Finite Constraint Domains

1
 Finite Constraint Domains
 Constraint satisfaction problems (CSP)
 A backtracking solver
 Node and arc consistency
 Bounds consistency
 Generalized consistency
 Optimization for arithmetic CSPs

2
 Finite Constraint Domains
 An important class of constraint domains
 Use to model constraint problems involving
choice: e.G. Scheduling, routing and
timetabling
 The greatest industrial impact of constraint
programming has been on these problems

3
 Constraint Satisfaction
Problems
 A constraint satisfaction problem (CSP)
consists of:
 A constraint C over variables x1,..., Xn
 A domain D which maps each variable xi to a
set of possible values d(xi)
 It is understood as the constraint
C  x1  D( x1)  xn  D( xn)

4
 Map Colouring
A classic CSP is the problem of coloring a map so that
no adjacent regions have the same color
Can the map of Australia be
colored with 3 colors ?
NT Q
WA  NT  WA  SA  NT  SA 
NT  Q  SA  Q  SA  NSW  WA
SA
SA  V  Q  NSW  NSW  V NSW

D(WA)  D( NT )  D( SA)  D(Q)  V

D( NSW )  D(V )  D(T ) 

T
{red , yellow, blue}
5
 4-queens
Place 4 queens on a 4 x 4 chessboard so that none can
take another.
Q1 Q2 Q3 Q4
Four variables Q1, Q2,
Q3, Q4 representing the 1
row of the queen in each
column. Domain of each 2
variable is {1,2,3,4}
3

One solution! --> 4

6
 4-queens
The constraints:

Not on the same row

Q1  Q2  Q1  Q3  Q1  Q4 
Q2  Q3  Q 2  Q 4  Q3  Q4 
Q1  Q2  1  Q1  Q3  2  Q1  Q4  3 
Not diagonally up
Q2  Q3  1  Q2  Q 4  2  Q 3  Q4  1 
Q1  Q2  1  Q1  Q3  2  Q1  Q4  3 
Not diagonally down Q2  Q3  1  Q 2  Q 4  2  Q3  Q 4  1

7
 Smugglers Knapsack
Smuggler with knapsack with capacity 9, who
needs to choose items to smuggle to make
profit at least 30
object profit size
whiskey 15 4
perfume 10 3
cigarretes 7 2

4W  3 P  2C  9  15W  10 P  7C  30

What should be the domains of the variables? 8

 Systematic Search Methods
 exploring the solution space
 complete and sound
 efficiency issues exploring subspace

 Backtracking (BT)
 Generate & Test (GT)
exploring
Z
individual assignments
Y
X9
 Generate & Test
 probably the most general problem solving method
 Algorithm:
generate labelling
test satisfaction

Drawbacks: Improvements:
blind generator smart generator
--> local search
late discovery of testing within generator
inconsistencies --> backtracking

10
 Backtracking (BT)
 incrementally extends a partial solution towards a complete
solution
 Algorithm:
assign value to variable A1
check consistency
until all variables labelled B 2
1
 Drawbacks:
 thrashing C 2
1 1
 redundant work
 late detection of conflict
D
1 2 1 2 1


A = D, B  D, A+C < 4
11
 GT & BT - Example
 Problem:
X::{1,2}, Y::{1,2}, Z::{1,2}
X = Y, X  Z, Y > Z
X Y Z test X Y Z test
1 1 1 fail 1 1 1 fail
1 1 2 fail 2 fail
1 2 1 fail 2 fail
1 2 2 fail 2 1 fail
2 1 1 fail 2 1 passed
2 1 2 fail
2 2 1 passed
generate & test backtracking
12
 Simple Backtracking Solver
 The simplest way to solve CSPs is to enumerate
the possible solutions
 The backtracking solver:
 Enumerates values for one variable at a time
 Checks that no prim. Constraint is false at each stage
 Assume satisfiable(c) returns false when
primitive constraint c with no variables is
unsatisfiable

13
 Partial Satisfiable
 Check whether a constraint is unsatisfiable
because of a prim. Constraint with no vars
 Partial_satisfiable(c)
Partial_satisfiable
 For each primitive constraint c in C
 If vars(c) is empty

 If satisfiable(c) = false return false

 Return true

14
 Backtrack Solve
 Back_solve(c,d)
Back_solve
 If vars(c) is empty return partial_satisfiable(c)
partial_satisfiable
 Choose x in vars(c)
 For each value d in d(x)
 Let C1 be C with x replaced by d

