alo f buf cse Research in Knowledge Representation
alo f buf @ cse Research in Knowledge Representation and Reasoning Stuart C. Shapiro Department of Computer Science & Engineering Center for Multi. Source Information Fusion Center for Cognitive Science University at Buffalo, The State University of New York Fall 2006 S. C. Shapiro
alo @ cse f buf MGLAIR Agent Architecture Mind Body KL (SNe. PS) Independent of lower-body implementation PMLa PMLb I/P s o c k e t s Dependent on lower-body implementation PMLc Proprioception Speech Hearing SAL Vision Motion Fall 2006 S. C. Shapiro 2 W O R L D
alo f buf @ cse SNe. PS is a Logic-Based Frame-Based Network-Based knowledge representation, reasoning, and acting system. 3 Fall 2006 S. C. Shapiro
alo @ cse f buf SNe. PS Is Logic-Based wff 1!: all(x)(Isa(x, dog) => Property(x, four-legged)) wff 2!: Isa(Toto, dog) : Property(Toto, ? x)? wff 3!: Property(Toto, four-legged) : Isa(Fala, dog)! wff 6!: Property(Fala, four-legged) wff 5!: Isa(Fala, dog) : askwh Property(? x, four-legged) Fala: Fala Toto: Toto : list-wffs wff 6!: Property(Fala, four-legged) wff 5!: Isa(Fala, dog) wff 3!: Property(Toto, four-legged) wff 2!: Isa(Toto, dog) wff 1!: all(x)(Isa(x, dog) => Property(x, four-legged)) 4 Fall 2006 S. C. Shapiro
alo @ cse f buf SNe. PS Is Frame-Based : %(describe *nodes) (m 6! (a 1 Fala) (a 2 four-legged) (r Property)) (m 5! (a 1 Fala) (a 2 dog) (r Isa)) (m 3! (a 1 Toto) (a 2 four-legged) (r Property)) (m 2! (a 1 Toto) (a 2 dog) (r Isa)) (m 1! (forall v 1) (ant (p 1 (a 1 v 1) (a 2 dog) (r Isa))) (cq (p 2 (a 1 v 1) (a 2 four-legged) (r Property)))) : %(describe (assert r Isa a 1 (Fido Rover Lassie) a 2 (dog pet))) (m 7! (a 1 Lassie Rover Fido) (a 2 pet dog) (r Isa)) wff 7!: Isa({Lassie, Rover, Fido}, {pet, dog}) : Property(Rover, ? x)? wff 8!: Property(Rover, four-legged) 5 Fall 2006 S. C. Shapiro
alo f buf @ cse SNe. PS Is Network-Based : define-frame Ako(nil subclass superclass) Ako(x 1, x 2) will be represented by {<subclass, x 1>, <superclass, x 2>} : Ako({man, dog}, mammal). wff 1!: Ako({dog, man}, mammal) : Ako({mammal, fish}, vertebrate). wff 2!: Ako({fish, mammal}, vertebrate) : Ako(vertebrate, animal). wff 3!: Ako(vertebrate, animal) : %(define-path subclass (compose subclass (kstar (compose superclass- ! subclass)))). . . : askwh Ako(? x, animal) vertebrate: vertebrate fish: fish mammal: mammal dog: dog man: man 6 Fall 2006 S. C. Shapiro
alo f buf @ cse Procedural Attachment A predicate or function symbol may be attached to a user-written procedure so instances may be computed in the underlying programming language. 7 Fall 2006 S. C. Shapiro
alo @ cse f buf Example of Procedural Attachment : Diff(7, 3, ? x)? wff 24!: Diff(7, 3, 4) : Diff(10, ? x, 7)? wff 25!: Diff(10, 3, 7) : Diff(? x, 5, 7)? wff 26!: Diff(12, 5, 7) : Diff(15, 8, 7)? wff 314!: Diff(15, 8, 7) : Diff(15, 8, 9)? wff 316!: ~Diff(15, 8, 9) 8 Fall 2006 S. C. Shapiro
alo f buf @ cse Building Domain all(x)(on. Floor(x) => {(<(x, 3) => location(below. Ground)), (<(2, x) => location(above. Ground))}). 9 Fall 2006 S. C. Shapiro
