Constraint Satisfaction Problems

Must be
Hot&Sour
Soup
No
Chicken
Appetizer Peanuts
Dish
Total Cost
< $30
Pork Dish No
Vegetable Peanuts

Not Both Seafood Rice

Spicy Not
Chow Mein
Constraint Network
1
 Formal Definition of CSP
• A constraint satisfaction problem (CSP) is
a triple (V, D, C) where
– V is a set of variables X1, ... , Xn.
– D is the union of a set of domain sets
D1,...,Dn, where Di is the domain of possible
values for variable Xi.
– C is a set of constraints on the values of the
variables, which can be pairwise (simplest
and most common) or k at a time.
2
 CSPs vs. Standard Search Problems

• Standard search problem:

•
– state is a "black box“ – any data structure that supports
successor function, heuristic function, and goal test
–
• CSP:
•
– state is defined by variables Xi with values from domain Di
–
– goal test is a set of constraints specifying allowable
combinations of values for subsets of variables
–

• Simple example of a formal representation language 3

 Example: Map-Coloring

• Variables WA, NT, Q, NSW, V, SA, T

• Domains Di = {red,green,blue}
• Constraints: adjacent regions must have different colors
•
• e.g., WA ≠ NT, or (WA,NT) in {(red,green),(red,blue),(green,red),
(green,blue),(blue,red),(blue,green)}
•
4
 Example: Map-Coloring

• Solutions are complete and consistent

assignments, e.g., WA = red, NT = green,Q =
red,NSW = green,V = red,SA = blue,T = green
• 5
 Constraint graph
• Binary CSP: each constraint relates two variables
•
• Constraint graph: nodes are variables, arcs are
constraints

6
 Varieties of constraints
• Unary constraints involve a single variable,
– e.g., SA ≠ green
–

• Binary constraints involve pairs of variables,

– e.g., value(SA) ≠ value(WA)
–

• Higher-order constraints involve 3 or more

variables,
– e.g., cryptarithmetic column constraints
7
 Example: Cryptarithmetic

• Variables:
{F, T, U, W, R, O, X1, X2, X3}
• Domains: {0,1,2,3,4,5,6,7,8,9}
• Constraints: Alldiff (F,T,U,W,R,O)
•
– O + O = R + 10 · X1
–
– X1 + W + W = U + 10 · X2
– 8
– X2 + T + T = O + 10 · X3
 Example: Latin Squares Puzzle

X1 X2 X3 X4 red RT RS RC RO
X5 X6 X7 X8 green GT GS GC GO
X9 X10 X11 X12 blue BT BS BC BO
X13 X14 X15 X16 yellow YT YS YC YO

Variables Values

Constraints: In each row, each column, each major diagonal, there must
be no two markers of the same color or same shape.

How can we formalize this?

V:
D:
C:
9
 Real-world CSPs
• Assignment problems
– e.g., who teaches what class
–
• Timetabling problems
•
– e.g., which class is offered when and where?
–
• Transportation scheduling
•
• Factory scheduling

Notice that many real-world problems involve 10

real-valued variables
 The Consistent Labeling Problem

• Let P = (V,D,C) be a constraint satisfaction problem.

• An assignment is a partial function f : V -> D that assigns

a value (from the appropriate domain) to each variable

• A consistent assignment or consistent labeling is an

assignment f that satisfies all the constraints.

• A complete consistent labeling is a consistent labeling

in which every variable has a value.

11
 Standard Search Formulation

• state: (partial) assignment

• initial state: the empty assignment { }
• successor function: assign a value to an unassigned variable that
does not conflict with current assignment
 fail if no legal assignments

• goal test: the current assignment is complete

(and is a consistent labeling)

1. This is the same for all CSPs regardless of application.

2. Every solution appears at depth n with n variables

 we can use depth-first search.

3. Path is irrelevant, so we can also use complete-state formulation.

12
What Kinds of Algorithms are used for CSP?

• Backtracking Tree Search

• Tree Search with Forward Checking

• Tree Search with Discrete Relaxation (arc consistency,

k-consistency)

• Many other variants

• Local Search using Complete State Formulation

13
 Backtracking Tree Search
• Variable assignments are commutative}, i.e.,
[ WA = red then NT = green ] same as [ NT = green then WA = red ]

• Only need to consider assignments to a single variable at each node.

• Depth-first search for CSPs with single-variable assignments is called

backtracking search.

• Backtracking search is the basic uninformed algorithm for CSPs.

• Can solve n-queens for n ≈ 25.

14
 Graph Matching Example
Find a subgraph isomorphism from R to S.
R 1 2

(1,a) (1,b) (1,c) (1,d) (1,e)

3 4
(2,a) (2,b) (2,c) (2,d) (2,e)
X X X

S e
(3,a) (3,b) (3,c) (3,d) (3,e) (3,a) (3,b) (3,c) (3,d) (3,e)
X X X X X X X X X
a c
(4,a) (4,b) (4,c) (4,d) (4,e)
X X X X

b d
How do we formalize this problem? 15
Backtracking Search

16
Backtracking Example

17
Backtracking Example

18
Backtracking Example

19
Backtracking Example

20
 Improving Backtracking Efficiency
• General-purpose methods can give huge
gains in speed:
•
– Which variable should be assigned next?
–
– In what order should its values be tried?
–
– Can we detect inevitable failure early?
–
21
 Most Constrained Variable
• Most constrained variable:
choose the variable with the fewest legal values