 If partial_satisfiable(c1)
partial_satisfiable( then
 If back_solve(c1,d)
back_solve then return true
 Return false
15
 Backtracking Solve
X  Y Y  Z D( X )  D(Y )  D( Z )  {1,2}
X  Y Y  Z
X 1 X 2
Choose
Variablevar
Choose XZ
var X
Y
Y
domain
domain {1,2}
domain {1,2}
{1,2} 1 Y Y  Z 2  Y Y  Z
Y 1 Y2 2  1 1  Z 2  22  Z

1  1 1  Z 1 22  Z
Z 1 No variables,
Z 2
partial_satisfiable
No variables, false and false
and false 1 22 1 1 22  2 16
 Consistency Techniques
 removing inconsistent values from variables’ domains
 graph representation of the CSP
 binary and unary constraints only (no problem!)
 nodes = variables
 edges = constraints A>5
 node consistency (NC)
A
 arc consistency (AC) A<
A
 path consistency (PC) B C
 (strong) k-consistency C
B B=
C
17
 Node and Arc Consistency
 Basic idea: find an equivalent CSP to the
original one with smaller domains of vars
 Key: examine 1 prim.Constraint c at a time
 Node consistency: (vars(c)={x}) remove any
values from domain of x that falsify c
 Arc consistency: (vars(c)={x,y}) remove any
values from d(x) for which there is no value in
d(y) that satisfies c and vice versa
18
 Node Consistency
 Primitive constraint c is node consistent
with domain D if |vars(c)| /=1 or
 If vars(c) = {x} then for each d in d(x)
 X assigned d is a solution of c
 A CSP is node consistent if each prim.
Constraint in it is node consistent

19
 Node Consistency Examples
Example CSP is not node consistent (see Z)
X  Y Y  Z  Z  2
D( X )  D(Y )  D( Z )  {1,2,3,4}

This CSP is node consistent

X  Y Y  Z  Z  2
D( X )  D(Y )  {1,2,3,4}, D( Z )  {1,2}

The map coloring and 4-queens CSPs are node

consistent. Why?
20
 Achieving Node Consistency
 Node_consistent(c,d)
Node_consistent
 For each prim. Constraint c in C
 D := node_consistent_primitive(c,
node_consistent_primitive D)
 Return D
 Node_consistent_primitive(c,
Node_consistent_primitive D)
 If |vars(c)| =1 then
 Let {x} = vars(c)

 Return ):  {d  D( x )|{x  d } is a solution of c}

D( xD
21
 Arc Consistency
 A primitive constraint c is arc consistent
with domain D if |vars{c}| != 2 or
 Vars(c) = {x,y} and for each d in d(x) there
exists e in d(y) such that
{x  d , y  e} is a solution of c
 And similarly for y
 A CSP is arc consistent if each prim.
Constraint in it is arc consistent
22
 Arc Consistency Examples
This CSP is node consistent but not arc consistent
X  Y Y  Z  Z  2
D( X )  D(Y )  {1,2,3,4}, D( Z )  {1,2}
For example the value 4 for X and X < Y.
The following equivalent CSP is arc consistent
X  Y Y  Z  Z  2
D( X )  D(Y )  D( Z )  

The map coloring and 4-queens CSPs are also arc

consistent. 23
 Achieving Arc Consistency
 Arc_consistent_primitive(c,
Arc_consistent_primitive D)
 If |vars(c)| = 2 then
D( x ):  {d  D( x )| exists e  D( y ),
{x  d , y  e} is a soln of c}
D( y ):  {e  D( y )| exists d  D( x ),
 Return D {x  d , y  e} is a soln of c}
 Removes values which are not arc consistent with c

24
 Achieving Arc Consistency
 Arc_consistent(c,d)
Arc_consistent
 Repeat
 W := d

 For each prim. Constraint c in C

 D := arc_consistent_primitive(c,d)
arc_consistent_primitive
 Until W = D
 Return D
 A very naive version (there are much better)
25
 Using Node and Arc Cons.
 We can build constraint solvers using the
consistency methods
 Two important kinds of domain
 False domain: some variable has empty domain
 Valuation domain: each variable has a
singleton domain
 Extend satisfiable to CSP with val. Domain

