Presentation is loading. Please wait.

Presentation is loading. Please wait.

FLUX. FLUX Predicates holds(X, Z) –Asserts X in Z knows(X, Z) –X is known to be true in Z knows_val(X, V, Z) – update(Z1, FAdd, FRemove, Z2) –Z2 = Z1.

Similar presentations


Presentation on theme: "FLUX. FLUX Predicates holds(X, Z) –Asserts X in Z knows(X, Z) –X is known to be true in Z knows_val(X, V, Z) – update(Z1, FAdd, FRemove, Z2) –Z2 = Z1."— Presentation transcript:

1 FLUX

2 FLUX Predicates holds(X, Z) –Asserts X in Z knows(X, Z) –X is known to be true in Z knows_val(X, V, Z) – update(Z1, FAdd, FRemove, Z2) –Z2 = Z1 + FAdd – FRemove execute(A, Z1, Z2) –Z2 is the state after executing A in Z1

3 FLUX as a Modeling Language perform(+Action, -Perceptions). –Executes the action on the environment simulator –Returns the perception list state_update(Z1, Action, Z2, Perceptions) –Updates agent’s internal world model from Z1 to Z2, provided Action has been executed and Perceptions have been detected

4 Non FLUX modeling Action preconditions (the poss/2 or 3 predicate) –Some examples do not include it –The underlying code does not uses it –Each example models it in a different way... poss(Action, State) poss(Action, Perceptions, State) Model initialization Planing Environment simulator

5 CHR Operational Semantics in FLUX

6 Agent View World Model –Execution State –Program Rules Actions –Transitions Execution –While a final state has not been reached Apply transitions non-deterministically

7 World Model rule0 @ a, b \ c, d g0 | b0. rule1 @ x, y ==> g1 | b1. init(Z) :- holds(goal(g0), Z),..., holds(goal(gn), Z), holds(udstore(u0), Z),..., holds(udstore(un), Z), holds(bicstore(b0), Z),..., holds(bicstore(bn), Z), holds(rule(a, b \ c, d g0 | b0), Z), holds(rule(x, y ==> g1 | b1), Z).

8 Helper Predicates isBIC(true). isBIC(false). isBIC(_ == _). isUDC(X) :- not isBIC(X).

9 Helper Predicates getVars(Constraints, Variables) –getVars([p(X), q(Y,X)], W), W = [X,Y]. renameVars(HeadKeep0, HeadRemove0, Guard0, Body0, HeadKeep, HeadRemove, Guard, Body) findMatchingConstraints(Z, Constraints, Matching)

10 Helper Predicates generateMatchingEquations(Constraints, Matching, Equations) –generateMatchingEquations([c(X),d(Y)], [c(1), d(2)], W = [X == 1, Y == 2].

11 Helper Predicates entails(Z, V, C, Z1) –Z = –Z1 = –Z  Z1 –

12 Helper Predicates processBics(Z, _, [], Z). processBics(Z, LV, [B|R], Zf) :- entails(Z,LV,B, Z1), update(Z1,[bicstore(B)], [], Z2), processBics(Z2, LV, R,Zf).

13 Helper Predicates removeMatching(Z1, [], Z1). removeMatching(Z1, [C|R], Z2) :- removeMatching(Z1, R, Z12), update(Z12, [], [udstore(C)], Z2). addBody(Z1, [], Z1). addBody(Z1, [C|R], Z2) :- addBody(Z1, R, Z12), update(Z12, [goal(C)], [], Z2). addEquations(Z1, [], Z1). addEquations(Z1, [C|R], Z2) :- addEquations(Z1, R, Z12), update(Z12, [bicstore(C)], [], Z2).

14 state_update(Z1, introduce, Z2, []) :- (knows_val([X], goal(X), Z1), isUDC(X), update(Z1, [udstore(X)], [goal(X)], Z2)) ; Z1 = Z2.

15 state_update(Z1, solve, Z2, []) :- (knows_val([X], goal(X), Z1), isBIC(X), getVars([X], LV), entails(Z1, LV, X, Z2)) ; Z1 = Z2.

16 state_update(Z1, simplify, Zf, []) :- (knows_val([HeadRemove0, Guard0, Body0], rule(HeadRemove0 Guard0 : Body0), Z1), renameVars([], HeadRemove0, Guard0, Body0, _, HeadRemove, Guard, Body),...

