Upload
estevan-daley
View
221
Download
1
Embed Size (px)
Citation preview
1
Finite Constraint DomainsFinite Constraint Domains
2
Finite Constraint DomainsFinite Constraint Domains
Constraint satisfaction problems (CSP) A backtracking solver Node and arc consistency Bounds consistency Generalized consistency Optimization for arithmetic CSPs
3
Finite Constraint DomainsFinite 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
4
Constraint Satisfaction Constraint Satisfaction ProblemsProblems
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 x D x xn D xn 1 1( ) ( )
5
Map ColouringMap Colouring
A classic CSP is the problem of coloring a map so that no adjacent regions have the same color
WA
NT
SA
Q
NSW
V
T
Can the map of Australia be colored with 3 colors ?
WA NT WA SA NT SA
NT Q SA Q SA NSW
SA V Q NSW NSW V
D WA D NT D SA D Q
D NSW D V D T
red yellow blue
( ) ( ) ( ) ( )
( ) ( ) ( )
{ , , }
6
4-queens4-queens
Place 4 queens on a 4 x 4 chessboard so that none can take another.
Q1 Q2 Q3 Q4
1
2
3
4
Four variables Q1, Q2, Q3, Q4 representing the row of the queen in each column. Domain of each variable is {1,2,3,4}
One solution! -->
7
4-queens4-queens
The constraints:
Q Q Q Q Q Q
Q Q Q Q Q Q
Q Q Q Q Q Q
Q Q Q Q Q Q
Q Q Q Q Q Q
Q Q Q Q Q Q
1 2 1 3 1 4
2 3 2 4 3 4
1 2 1 1 3 2 1 4 3
2 3 1 2 4 2 3 4 1
1 2 1 1 3 2 1 4 3
2 3 1 2 4 2 3 4 1
Not on the same row
Not diagonally up
Not diagonally down
8
Smugglers KnapsackSmugglers Knapsack
Smuggler with knapsack with capacity 9, who needs to choose items to smuggle to make profit at least 30
object profit size
whiskey
perfume
cigarretes
15 4
10 3
7 2
4 3 2 9 15 10 7 30W P C W P C
What should be the domains of the variables?
9
Systematic Search MethodsSystematic Search Methods
exploring the solution space complete and sound efficiency issues
Backtracking (BT) Generate & Test (GT)
exploringindividual assignments
exploring subspace
X
Z
Y
10
Generate & TestGenerate & 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 generatorinconsistencies --> backtracking
11
Backtracking (BT)Backtracking (BT)
incrementally extends a partial solution towards a complete solution
Algorithm:
assign value to variable
check consistency
until all variables labelled Drawbacks:
thrashing redundant work late detection of conflict
A
C
B
1
1
1
1
2
2
A = D, B D, A+C < 4
1
2
D21
1
12
GT & BT - ExampleGT & BT - Example
Problem:X::{1,2}, Y::{1,2}, Z::{1,2}
X = Y, X Z, Y > Z
generate & test backtracking
X Y Z test1 1 1 fail1 1 2 fail1 2 1 fail1 2 2 fail2 1 1 fail2 1 2 fail2 2 1 passed
X Y Z test1 1 1 fail
2 fail2 fail
2 1 fail2 1 passed
13
Simple Backtracking SolverSimple 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
14
Partial SatisfiablePartial Satisfiable
Check whether a constraint is unsatisfiable because of a prim. Constraint with no vars
Partial_satisfiablePartial_satisfiable(c) For each primitive constraint c in C
If vars(c) is empty If satisfiable(c) = false return false
Return true
15
Backtrack SolveBacktrack Solve
Back_solveBack_solve(c,d) If vars(c) is empty return partial_satisfiablepartial_satisfiable(c) Choose x in vars(c) For each value d in d(x)
Let C1 be C with x replaced by d If partial_satisfiable(partial_satisfiable(c1) then
If back_solveback_solve(c1,d) then return true Return false
16
Backtracking SolveBacktracking SolveX Y Y Z D X D Y D Z ( ) ( ) ( ) { , }1 2
X Y Y Z
Choose var X domain {1,2}
X 1
1 Y Y ZChoose var Y domain {1,2}
Y 1
1 1 1 Z
partial_satisfiablepartial_satisfiable false
Y 2
1 2 2 Z
Choose var Z domain {1,2}
Z 1
1 2 2 1 No variables, and false
Z 2
1 2 2 2
No variables, and false
Variable X domain {1,2}
X 2
2 Y Y ZChoose var Y domain {1,2}