26
 Node and Arc Cons. Solver
 D := node_consistent(C,D)
node_consistent
 D := arc_consistent(C,D)
arc_consistent
 if D is a false domain
 return false
 if D is a valuation domain
 return satisfiable(C,D)
 return unknown
27
 Node and Arc Solver Example
Colouring Australia: with constraints
WA  red  NT  yellow
WA NT SA Q NSW V T

Node
consistency

WA  NT WA  SA NT  SA
NT  Q SA  Q SA  NSW
SA  V Q  NSW NSW  V
28
 Node and Arc Solver Example
Colouring Australia: with constraints
WA  red  NT  yellow
WA NT SA Q NSW V T

Arc
consistency

WA  NT WA  SA NT  SA
NT  Q SA  Q SA  NSW
SA  V Q  NSW NSW  V 29
 Node and Arc Solver Example
Colouring Australia: with constraints
WA  red  NT  yellow
WA NT SA Q NSW V T

Arc
consistency

WA  NT WA  SA NT  SA
NT  Q SA  Q SA  NSW
SA  V Q  NSW NSW  V 30
 Node and Arc Solver Example
Colouring Australia: with constraints
WA  red  NT  yellow
WA NT SA Q NSW V T

Arc
consistency

Answer:
WA  NT WA  SA NT  SA
unknown NT  Q SA  Q SA  NSW
SA  V Q  NSW NSW  V 31
 Backtracking Cons. Solver
 We can combine consistency with the
backtracking solver
 Apply node and arc consistency before
starting the backtracking solver and after
each variable is given a value

32
 Back. Cons Solver Example
Q1 Q2 Q3 Q4

1
Therefore,
No value
There is no
wecan
need
be to
2
possible
choose to
assigned
value for
another value
Q3 in this 3
variable Q3!
for Q2.
case!
4

33
 Back. Cons Solver Example
Q1 Q2 Q3 Q4

1
backtracking,
Backtracking…
backtracking,
We cannot
find any 2
Find another
Find another
possible
value of value
Q1?
value
value for
of
for Q4 inQ3? 3
Q2?
No!
this case!
Yes, Q1 =2
4

34
Back. Cons Solver Example
Q1 Q2 Q3 Q4

35
Back. Cons Solver Example
Q1 Q2 Q3 Q4

36
 Node and Arc Solver Example
Colouring Australia: with constraints
WA  red  NT  yellow
WA NT SA Q NSW V T

Backtracking
enumeration

Select a variable with domain of more than 1, T

Add constraint T  red Apply consistency
Answer: true 37
 Is AC enough?
 empty domain => no solution
 cardinality of all domains is 1 => solution
 Problem: X
X::{1,2}, Y::{1,2}, Z::{1,2}
1 2
X  Y, X  Z, Y  Z

1 2
Y
Z
1 2

38
 Path Consistency (PC)
V2 V4
V3 V5
V0 V1
???

 Path (V0 … Vn) is path consistent iff for each pair of

compatible values x in D(V0) and y in D(Vn) there exists
an instantiation of the variables V1 ..Vn-1 such that all the
constraint (Vi,Vi+1) are satisfied
 CSP is path consistent iff each path is path consistent
 consistency along the path only

39
 Path Consistency (PC)
V2 V4
V3 V5
V0 V1
???

 checking paths of length 2 is enough

 Plus/Minus
+ detects more inconsistencies than AC
- extensional representation of constraints (01
matrix), huge memory consumption
- changes in graph connectivity
40
 K -consistency
 K-consistency
 consistent valuation o (K-1) variables can be
extended to K-th variable
 strong K-consistency
 J-consistency for each JK

41
 Is k-consistency enough ?
 If all the domains have cardinality 1=> solution
 If any domain is empty => no solution
 Otherwise ?

 strongly k-consistent constraint graph with k

nodes => completeness
 for some special kind of graphs weaker forms of
consistency are enough

42
 Consistency Completeness
 strongly N-consistent constraint graph with N nodes =>
solution
 strongly K-consistent constraint graph with N nodes
(K<N) => ???

