Using constraint programming framework for Semiqualitative Reasoning
Abstract
A semiqualitative model is the representation of the world by means of qualitative and quantitative knowledge. It is motivated by the lack of quantitative knowledge in determined parameters or variables of the model. We wish to obtain conclusions and generalization derived from these specific models .. In this work, we show as semiqualitative models may be transformed to interval constraint satisfaction problems(ICSP). The inference process in an ICSP is based in combining constraint propagation and bronch and bound algorithms. The interval domains of the variables are to propagate through the constraints. A defined operator must exist for every qualitative knowledge. lt transforms this knowledge to an interval constraint. The application of the constraint programming framework to a ICSP allows the obtaining of conclusions in a automatic way.
Full text
[Pie a95]. N. Pie a. Cu en T ends in Quali a i e
Reasoning and Applica ions, Monog a la
CIMNE, 33. In e na ional Cen e o
Nume ical Me hods in Enginee ing,
Ba celona.
[Pie a87]. N. Pie a. Connec ius de logiques no
es anda d com a ope ado s d'ag egació en
classi icació mul i a iable i econeixemen de
o mes. Ph. D. Thesis Uni e si a Poli ecnica
de Ca alunya, Ba celona.
[Raiman86]. O. Raiman. O de o Magni ude
Reasoning. P oceedings o AAAI Con e ence,
Philadelphia, 100-104.
(Schwei7.e &Skla 83]. B. Schweize and A. Skla .
P obalis ic Me ic Spaces, No h-Holland,
Ams e dam.
[T a é-Massuyes&Pie a89]. L. T a é-Massuyes
and N. Pie a. Toe O de s o Magni ude
Models as Quali a i e Algeb as. P oceedings
11 h IJCAl, De oi , 1261-1266.
[T a é-Massuyes e al.97]. L. T a é-Massuyes,
Ph. Dague and F. GueITin. Le Raisonnemen
Quali a i pou les Sciences de l'lngenieu , Ed.
He mes, Pa is.
74
Using cons ain p og amming amewo k o
Semiquali a i e Reasoning
R.M. Gasea, J.A. O ega, M. To o, J. To es and J.M. Co de o
Depa amen o de Lenguajes y Sis emas In o má icos.
Uni e sidad de Se illa
A da. Reina Me cedes s/n. Se illa (Spain)
{gasea ,jao ega, m o o,j o es,co de o }©lsi. us.es
Abs ac
A semiquali a i e model is he ep esen a ion o he wo ld by means o quali a i e and quan i a i e
knowledge. I is mo i a ed by he lack o quan i a i e knowledge in de e mined pa ame e s o a iables
o he model. We wish o ob ain conclusions and gene aliza ion de i ed om hese speci ic models ..
In his wo k, we show as semiquali a i e models may be ans o med o in e al cons ain sa is ac ion
p oblems(ICSP). The in e ence p ocess in an ICSP is based in combining cons ain p opaga ion and
b onch and bound algo i hms. The in e al domains o he a iables a e o p opaga e h ough he
cons ain s. A de ined ope a o mus exis o e e y quali a i e knowledge. l ans o ms his knowledge
o an in e al cons ain . The applica ion o he cons ain p og amming amewo k o a ICSP allows
he ob aining o conclusions in a au oma ic way.
Keywo ds: Quali a i e easoning, Cons ain P og amming, Consis ency echniques.
1 In oduc ion
Di e en communi ies ha e s udied he qua
li a i e ep esen a ion o he wo ld and hey
ha e p oposed o malisms o quali a i ely ea
son abou hese ep esen a ions. The main m
i a ions a e he lack o quan i a i e knowledge
and he excess o hem ha may ob ain conclu
sions and explana ions unde s andable by hu
mans. Gene ally he eal models a e no pu e
quali a i e o pu e quan i a i e bu a e no
mally composed o quan i a i e and quali a
i e knowledge. l is known as semiquali a i e
model. Ali his knowledge mus be conside ed
when hese models a e s udied.
In he eigh ies, he main concep s o he quali
a i e echniques appea ed in he a ea o A i
icial In elligence(AJ), by means o he publica-
ion o he sys ems ENVISION [DeKlee 84],
QSJM[Kuipe s86] and QPT[Fo bus84]. Di
en monog aphs ha e been published in he
90s [Weld90),[Fal ings92],[Kuipe s94],(Pie a95],
[Dague95],[T a e97], ha e lec he opics and
he cli e en de eloped echniques in quali a
i e easoning(QR). This quali a i e know
ledge is use ul o unde s and, in a simple way,
some o he p ope ies o he models, he
o e he simplici y is i s p incipal p ope y. l
ha e been applied in di e en asks, such as su
pe ision and diagnosis o chemical indus ial
p ocess[Penal a91l[Bousson93](Chang93J, ma
ine mo o (Mo eno93), he con ol [Lepe i 87],
analysis o dynamical sys ems (A acil93]
[Gasca98], so wa e analysis [Xan hakis94] and
concep ual design o s uc u es in enginee ing
[Bozzo98]. In a majo i y o he app oaches o
modeling quali a i e easoning in Al, i is as
sumed explici ly o implici ly ha quali a i e
75
1