2 1 1 Z 2 2 2 Z
17
Consistency TechniquesConsistency Techniques
removing inconsistent values from variables’ domains graph representation of the CSP
binary and unary constraints only (no problem!) nodes = variables edges = constraints
node consistency (NC) arc consistency (AC) path consistency (PC) (strong) k-consistency
A
B
C
A>5
AB
A<C
B=C
18
Node and Arc ConsistencyNode 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
19
Node ConsistencyNode 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
20
Node Consistency ExamplesNode Consistency Examples
Example CSP is not node consistent (see Z)X Y Y Z Z
D X D Y D Z
2
1 2 3 4( ) ( ) ( ) { , , , }
This CSP is node consistent
X Y Y Z Z
D X D Y D Z
2
1 2 3 4 1 2( ) ( ) { , , , }, ( ) { , }
The map coloring and 4-queens CSPs are node consistent. Why?
21
Achieving Node ConsistencyAchieving Node Consistency
Node_consistentNode_consistent(c,d) For each prim. Constraint c in C
D := node_consistent_primitivenode_consistent_primitive(c, D) Return D
Node_consistent_primitiveNode_consistent_primitive(c, D) If |vars(c)| =1 then
Let {x} = vars(c)
Return DD x d D x x d c( ): { ( )|{ } } is a solution of
22
Arc ConsistencyArc 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
And similarly for y A CSP is arc consistent if each prim.
Constraint in it is arc consistent
{ , }x d y e c is a solution of
23
Arc Consistency ExamplesArc Consistency Examples
This CSP is node consistent but not arc consistent X Y Y Z Z
D X D Y D Z
2
1 2 3 4 1 2( ) ( ) { , , , }, ( ) { , }
For example the value 4 for X and X < Y.
The following equivalent CSP is arc consistent
X Y Y Z Z
D X D Y D Z
2
( ) ( ) ( )
The map coloring and 4-queens CSPs are also arc consistent.
24
Achieving Arc ConsistencyAchieving Arc Consistency
Arc_consistent_primitiveArc_consistent_primitive(c, D) If |vars(c)| = 2 then
Return D Removes values which are not arc consistent with
c
D x d D x e D y
x d y e c
( ): { ( )| ( ),
{ , } }
exists
is a soln of
D y e D y d D x
x d y e c
( ): { ( )| ( ),
{ , } }
exists
is a soln of
25
Achieving Arc ConsistencyAchieving Arc Consistency
Arc_consistentArc_consistent(c,d) Repeat
W := d For each prim. Constraint c in C
D := arc_consistent_primitivearc_consistent_primitive(c,d) Until W = D Return D
A very naive version (there are much better)
26
Using Node and Arc Cons.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
27
Node and Arc Cons. SolverNode and Arc Cons. Solver
D := node_consistentnode_consistent(C,D) D := arc_consistentarc_consistent(C,D) if D is a false domain
return false if D is a valuation domain
return satisfiable(C,D) return unknown
28
Node and Arc Solver ExampleNode and Arc Solver ExampleColouring Australia: with constraints
WA NT WA SA NT SA
NT Q SA Q SA NSW
SA V Q NSW NSW V
WA red NT yellow
WA NT SA Q NSW V T
Node consistency
29
Node and Arc Solver ExampleNode and Arc Solver ExampleColouring Australia: with constraints
WA NT WA SA NT SA
NT Q SA Q SA NSW
SA V Q NSW NSW V
WA red NT yellow
WA NT SA Q NSW V T
Arc consistency
30
Node and Arc Solver ExampleNode and Arc Solver ExampleColouring Australia: with constraints
WA NT WA SA NT SA
NT Q SA Q SA NSW
SA V Q NSW NSW V
WA red NT yellow
WA NT SA Q NSW V T
Arc consistency
31
Node and Arc Solver ExampleNode and Arc Solver ExampleColouring Australia: with constraints
WA NT WA SA NT SA
NT Q SA Q SA NSW
SA V Q NSW NSW V
WA red NT yellow
WA NT SA Q NSW V T
Arc consistency
Answer:
unknown
32
Backtracking Cons. SolverBacktracking 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
33
Back. Cons Solver ExampleBack. Cons Solver Example
There is no possible value for
variable Q3!