A D
{1,2,3} {1,2,3}
path consistent   
but no solution 
B  C
{1,2,3 {1,2,3}
 Special graph structures }
 tree structured graph => (D)AC is enough

43
 hyper-arc consistency
 A primitive constraint c is hyper-arc
consistent with domain D if
 for each variable x in c and for each d in d(x) the valuation
x-->d can be extended to a solution of c
 A CSP is hyper-arc consistent if each prim.
Constraint in it is hyper-arc consistent
 NC  hyper-arc consistency for constraint with 1 var
 AC  hyper-arc consistency for constraint with 2 var

44
 Non binary constraints
 What about prim. constraints with more
than 2 variables?
 Each CSP can be transformed into an
equivalent binary CSP (dual encoding)
 k-consitncy
 hyper-arc consistency

45
 Dual encoding
 Each CSP can be transformed into an
equivalent binary CSP by ``dual encoding’’:
 k-ary constraint c is converted to a dual variable vc with
the domain consisting of compatible tuples
 for each pair of constraint c,c’ sharing some variable there
is a binary constraint between vc and vc’ restricting the
dual variables to tuples in which the original shared
variables take the same value

46
 Hyper-arc consistency
 hyper-arc consistency: extending arc
consistency to arbitrary number of variables
 (better than binary representation)
 Unfortunately determining hyper-arc
consistency is NP-hard (as expensive as
determinining if a CSP is satisfiable).
 Solution?

47
 Bounds Consistency
 arithmetic CSP: domains = integers
 range: [l..u] represents the set of integers {l,
l+1, ..., u}
 idea use real number consistency and only
examine the endpoints (upper and lower
bounds) of the domain of each variable
 Define min(D,x) as minimum element in
domain of x, similarly for max(D,x)
48
 Bounds Consistency
 A prim. constraint c is bounds consistent with
domain D if for each var x in vars(c)
 exist real numbers d1, ..., dk for remaining vars
x1, ..., xk such that min(D,xj),= dj<= max(D,xj) and
{x  min( D, x ), x1  d 1, xk is adk }
solution of c
 and similarly for
{x  max( D, x )}
 An arithmetic CSP is bounds consistent if all its
primitive constraints are
49
 Bounds Consistency Examples
X  3Y  5Z
D( X )  [2..7], D(Y )  [0..2], D( Z )  [ 1..2]

Not bounds consistent, consider Z=2, then X-3Y=10

But the domain below is bounds consistent
D( X )  [ 2..7], D(Y )  [ 0..2], D( Z )  [0..1]
Compare with the hyper-arc consistent domain

D( X )  {3,5,6}, D(Y )  {0,1,2}, D( Z )  {0,1} 50

 Achieving Bounds
Consistency
 Given a current domain D we wish to
modify the endpoints of domains so the
result is bounds consistent
 propagation rules do this

51
 Achieving Bounds
Consistency
Consider the primitive constraint X = Y + Z which
is equivalent to the three forms
X  Y  Z Y  X  Z Z  X Y

Reasoning about minimum and maximum values:

X  min( D, Y )  min( D, Z ) X  max( D, Y )  max( D, Z )
Y  min( D, X )  max( D, Z ) Y  max( D, X )  min( D, Z )
Z  min( D, X )  max( D, Y ) Z  max( D, X )  min( D, Y )

Propagation rules for the constraint X = Y + Z 52

 Achieving Bounds
Consistency
X YZ
D( X )  [4..8], D(Y )  [0..3], D( Z )  [2..2]

The propagation rules determine that:

( 0  2  ) 2  X  5 (  3  2)
( 4  2  ) 2  Y  6 (  8  2)
(4  3  ) 1  Z  8 (  8  0)
Hence the domains can be reduced to
D( X )  [4..5], D(Y )  [2..3], D( Z )  [2..2]
53
 More propagation rules
4W  3 P  2C  9
9 3 2
W   min( D, P )  min( D, C )
4 4 4
9 4 2
P   min( D,W )  min( D, C )
3 3 3
9 4 3
C   min( D, W )  min( D, P )
2 2 2
Given initial domain:
D(W )  [0..9], D( P )  [0..9], D(C )  [0..9]
9 9 9
We determine that W , P , C
4 3 2
new domain: D(W )  [0..2], D( P)  [0..3], D(C )  [0..544]
 Disequations Y  Z

Disequations give weak propagation rules, only when

one side takes a fixed value that equals the minimum or
maximum of the other is there propagation

D(Y )  [2..4], D( Z )  [2..3] no propagation

D(Y )  [2..4], D( Z )  [3..3] no propagation
D(Y )  [2..4], D( Z )  [2..2] prop D(Y )  [ 3..4], D( Z )  [2..2]

55
 Multiplication X  Y  Z
If all variables are positive its simple enough
X  min( D, Y )  min( D, Z ) X  max( D, Y )  max( D, Z )
Y  min( D, X ) / max( D, Z ) Y  max( D, X ) / min( D, Z )
Z  min( D, X ) / max( D, Y ) Z  max( D, X ) / min( D, Y )