alo @ cse f buf Primitive Acts A version of procedural attachment for implementing intelligent agents: : perform say(Welcome, "to you all. ") Welcome to you all. 10 Fall 2006 S. C. Shapiro
alo @ cse f buf Policies Connect propositions and acts: wheneverdo(location(below. Ground), withsome(f, on. Floor(f), say("It's dark here on floor", f), say("Where am I? ", ""))). wheneverdo(location(above. Ground), withsome(f, on. Floor(f), say("It's sunny outside floor", f), say("Where am I? ", ""))). 11 Fall 2006 S. C. Shapiro
alo SNe. BR: Belief Revision/ Assumption-Based Truth Maintenance f buf @ cse • Identify possible culprits of contradictions. • Disbelieve implications of disbelieved hypotheses. • Use state constraints to adjust beliefs: andor(1, 1){on. Floor(1), on. Floor(2), on. Floor(3), on. Floor(4)}. andor(1, 1){location(below. Ground), location(above. Ground)}. 12 Fall 2006 S. C. Shapiro
alo @ cse f buf Combined Use of SNe. BR & Procedural Attachment : perform believe(on. Floor(1)) It's dark here on floor 1 : location(? x)? wff 24!: ~location(above. Ground) wff 6!: location(below. Ground) : perform believe(on. Floor(4)) It's sunny outside floor 4 : location(? x)? wff 33!: ~location(below. Ground) wff 7!: location(above. Ground) 13 Fall 2006 S. C. Shapiro
alo f buf @ cse BR with Multiple Sources wff 1: all(x)(andor(0, 1){mammal(x), fish(x)}) wff 2: all(x)(fish(x) <=> has(x, scales)) wff 4: all(x)(whale(x) => fish(x)) wff 5: Source(Melville, all(x)(whale(x) => fish(x))) wff 6: all(x)(whale(x) => mammal(x)) wff 7: Source(Darwin, all(x)(whale(x) => mammal(x))) wff 8: Sgreater(Darwin, Melville) wff 11: free(Willy) and whale(Willy) Note: Source & Sgreater props are regular object-language props. 14 Fall 2006 S. C. Shapiro
alo f buf @ cse Finding the Contradiction : has(Willy, scales)? I infer fish(Willy) I infer has(Willy, scales) I infer mammal(Willy) I infer it is not the case that wff 14: fish(Willy) 15 Fall 2006 S. C. Shapiro
alo f buf @ cse Using Source Credibility A contradiction was detected within context default-defaultct. The contradiction involves the newly derived proposition: wff 17: ~fish(Willy) {<der, {wff 1, wff 6, wff 11}>} and the previously existing proposition: wff 14: fish(Willy) {<der, {wff 4, wff 11}>} The least believed hypothesis: (wff 4) The most common hypothesis: (nil) The hypothesis supporting the fewest wffs: (wff 1) I removed the following belief: wff 4: all(x)(whale(x) => fish(x)) I no longer believe the following 2 propositions: wff 14: fish(Willy) wff 13: has(Willy, scales) 16 Fall 2006 S. C. Shapiro
alo f buf @ cse Conclusions • MGLAIR is an agent architecture – For connecting reasoning with sensing and acting • SNe. PS is a – – Logic-based Frame-based Network-based Knowledge representation, reasoning, and acting system. • Procedural attachment provides – Sensing, acting, computing at the “subcognitive” layers • SNe. BR does belief revision & truth maintenance. • Source meta-knowledge may be entered and used. 17 Fall 2006 S. C. Shapiro
- Slides: 17