17 findMatchingConstraints(Z1, HeadRemove,Matching), generateMatchingEquations(HeadRemove, Matching, Equations), getVars(HeadRemove, HeadVars), getVars(Guard, GuardVars), subtract(GuardVars, HeadVars, GuardLocalVars), processBics(Z1, HeadVars, Equations, Z2), processBics(Z2, GuardLocalVars, Guard, Z3),...

18 ... ((knows(bicstore(false), Z3), fail) ; (not knows(bicstore(false), Z3), removeMatching(Z3, Matching, Z4), addBody(Z4, Body, Z5), addEquations(Z5, Equations, Zf) ) ) ; Z1 = Zf.

19 CHOR Semantics in FLUX

20 CHR + O F-Atoms are syntax sugar for special UD constraints using Kifer’s codification –O:C  isa(O,C) –C::S  sub(C,S) –X[F->V]  fd(X,F,V) A new transition was added. It completes the current UD-store with information inherited by the means of inheritance Syntatic constraints: –O:C and C::S cannot appear on HeadRemove

21 Helper Predicates isFAtom(isa(_,_)). isFAtom(sub(_,_)). isFAtom(fd(_,_,_)).

22 Inherit Taxonomy Completion Taxonomy Validation Feature Inheritance validadeCycles/2 propagateSub/2 propagateIsa/2 computeInheritedFeatures/2 computeOverriding/3 cancelOverridenByCS/3

23 state_update(Z1, inherit, Zf, []) :- (not (knows_val([X], goal(X), Z1), isFAtom(X)), handleTypeInheritance(Z1, Z2), handleFeatureInheritance(Z2,Zf)) ; Z1 = Zf. handleTypeInheritance(Z1, Zf) :- propagateSub(Z1,Z2), propagateIsa(Z2,Z3), validateCycles(Z3,Zf).

24 handleFeatureInheritance(Z,Zf) :- computeInheritedFeatures(Z,Candidates), computeOverriding(Z,Candidates, CandidatesWithOutOverriden), cancelOverridenByCS(Z, CandidatesWithOutOverriden, CandidatesWithOutOverridenExt), computeSourceValue(Z,CandidatesWithOutOverridenExt, CandidatesFinal, Conflicts), transformCandidates(CandidatesFinal, NewFeatures), update(Z, NewFeatures, [], Z1), getVars(Conflicts, VarC), processBics(Z1, VarC, Conflicts, Zf).

25 Examples

26 Clyde rule1 @@ main clyde:royalElephant, royalElephant::elephant, elephant[color->gray], royalElephant[color->white) ].

27 Clyde (Results) clyde[color->white] clyde:elephant royalElephant[color->white] elephant[color->gray] royalElephant::elephant clyde:royalElephant

28 Nixon diamond rule1 @@ main quaker[policy->pacifist], republican[policy->hawk], nixon:quaker, nixon:republican.

29 Nixon Diamond (Results) Source Based Multiple Inheritance nixon : republican nixon : quaker republican[policy -> hawk] quaker[policy -> pacifist]

30 Nixon Diamond (Results) Value Based Multiple Inheritance nixon[policy->pacifist] nixon[policy->hawk] nixon:republican nixon:quaker Republican[policy->hawk] quaker[policy->pacifist] hawk==pacifist pacifist==hawk false

31 Example 2.5.1 (Adapt.) rule1 @@ main c1::c2, c2[m->a]. rule2 @@ c1[m->a] \ c2[m->a] c2[m->b].

32 Example 2.5.1 (Adapt.) (Results) c1[m->a] c2[m->b] c1::c2

33 Example 2.5.3 rule1 @@ main c1:c2, c3::c2, c2[m->a], c3[m->b]. rule2 @@ c1[m->a] ===> c1:c3.

34 Example 2.5.3 (Results) c1 : c3 c1[m-> a] c3[m->b] c2[m->a] c3 :: c2 c1 : c2


Download ppt "FLUX. FLUX Predicates holds(X, Z) –Asserts X in Z knows(X, Z) –X is known to be true in Z knows_val(X, V, Z) – update(Z1, FAdd, FRemove, Z2) –Z2 = Z1."

Similar presentations


Ads by Google