Example: D( X )  [4..8], D(Y )  [1..2], D( Z )  [1..3]

becomes: D( X )  [4..6], D(Y )  [2..2], D( Z )  [2..3]

But what if variables can be 0 or negative? 56

 Multiplication X  Y  Z
Calculate X bounds by examining extreme values
X  minimum{min( D, Y )  min( D, Z ), min( D, Y )  max( D, Z )
max( D, Y )  min( D, Z ), max( D, Y )  max( D, Z )}
Similarly for upper bound on X using maximum
BUT this does not work for Y and Z? As long as
min(D,Z) <0 and max(D,Z)>0 there is no bounds
restriction on Y
X  Y  Z { X  4, Y  d , Z  4 / d }
Recall we are using real numbers (e.g. 4/d) 57
 Multiplication X  Y  Z
We can wait until the range of Z is non-negative or
non-positive and then use rules like
Y  minimum{min( D, X ) / min( D, Z ), min( D, X ) / max( D, Z )
max( D, X ) / min( D, Z ), max( D, X ) / max( D, Z )}
division by 0:
 ve  ve 0
     
0 0 0

58
 Bounds Consistency Algm
 Repeatedly apply the propagation rules for
each primitive constraint until there is no
change in the domain
 We do not need to examine a primitive
constraint until the domains of the variables
involve are modified

59
 Bounds Consistency Example

Smugglers knapsack problem (no whiskey available)

capacity profit
4W  3 P  2C  9  15W  10 P  7C  30
D((W
D  [[00....00],
W))  ], D
D(( P  [[0
P))  1..
..39
3],], D C))  [[00....449]]
D((C
 
WW 102
W / 15
289 4 PPP92
//15 12 / 10C C
//310 9/02/ 60
C 7 /7
Continuing there is no further change
Note how we had to reexamine the profit
constraint
60
 Bounds consistency solver
 D := bounds_consistent(C,D)
bounds_consistent
 if D is a false domain
 return false
 if D is a valuation domain
 return satisfiable(C,D)
 return unknown

61
 Back. Bounds Cons. Solver
 Apply bounds consistency before starting
the backtracking solver and after each
variable is given a value

62
 Back. Bounds Solver Example
Smugglers knapsack problem (whiskey available)
capacity profit
4W  3 P  2C  9  15W  10 P  7C  30
Current domain:
D W))  [[00....9
D((W 20],],D P))  [[0
D((P ..3
1..19],
], D C))  [[3
D((C 9]]]
00....344

Initial bounds consistency

W=0
P=1 Solution Found: return true
63
(0,1,3)
 Back. Bounds Solver Example
Smugglers knapsack problem (whiskey available)
capacity profit
4W  3 P  2C  9  15W  10 P  7C  30
Current domain:
D(W
DD (W)[0)..{
)(W 
00],
},D
,D(P
(P) )[1{
..33},
,],DD
D
(C ) ))
((C
C {[00}..4]

Backtrack
Backtrack Initial bounds consistency
W=0 W=1 W=2
P=1 P=2 P=3 (1,1,1) (2,0,0)
(0,1,3) false (0,3,0) No more solutions64
 Generalized Consistency
 Can use any consistency method with any
other communicating through the domain,
 node consistency : prim constraints with 1 var
 arc consistency: prim constraints with 2 vars
 bounds consistency: other prim. constraints
 Sometimes we can get more information by
using complex constraints and special
consistency methods
65
 Alldifferent
 alldifferent({V1,...,Vn}) holds when each
variable V1,..,Vn takes a different value
 alldifferent({X, Y, Z}) is equivalent to
X  Y  X  Z Y  Z
 Arc consistent with domain
D( X )  {1,2}, D(Y )  {1,2}, D( Z )  {1,2}
 BUT there is no solution! specialized
consistency for alldifferent can find it
66
 Alldifferent Consistency
 let c be of the form alldifferent(V)
 while exists v in V where D(v) = {d}
 V := V - {v}
 for each v’ in V
 D(v’) := D(v’) - {d}