Q1 Q2 Q3 Q4
1
2
3
4
No value can be
assigned toQ3 in this
case!
Therefore,we need to
choose another value
for Q2.
34
Back. Cons Solver ExampleBack. Cons Solver ExampleQ1 Q2 Q3 Q4
1
2
3
4
We cannot find any
possible valuefor Q4 inthis case!
Backtracking…
Find another value for Q3?
No!
backtracking,
Find anothervalue of Q2?
No!
backtracking,
Find anothervalue of Q1?
Yes, Q1 = 2
35
Back. Cons Solver ExampleBack. Cons Solver ExampleQ1 Q2 Q3 Q4
1
2
3
4
36
Back. Cons Solver ExampleBack. Cons Solver ExampleQ1 Q2 Q3 Q4
1
2
3
4
37
Node and Arc Solver ExampleNode and Arc Solver ExampleColouring 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
38
Is AC enough?Is AC enough?
empty domain => no solution cardinality of all domains is 1 => solution Problem:
X::{1,2}, Y::{1,2}, Z::{1,2} X Y, X Z, Y Z
1 2
1 2
1 2
X
YZ
39
Path Consistency (PC)Path Consistency (PC)
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
V0 V1
V3
V2 V4 V5
???
40
Path Consistency (PC)Path Consistency (PC)
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
V0 V1
V3
V2 V4 V5
???
41
K -consistencyK -consistency
K-consistency consistent valuation o (K-1) variables can be
extended to K-th variable strong K-consistency
J-consistency for each JK
42
Is k-consistency enough ?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
43
Consistency CompletenessConsistency Completeness strongly N-consistent constraint graph with N nodes
=> solution strongly K-consistent constraint graph with N nodes
(K<N) => ???
path consistentbut no solution
Special graph structures tree structured graph => (D)AC is enough
A
B C
D
{1,2,3}
{1,2,3}
{1,2,3}
{1,2,3}
44
hyper-arc consistencyhyper-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
45
Non binary constraints 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
46
Dual encodingDual 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
47
Hyper-arc consistencyHyper-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?