• a.k.a. minimum remaining values (MRV)

heuristic
•
22
 Most Constraining Variable
• Tie-breaker among most constrained
variables
• Most constraining variable:
•
– choose the variable with the most constraints
on remaining variables
–

23
 Least Constraining Value
• Given a variable, choose the least
constraining value:
•
– the one that rules out the fewest values in the
remaining variables
–

• Combining these heuristics makes 1000

24
 Forward Checking
(Haralick and Elliott, 1980)

Variables: U = {u1, u2, … , un}

Values: V = {v1, v2, … , vm}
Constraint Relation: R = {(ui,v,uj,v’) | ui having value
v is compatible with uj having label v’}

ui,v uj,v’

If (ui,v,uj,v’) is not in R, they are incompatible,

meaning if ui has value v, uj cannot have value v’. 25
 Forward Checking
Forward checking is based on the idea that
once variable ui is assigned a value v,
then certain future variable-value pairs (uj,v’)
become impossible.

ui,v)

uj,v’ uj,v’

Instead of finding this out at many places on the tree,

we can rule it out in advance.
26
 Data Structure for Forward Checking
Future error table (FTAB)
One per level of the tree (ie. a stack of tables)

v1 v2 . . . vm
u1
u2 What does it mean if a
: whole row becomes 0?
un

At some level in the tree,

for future (unassigned) variables u
FTAB(u,v) = 1 if it is still possible to assign v to u
0 otherwise
27
How do we incorporate forward checking into a
backtracking depth-first search?

28
 Book’s Forward Checking Example

• Idea:
– Keep track of remaining legal values for unassigned variables
– Terminate search when any variable has no legal values
–

29
 Forward Checking
• Idea:
– Keep track of remaining legal values for unassigned variables
– Terminate search when any variable has no legal values
–

30
 Forward Checking
• Idea:
– Keep track of remaining legal values for unassigned variables
– Terminate search when any variable has no legal values
–

31
 Forward Checking
• Idea:
– Keep track of remaining legal values for unassigned variables
– Terminate search when any variable has no legal values
–

32
 Constraint Propagation
• Forward checking propagates information from assigned
to unassigned variables, but doesn't provide early
detection for all failures:
•

• NT and SA cannot both be blue!

•
33
• Constraint propagation repeatedly enforces constraints
 Arc Consistency
• Simplest form of propagation makes each arc consistent
• X Y is consistent iff
•
for every value x of X there is some allowed value y of Y

34
 Arc Consistency
• Simplest form of propagation makes each arc consistent
• X Y is consistent iff
•
for every value x of X there is some allowed value y of Y

35
 Arc Consistency
• Simplest form of propagation makes each arc consistent
• X Y is consistent iff
•
for every value x of X there is some allowed value y
of Y

• If X loses a value, neighbors of X need to be rechecked

•
36
 Arc consistency
• Simplest form of propagation makes each arc consistent
• X Y is consistent iff
•
for every value x of X there is some allowed value y
of Y

• If X loses a value, neighbors of X need to be rechecked

• Arc consistency detects failure earlier than forward
checking
• 37
 Arc Consistency Algorithm AC-3
Sometimes called Discrete Relaxation

• Time complexity: O(n2d3)

• 38
 Putting It All Together
• backtracking tree search
• with forward checking
• add arc-consistency
– For each pair of future variables (ui,uj)
– Check each possible remaining value v of ui
– Is there a compatible value w of uj?
– If not, remove v from possible values for ui
(set FTAB(ui,v) to 0)
39
 Comparison of Methods

• Backtracking tree search is a blind search.

• Forward checking checks constraints between the

current variable and all future ones.

• Arc consistency then checks constraints between

all pairs of future (unassigned) variables.

• What is the complexity of a backtracking tree search?

• How do forward checking and arc consistency affect that?

40
 k-consistency
(from Haralick and Shapiro, 1979,
The Consistent Labeling Problem: Part I)
Variables: U = {u1, u2, … , un}
Values: V = {v1, v2, … , vm}
Constraint Relation: R = {(u1,v1,u2,v2, … uk,vk) |
u1 having value v1, u2 having value v2,…
uk having value vk are mutually compatible}

hyperarc

41
 k-consistency

The kp discrete relaxation operator tried to

extend k-tuples of consistent variables and values
to (k+p)-tuples of consistent variables and values
in order to end up with a complete labeling consistent
over all n variables and their values.

It did a great job of pruning the search, but it was

very expensive to run.

42
 Local Search for CSPs
• Hill-climbing, simulated annealing typically work with
"complete" states, i.e., all variables assigned
•

• To apply to CSPs:
•
– allow states with unsatisfied constraints
–
– operators reassign variable values
–

• Variable selection: randomly select any conflicted

variable
•
43
 Example: 4-Queens
• States: 4 queens in 4 columns (44 = 256 states)
•
• Actions: move queen in column
•
• Goal test: no attacks
•
• Evaluation: h(n) = number of attacks
•

44
• Given random initial state, can solve n-queens in almost
 Summary
• CSPs are a special kind of problem:
•
– states defined by values of a fixed set of variables
–
– goal test defined by constraints on variable values
–

• Backtracking = depth-first search with one variable assigned per

node
•

• Variable ordering and value selection heuristics help significantly

•

• Forward checking prevents assignments that guarantee later failure

•

• Constraint propagation (e.g., arc consistency) does additional work45

to constrain values and detect inconsistencies