 DV := union of all D(v) for v in V

 if |DV| < |V| then return false domain
 return D
67
 Alldifferent Examples
alldifferent ({ X , Y , Z})
D( X )  {1,2}, D(Y )  {1,2}, D( Z )  {1,2}
DV = {1,2}, V={X,Y,Z} hence detect unsatisfiability
alldifferent ({ X , Y , Z , T})
D( X )  {1,2}, D(Y )  {1,2}, D( Z )  {1,2}, D( T )  {2,3,4,5}
DV = {1,2,3,4,5}, V={X,Y,Z,T} don’t detect unsat.
Maximal matching based consistency could
X Y Z T

1 2 3 4 5
68
 Other Complex Constraints
cumulative([ S1 , , Sn ],[ D1 , , Dn ],[ R1 , , Rn ], L)
 schedule n tasks with start times Si and
durations Di needing resources Ri where L
resources are available at each moment
element ( I ,[V1 , ,Vn ], X )
 array access if I = i, then X = Vi and if X !=
Vi then I != i

69
 Optimization for CSPs
 Because domains are finite can use a solver
to build a straightforward optimizer
 retry_int_opt(C,
retry_int_opt D, f, best)
 D2 := int_solv(C,D)
int_solv
 if D2 is a false domain then return best
 let sol be the solution corresponding to D2
 return retry_int_opt(C
retry_int_opt /\ f < sol(f), D, f, sol)

70
 Retry Optimization Example
Smugglers knapsack problem (optimize profit)
minimize 15
minimize 15WW10 10PP77CCsubject
subjectto to
capacity
capacity profit
profit
W33PP22CC99  15
44W 15W W10 10PP77CC30 30
 15WD(W10
) P[0..715 D
9C],W 10)
( 31
P 
P [015 9],D
W
7..C 10 CP) 
(31  7[ 0C..9]32
D W))[[00....99],],D
D((W D((PP))[[00....99],],D
D((CC))[[00....99]]

First
Next
No next
solution
solution
solution!
found: (W) )[0[1....01],],D
found: DD(W D((PP)) [[11....11],],D C)) [[31..1
D((C 3]

Corresponding
Return sol{W
solution sol
best solution {W  01
, ,PP 11, C
, C 31}}
sol ( f )  32
31 71
 Backtracking Optimization
 Since the solver may use backtrack search
anyway combine it with the optimization
 At each step in backtracking search, if best
is the best solution so far add the constraint
f < best(f)

72
 Back. Optimization Example
Smugglers knapsack problem (whiskey available)
capacity
Smugglers profitavailable)
knapsack problem (whiskey
 3 P  2C  9
4W capacity  15W  10 P  7C  30
profit
4W  3 P  2C15
W9  10 W
 P15 7C10P
31 7C  30
Current domain:
D W))  [[00....9
D((W 20],],D P))  [[0
D((P ..3
1..19],
], D C))  [[3
D((C 9]]]
00....344

Initial bounds consistency

W=0
Solution Found: add constraint
P=1
73
(0,1,3)
 Back. Optimization Example
Smugglers knapsack problem (whiskey available)
capacity profit
4W  3 P  2C  9  15W  10 P  7C  30
 15W  10 P  7C  32
31

Initial bounds consistency

W=0 W=1 W=2
P=1 P=2 P=3 (1,1,1) false
74
(0,1,3) false false Modify constraint
Return last sol (1,1,1)
 Branch and Bound Opt.
 The previous methods,unlike simplex don't use the
objective function to direct search
 branch and bound optimization for (C,f)
 use simplex to find a real optimal,
 if solution is integer stop
 otherwise choose a var x with non-integer opt value d
and examine the problems

  xcurrent
(Cthe
use   d best (C  to
, f )solution x constrain
 d  , f )prob.
75
 Branch and Bound Example
Smugglers knapsack problem
{W 
 211 / 4, P  0, CW  03}
{W P 20, P  1 / 3, CP 10} false

W 0 2, P  0C, C
{C  6 1/ 4, P  W1, 
 1 1 / 2{}W W C 2 0}
{W  2,(2,0,0)
Solution P  0=, C30 0} {W P 1,1P  P5  / 32, Cfalse
 0}

{W  17 / 4, P  W0, C2  Solution
{}
1W  1,(1,1,1)
P  1= , C32 1}
C  21, P  C
{W  3 5 false
0, C / 2} Worse than best sol
false W  03 / 4, W
{W P  10, C  3}
false false 76
 Finite Constraint Domains
Summary
 CSPs form an important class of problems
 Solving of CSPs is essentially based on
backtracking search
 Reduce the search using consistency methods
 node, arc, bound, generalized
 Optimization is based on repeated solving or
using a real optimizer to guide the search

77