48
Bounds ConsistencyBounds 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)
49
Bounds ConsistencyBounds 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 is a solution of c
and similarly for An arithmetic CSP is bounds consistent if all
its primitive constraints are
{ min( , ), , }x D x x d xk dk 1 1
{ max( , )}x D x
50
Bounds Consistency ExamplesBounds Consistency Examples
X Y Z
D X D Y D Z
3 5
2 7 0 2 1 2( ) [ .. ], ( ) [ .. ], ( ) [ .. ]
Not bounds consistent, consider Z=2, then X-3Y=10
But the domain below is bounds consistent
D X D Y D Z( ) [ .. ], ( ) [ .. ], ( ) [ .. ] 2 7 0 2 0 1
Compare with the hyper-arc consistent domain
D X D Y D Z( ) { , , }, ( ) { , , }, ( ) { , } 3 5 6 0 1 2 0 1
51
Achieving Bounds Achieving Bounds ConsistencyConsistency
Given a current domain D we wish to modify the endpoints of domains so the result is bounds consistent
propagation rules do this
52
Achieving Bounds Achieving Bounds ConsistencyConsistency
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 D Y D Z X D Y D Z
Y D X D Z Y D X D Z
Z D X D Y Z D X D Y
min( , ) min( , ) max( , ) max( , )
min( , ) max( , ) max( , ) min( , )
min( , ) max( , ) max( , ) min( , )
Propagation rules for the constraint X = Y + Z
53
Achieving Bounds Achieving Bounds ConsistencyConsistency
X Y Z
D X D Y D Z
( ) [ .. ], ( ) [ .. ], ( ) [ .. ]4 8 0 3 2 2
The propagation rules determine that:
( ) ( )
( ) ( )
( ) ( )
0 2 2 5 3 2
4 2 2 6 8 2
4 3 1 8 8 0
X
Y
Z
Hence the domains can be reduced to
D X D Y D Z( ) [ .. ], ( ) [ .. ], ( ) [ .. ] 4 5 2 3 2 2
54
More propagation rulesMore propagation rules4 3 2 9
9
4
3
4
2
49
3
4
3
2
39
2
4
2
3
2
W P C
W D P D C
P D W D C
C D W D P
min( , ) min( , )
min( , ) min( , )
min( , ) min( , )
Given initial domain:D W D P D C( ) [ .. ], ( ) [ .. ], ( ) [ .. ] 0 9 0 9 0 9
We determine that
new domain:
W P C 9
4
9
3
9
2, ,
D W D P D C( ) [ .. ], ( ) [ .. ], ( ) [ .. ] 0 2 0 3 0 4
55
Disequations Disequations
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
Y Z
D Y D Z
D Y D Z
D Y D Z D Y D Z
( ) [ .. ], ( ) [ .. ]
( ) [ .. ], ( ) [ .. ]
( ) [ .. ], ( ) [ .. ] ( ) [ .. ], ( ) [ .. ]
2 4 2 3
2 4 3 3
2 4 2 2 3 4 2 2
no propagation
no propagation
prop
56
MultiplicationMultiplication X Y Z
If all variables are positive its simple enoughX D Y D Z X D Y D Z
Y D X D Z Y D X D Z
Z D X D Y Z D X D Y
min( , ) min( , ) max( , ) max( , )
min( , ) / max( , ) max( , ) / min( , )
min( , ) / max( , ) max( , ) / min( , )
But what if variables can be 0 or negative?
Example:
becomes:
D X D Y D Z( ) [ .. ], ( ) [ .. ], ( ) [ .. ] 4 8 1 2 1 3
D X D Y D Z( ) [ .. ], ( ) [ .. ], ( ) [ .. ] 4 6 2 2 2 3
57
MultiplicationMultiplication X Y Z
Calculate X bounds by examining extreme values
X D Y D Z D Y D Z
D Y D Z D Y D Z
minimum{min( , ) min( , ), min( , ) max( , )
max( , ) min( , ), max( , ) max( , )}
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 Y d Z d { , , / } 4 4
Recall we are using real numbers (e.g. 4/d)
58
MultiplicationMultiplication X Y Z
We can wait until the range of Z is non-negative or non-positive and then use rules like
Y D X D Z D X D Z
