Logik für Informatiker: PROLOG Part 3: SLD Trees

Simply Logical – Chapter 3 © Peter Flach, 2000

1

Andreas Karwath

&

Wolfram Burgard

(original slides by Peter Flach)

Logik für Informatiker: PROLOG Part 3: SLD Trees


2

SLD Resolution and SLD Trees


3

SLD resolution

Resolution in Prolog based on definite clause logic

definite clauses: one atom in the head

Resolution strategy: which literal and which input clause is chosen for resolution?

Strategy in Prolog is called SLD resolution:

S selection rule (in Prolog from left to right) L linear resolution D definite clauses

Leftmost literal in query is selected; clauses are selected in textual order

SLD trees: different from proof trees: only show resolvents, but all possible resolution steps

Prolog searches SLD tree depthfirst !


4

:-teaches(peter,ai_techniques)

:-teaches(peter,expert_systems)

:-teaches(peter,computer_science)

?-student_of(S,peter)

SLDtree

:-follows(S,C),teaches(peter,C):-follows(S,C),teaches(peter,C)

:-teaches(peter,expert_systems)

:-teaches(peter,computer_science) :-teaches(peter,ai_techniques)

?-student_of(S,peter)

student_of(X,T):-follows(X,C),teaches(T,C).follows(paul,computer_science).follows(paul,expert_systems).follows(maria,ai_techniques).teaches(adrian,expert_systems).teaches(peter,ai_techniques).teaches(peter,computer_science).

p.445

[][]


5

Infinite SLDtrees

brother_of(X,Y):-brother_of(Y,X).brother_of(paul,peter).

brother_of(paul,peter).brother_of(peter,adrian).brother_of(X,Y):- brother_of(X,Z),

brother_of(Z,Y).

?-brother_of(paul,B)

[] :-brother_of(paul,Z),brother_of(Z,B)

:-brother_of(paul,Z1),brother_of(Z1,Z),brother_of(Z,B):-brother_of(peter,B)

[] :-brother_of(peter,Z),brother_of(Z,B)

•••

•••

•••

:-brother_of(B,peter)

[]

?-brother_of(peter,B)

:-brother_of(peter,B)

:-brother_of(B,peter)

[]

p.456


6

Depthfirst search

?-plist(L)

plist([]).plist([H|T]):-

p(H),plist(T).

p(1). p(2).[]

L = []

p.47

:-p(H1),plist(T1)

:-plist(T1)

[]

L = [1]

:-p(H1),plist(T1)

:-plist(T1)

[]

L = [1,1]

•••

:-plist(T1)

:-plist(T1)

[]

L = [1,2]

•••

[]

L = [2]

:-p(H1),plist(T1)

:-plist(T1):-plist(T1)

[]

L = [2,1]

[]

L = [2,2]

•••

•••

?-plist(L).L=[];L=[1];L=[1,1];…


7

Summary

First problem: trapped in infinite subtrees

Prolog is incomplete by the way the SLD tree is searched (depth-first; a deliberate design decision for memory efficiency)

if breadth-first search were used, then it would be refutation complete!

Second problem: looping if no (further) solution for any infinite

SLD tree

due to semi-decidability of full clausal logic(infinity of Herbrand base)


8

Programming Example:Monkey and banana


9

Problem: Monkey and Banana

The Problem

There is a monkey at the door into a room. In the middle of the room a banana is hanging from the ceiling. The monkey is hungry and wants to get the banana, but he cannot stretch high enough from the floor. At the window of the room there is a box that the monkey can use. The monkey can perform the following actions: walk on the floor, climb the box, push the box around (if he is already at it), and grasp the banana if he is standing on the box and directly underneath the banana.

Can the monkey grasp the banana?


10

banana

Problem: Monkey and Banana

door

window

box

mokey

middle


11

Monkey and Banana: States

state( MR, MB , P , B )

Monkey‘s Room Position

•at_door•middle•at_window

Monkey‘s Box Position

•on_floor•on_box

Box Position

•at_door•middle•at_window

Banana

•has•has_not


12

Monkey and Banana: Actions I

Possible actions:

walk, push, climb, grasp

Defined using:

move(OldState, Action, NewState)

OldState NewState

Action


13

Monkey and Banana: Actions II

Grasp Banana:

move( state(middle, on_box, middle, has_not),grasp,state(middle, on_box, middle, has)

).

move( state(P, on_floor, P, H),climb,state(P, on_box, P, H) ).

Climb Box:


14

Monkey and Banana: Actions III

Push Box:

move( state(P1, on_floor, P1, H),push(P1, P2),state(P2, on_floor, P2, H)).

move( state(P1, on_floor, P, H),walk(P1, P2),state(P2, on_floor, P, H)).

