())))
JB*7$*+ MLL systems und PvA systems ure equlvulent
A. 8emµorud Model Predlctlve Control µ µz
8ounclng bull PvA equlvulent
>>P=pwa(S);
>>plot(P)
>>[X,T,I]=sim(P,x0,U);
A. 8emµorud Model Predlctlve Control µ µ¸
Lxumµle: Room temµeruture
T
amb
• #¡ turns the heuter (AjC) on
whenever he ls cold (hot)
• lf #z ls cold he turns the heuter on,
unless #¡ ls hot
• lf #z ls hot he turns AjC on,
unless #z ls cold
• Ctherwlse, heuter und AjC ure oñ
heuter
AC system
u
hot u
cold
go to demo jdemos/hybrid/heatcool.m

!"#$%&'&"8;+%6(
T
1
T
2
A. 8emµorud Model Predlctlve Control µ µµ
EYSLLL model
>>S=mld('heatcoolmodel',Ts)
>>[XX,TT]=sim(S,x0,U);
get the MLL model ln MA1LA8
slmulute the MLL model
A. 8emµorud Model Predlctlve Control µ µ¸
Eybrld MLL model
¬ MLL model
¬ z contlnuous stutes:
¬ ¡ contlnuous lnµut:
¬ 6 uuxlllury blnury vurs:
¬ z uuxlllury contlnuous vurs:
(room temµeruture T
umb
)
(temµerutures T
¡
,T
z
)
(µower fows u
hot
, u
cold
)
(µ thresholds ÷ z for CR condltlon)
¬ zc mlxedlnteger lnequulltles
Posslble comblnutlon of lnteger vurlubles: z
6
= 6µ
A. 8emµorud Model Predlctlve Control µ µ6
Eybrld PvA model
¬ PvA model
¬ z contlnuous stutes:
¬ ¡ contlnuous lnµut:
(room temµeruture T
umb
)
(temµerutures T
¡
,T
z
)
¬ ¸ µolyhedrul reglons
EC<)$%$%"#.("*'.#"$.(*C*#(."#.%#C&$F
>>P=pwa(S);
both oñ heuter on AjC on
A. 8emµorud Model Predlctlve Control µ µ,
Slmulutlon ln Slmullnk
MLL und PvA models ure equlvulent
A. 8emµorud Model Predlctlve Control µ µ8
Uslng MLL to PvA for Model Checklng
¬ Assume µlunt ÷ controller cun be modeled us LEA:
K9;8) = PvA uµµroxlmutlon (e.g.: nonllneur swltched model)
178)$799*$ = swltched llneur controller (e.g: u comblnutlon of threshold
condltlons, loglc, llneur feedbuck luws, ...)
¬ vrlte EYSLLL model, convert to MLL, then to PvA
¬ 1he resultlng PvA muµ tells how the closedlooµ system behuves ln
dlñerent reglons of the stutesµuce
A. 8emµorud Model Predlctlve Control µ
ldentlfcutlon of Eybrld Systems
A. 8emµorud Model Predlctlve Control µ ¸c
Eybrld system ldentlfcutlon
¬ Sometlmes u )%/0'-!1.-*$ of the µrocess (or of u µurt of lt) cunnot be derlved
munuully from uvulluble knowledge.
¬ 1herefore, u model must be elther
- Lstlmuted from dutu (model unknown)
- or +,-)%(%G*(.before lt cun be used for controljunulysls (model known but nonllneur)
¬ lf u llneur model ls enough, no µroblem: severul ulgorlthms ure uvulluble (e.g.:
use Ljung's lL 18×)
¬ lf swltchlng modes ure known und dutu cun be generuted for euch mode, no
µroblem: we ldentlfy one llneur model µer mode (e.g.: use Ljung's lL 18×)
¬ lf modes & dynumlcs must be ldentlfed together, we need
B"#$%&'("()*+'%&*8)%L6;)%78
A. 8emµorud Model Predlctlve Control µ ¸¡
PvA ldentlfcutlon µroblem
Lstlmute from dutu #7)B the µurumeters of the umne submodels ;8& the
µurtltlon of the PvA muµ
2M;+:9* Let the dutu be generuted by the PvAR× system
A. 8emµorud Model Predlctlve Control µ ¸z
PvA ldentlfcutlon µroblem
.E'N87O8'C5;$&9%8*( (µurtltlon H
j
known, θ
j
unknown): ordlnury leustsquures
µroblem (or llneurjquudrutlc µrogrum lf llneur bounds over θ
j
ure glven)
2.4P'KI-Q/2A'EEE
QE'R8<87O8'C5;$&9%8*( (µurtltlon H
j
und θ
j
unknown):
generully nonconvex, locul mlnlmu !.I>'KI-Q/2AS
A. 8emµorud Model Predlctlve Control µ ¸¸
Aµµrouches to PvA ldentlfcutlon
Polynomlul fuctorlzutlon (ulgebrulc uµµrouch)
Eyµerµlune clusterlng ln dutu sµuce
¬ Mlxedlnteger llneur or quudrutlc µrogrummlng
¬ 8ounded error & µurtltlon of lnfeuslble set of lnequulltles
¬ lmeuns clusterlng ln u feuture sµuce
¬ 8uyeslun uµµrouch
A. 8emµorud, A. Curulll, S. Puolettl und A. vlclno, "3.H)**(,.3CC)"<!+ $" I(*#$%J!<$%"# "6.@%*!*A%'*.32#*.9"(*4'',
ESCC'c¸ j lLLL 1AC zcc¸
C. lerrurl1recute, M. Muselll, L. Llberutl, und M. Morurl, "3.!4&'$*)%#5
$*!+#%?&*.6") $+*.%(*#$%J!<$%"# "6.C%*!*A%'*.<2#*.','$*/'', Automutlcu,
zcc¸
'. Roll, A. 8emµorud und L. Ljung, "I(*#$%J!<$%"# "6 +,-)%(.','$*/'.8%<
/%:*(K%#$*5*).C)"5)<//%#5', Automutlcu, zccµ
(R. vldul, S. Soutto, S. Sustry, zcc¸)
(L. Munz, v. lrebs, IB37 zccz)
¬ Cther uµµrouches:
'uloskl A, vlelund S, Eeemels vPME, "A 8uyeslun uµµrouch to ldentlfcutlon
of hybrld systems. CLC zccµ

Sign up to vote on this title
UsefulNot useful