D X D Z D X D Z
minimum{min( , ) / min( , ), min( , ) / max( , )
max( , ) / min( , ), max( , ) / max( , )}
division by 0:
ve ve
0 0
0
0
59
Bounds Consistency AlgmBounds 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
60
Bounds Consistency ExampleBounds Consistency Example
Smugglers knapsack problem (no whiskey available)
capacity profit
W P C W P C4 3 2 9 15 10 7 30
D W D P D C( ) [ .. ], ( ) [ .. ], ( ) [ .. ] 0 0 0 9 0 9
W P C 102 15 12 10 60 7/ / /W P C 9 4 9 3 9 2/ / /
D W D P D C( ) [ .. ], ( ) [ .. ], ( ) [ .. ] 0 0 0 3 0 4
W P C 28 15 2 10 0 7/ / /
D W D P D C( ) [ .. ], ( ) [ .. ], ( ) [ .. ] 0 0 1 3 0 4
Continuing there is no further change
Note how we had to reexamine the profit constraint
61
Bounds consistency solverBounds consistency solver
D := bounds_consistentbounds_consistent(C,D) if D is a false domain
return false if D is a valuation domain
return satisfiable(C,D) return unknown
62
Back. Bounds Cons. SolverBack. Bounds Cons. Solver
Apply bounds consistency before starting the backtracking solver and after each variable is given a value
63
Back. Bounds Solver ExampleBack. Bounds Solver Example
Smugglers knapsack problem (whiskey available)capacity profit
W P C W P C4 3 2 9 15 10 7 30
D W D P D C( ) [ .. ], ( ) [ .. ], ( ) [ .. ] 0 9 0 9 0 9Current domain:
Initial bounds consistency
D W D P D C( ) [ .. ], ( ) [ .. ], ( ) [ .. ] 0 2 0 3 0 4
W = 0
D W D P D C( ) [ .. ], ( ) [ .. ], ( ) [ .. ] 0 0 1 3 0 4
P = 1
D W D P D C( ) [ .. ], ( ) [ .. ], ( ) [ .. ] 0 0 1 1 3 3
(0,1,3)
Solution Found: return true
64
Back. Bounds Solver ExampleBack. Bounds Solver Example
Smugglers knapsack problem (whiskey available)capacity profit
W P C W P C4 3 2 9 15 10 7 30 Current domain:
Initial bounds consistencyBacktrack
D W D P D C( ) [ .. ], ( ) [ .. ], ( ) [ .. ] 0 0 1 3 0 4
P = 2
D W D P D C( ) , ( ) , ( )
false
Backtrack
D W D P D C( ) [ .. ], ( ) [ .. ], ( ) [ .. ] 0 0 1 3 0 4
P = 3
D W D P D C( ) { }, ( ) { }, ( ) { } 0 3 0
W = 0
P = 1
(0,1,3) (0,3,0)
W = 1
(1,1,1)
W = 2
(2,0,0)
No more solutions
65
Generalized ConsistencyGeneralized 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
66
AlldifferentAlldifferent
alldifferent({V1,...,Vn}) holds when each variable V1,..,Vn takes a different value alldifferent({X, Y, Z}) is equivalent to
Arc consistent with domain
BUT there is no solution! specialized consistency for alldifferent can find it
X Y X Z Y Z
D X D Y D Z( ) { , }, ( ) { , }, ( ) { , } 1 2 1 2 1 2
67
Alldifferent ConsistencyAlldifferent 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
68
Alldifferent ExamplesAlldifferent Examplesalldifferent X Y Z
D X D Y D Z
({ , , })
( ) { , }, ( ) { , }, ( ) { , } 1 2 1 2 1 2
DV = {1,2}, V={X,Y,Z} hence detect unsatisfiabilityalldifferent X Y Z T
D X D Y D Z D T
({ , , , })
( ) { , }, ( ) { , }, ( ) { , }, ( ) { , , , } 1 2 1 2 1 2 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
69
Other Complex ConstraintsOther Complex Constraints
schedule n tasks with start times Si and durations Di needing resources Ri where L resources are available at each moment
array access if I = i, then X = Vi and if X != Vi then I != i