Walk:


15

Monkey and Banana: Actions Summary

move(state(middle, on_box, middle, has_not),grasp,state(middle, on_box, middle, has)).

move(state(P, on_floor, P, H),climb,state(P, on_box, P, H)).

move(state(P1, on_floor, P1, H),push(P1, P2),state(P2, on_floor, P2, H)).

move(state(P1, on_floor, P, H),walk(P1, P2),state(P2, on_floor, P, H)).


16

Monkey and Banana: Goals

How can the monkey get the banana?

canget(state(_P1, _F, _P2, has)).canget(OldState) :-

move(OldState, Move, NewState),canget(NewState).

•Can the monkey get the banana from the initial state?

?- canget( state(at_door, on_floor,at_window, has_not)).


17

Monkey and Banana: Summary

move(state(middle, on_box, middle, has_not),grasp,state(middle, on_box, middle, has)).

move(state(P, on_floor, P, H),climb,state(P, on_box, P, H)).

move(state(P1, on_floor, P1, H),push(P1, P2),state(P2, on_floor, P2, H)).

move(state(P1, on_floor, P, H),walk(P1, P2),state(P2, on_floor, P, H)).

canget(state(_, _, _, has)).canget(OldState) :-

move(OldState, Move, NewState),canget(NewState).


18

Monkey and Banana: Trace

| ?- canget(state(at_door,on_floor,at_window,has_not)).Call: canget(state(at_door,on_floor,at_window,has_not)) ?Call: move(state(at_door,on_floor,at_window,has_not),_989,_990) ?Exit: move(state(at_door,on_floor,at_window,has_not),walk(at_door,_1476),state(_1476,on_floor,at_window,has_not)) ?Call: canget(state(_1476,on_floor,at_window,has_not)) ?Call: move(state(_1476,on_floor,at_window,has_not),_2415,_2416) ?Exit: move(state(at_window,on_floor,at_window,has_not),climb,state(at_window,on_box,at_window,has_not)) ?Call: canget(state(at_window,on_box,at_window,has_not)) ?Call: move(state(at_window,on_box,at_window,has_not),_3841,_3842) ?Fail: move(state(at_window,on_box,at_window,has_not),_3841,_3842) ?Fail: canget(state(at_window,on_box,at_window,has_not)) ?Redo: move(state(at_window,on_floor,at_window,has_not),climb,state(at_window,on_box,at_window,has_not)) ?Exit: move(state(at_window,on_floor,at_window,has_not),push(at_window,_2902),state(_2902,on_floor,_2902,has_not)) ?Call: canget(state(_2902,on_floor,_2902,has_not)) ?Call: move(state(_2902,on_floor,_2902,has_not),_3844,_3845) ?Exit: move(state(_2902,on_floor,_2902,has_not),climb,state(_2902,on_box,_2902,has_not)) ?Call: canget(state(_2902,on_box,_2902,has_not)) ?Call: move(state(_2902,on_box,_2902,has_not),_5270,_5271) ?Exit: move(state(middle,on_box,middle,has_not),grasp,state(middle,on_box,middle,has)) ?Call: canget(state(middle,on_box,middle,has)) ?Exit: canget(state(middle,on_box,middle,has)) ?Exit: canget(state(middle,on_box,middle,has_not)) ?Exit: canget(state(middle,on_floor,middle,has_not)) ?Exit: canget(state(at_window,on_floor,at_window,has_not)) ?Exit: canget(state(at_door,on_floor,at_window,has_not)) ?yes


19

Monkey and Banana: Notes

Order of clauses in monkeybanana example:

1. grasp

2. climb

3. push

4. walk

Changing this order can cause Prolog to loop indefinitely !


20

Getting the necessary action sequence as output:

canget(state(_, _, _, has),[]).canget(OldState, [Move| Actions]) :-

move(OldState, Move, NewState),canget(NewState, Actions).

?- canget(state(at_door, on_floor,at_window, has_not),Actions).

Actions = [walk(at_door, at_window), push(at_window, middle), climb, grasp]

Yes


21

Why is there is an infinite number of solutions:

?- canget(state(at_door, on_floor, at_window, has_not),Actions).

Actions = [walk(at_door, at_window), push(at_window, middle), climb, grasp] ;

Actions = [walk(at_door, at_window), push(at_window, _G276), push(_G276, middle), climb, grasp] ;

Actions = [walk(at_door, at_window), push(at_window, _G276), push(_G276, _G287), push(_G287, middle), climb, grasp] ;

Date post:	18-Nov-2021
Category:	Documents
Upload:	others
View:	3 times
Download:	0 times

Logik für Informatiker: PROLOG Part 3: SLD Trees

Documents