cumulative S S D D R R Ln n n([ , , ],[ , , ],[ , , ], )1 1 1
element I V V Xn( ,[ , , ], )1
70
Optimization for CSPsOptimization for CSPs
Because domains are finite can use a solver to build a straightforward optimizer
retry_int_optretry_int_opt(C, D, f, best) D2 := int_solvint_solv(C,D) if D2 is a false domain then return best let sol be the solution corresponding to D2 return retry_int_optretry_int_opt(C /\ f < sol(f), D, f, sol)
71
Retry Optimization ExampleRetry Optimization Example
Smugglers knapsack problem (optimize profit)minimize subject to
15 10 7
4 3 2 9 15 10 7 30
0 9 0 9 0 9
W P Ccapacity profit
W P C W P C
D W D P D C( ) [ .. ], ( ) [ .. ], ( ) [ .. ]
First solution found: D W D P D C( ) [ .. ], ( ) [ .. ], ( ) [ .. ] 0 0 1 1 3 3
Corresponding solution sol W P C{ , , } 0 1 3
sol f( ) 31
minimize subject to
15 10 7
4 3 2 9 15 10 7 30
15 10 7 31
0 9 0 9 0 9
W P Ccapacity profit
W P C W P C
W P C
D W D P D C( ) [ .. ], ( ) [ .. ], ( ) [ .. ]
Next solution found: D W D P D C( ) [ .. ], ( ) [ .. ], ( ) [ .. ] 1 1 1 1 1 1
sol W P C{ , , } 1 1 1
sol f( ) 32
minimize subject to
15 10 7
4 3 2 9 15 10 7 30
15 10 7 31 15 10 7 32
0 9 0 9 0 9
W P Ccapacity profit
W P C W P C
W P C W P C
D W D P D C( ) [ .. ], ( ) [ .. ], ( ) [ .. ]
No next solution!
Return best solution
72
Backtracking OptimizationBacktracking 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)
73
Back. Optimization ExampleBack. Optimization Example
Smugglers knapsack problem (whiskey available)capacity profit
W P C W P C4 3 2 9 15 10 7 30
D W D P D C( ) [ .. ], ( ) [ .. ], ( ) [ .. ] 0 9 0 9 0 9Current domain:
Initial bounds consistency
D W D P D C( ) [ .. ], ( ) [ .. ], ( ) [ .. ] 0 2 0 3 0 4
W = 0
D W D P D C( ) [ .. ], ( ) [ .. ], ( ) [ .. ] 0 0 1 3 0 4
P = 1
D W D P D C( ) [ .. ], ( ) [ .. ], ( ) [ .. ] 0 0 1 1 3 3
(0,1,3)
Solution Found: add constraint
Smugglers knapsack problem (whiskey available)capacity profit
W P C W P C
W P C
4 3 2 9 15 10 7 30
15 10 7 31
74
Back. Optimization ExampleBack. Optimization Example
Initial bounds consistency
capacity profit
W P C W P C
W P C
4 3 2 9 15 10 7 30
15 10 7 31
P = 2
false
P = 3
false
W = 1
(1,1,1)
Modify constraint
capacity profit
W P C W P C
W P C
4 3 2 9 15 10 7 30
15 10 7 32
W = 0
P = 1
(0,1,3)
Smugglers knapsack problem (whiskey available)
W = 2
false
Return last sol (1,1,1)
75
Branch and Bound Opt.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
use the current best solution to constrain prob. ( , ) ( , )C x d f C x d f
76
Branch and Bound ExampleBranch and Bound ExampleSmugglers knapsack problem
{ / , , }W P C 11 4 0 0W 2 W 3
{ , / , }W P C 2 1 3 0P 0 P 1
{ , , / }W P C 2 0 1 2C 0 C 1
{ , , }W P C 2 0 0Solution (2,0,0) = 30
{ / , , }W P C 7 4 0 1W 1 W 2
{ , , / }W P C 1 0 5 2C 2 C 3
false { / , , }W P C 3 4 0 3W 0 W 1
false false
false
{ / , , }W P C 6 4 1 0W 1 W 2
{ , / , }W P C 1 5 3 0P 1 P 2
{ , , }W P C 1 1 1Solution (1,1,1) = 32
Worse than best sol
false
false
77
Finite Constraint Domains Finite Constraint Domains SummarySummary
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