CS 4700:

Foundations of Artificial Intelligence

Bart Selman
[email protected]

Module:
Constraint Satisfaction
Chapter 6, R&N
(Completes part II – Problem Solving)

1
Bart Selman
CS4700
 Outline

Constraint Satisfaction Problems (CSP)

Backtracking search for CSPs

2
Bart Selman
CS4700
Key issue: So far, we have treated nodes in
search trees as “black boxes,” only looked
inside to check whether the node is a goal state.

In CSPs, we want to “look inside the nodes”

and exploit problem structure during the
search. Sometimes, reasoning or inference
(“propagation techniques”) will
led us find solutions without any search!

3
Bart Selman
CS4700
Motivational Example:
8-Queens

Usual goal:
place 8 non-attacking
queens.

Already shown
effectiveness of local
search.

4
Bart Selman
CS4700
 Generic (DFS/BFS) search

Is this how you would

program this problem?

Node is a (partial) state of the board

Action: “place a queen on board”
Goal test: are there 8 non-attacking queens?
How many placements of 8 queens will be considered?

The DFS/BFS tree will enumerate up to

648 combinations (assume limit depth to 8).

64^8 = 2^48 = 2.8 x 10^14

Note redundancy: Q1 in (1,3), Q2 in (2,7) …
vs. Q1 in (2,7), Q2 in (1,3) … 5
Bart Selman
CS4700
 Alternative: Use “factored representation.”
Factoring refers to “splitting things into smaller parts.”

State has internal structure.

1) Here, a set of 8 variables.
x_1, x_2, … x_8.
2) Each with domain {1,…8}, the
possible variable values.
3) A set of constraints:
e.g. no two vars can be assigned
the same value. (Why?)
Each variable gives the position
of a queen in a row.
Number of board states to explore:
“only” 88 = 16.7 x 10^6 combinations.

Set of vars, set of possible values for each vars

& set of constraints defines a CSP.
6
Bart Selman
CS4700
Set of vars, set of possible values for each vars & set of constraints
defines a CSP.

A solution to the CSP is an assignment of values to the variables so that

all constraints are satisfied (no “violated constraints.”)

A CSP is inconsistent if no such solution exists.

Eg try to place 9 non-attacking queens on an 8x8 board.

Hmm. Can a search program figure out that you can’t place 101 queens
on 100x100 board?
Not so easy! Most search approaches can’t. Need much more clever
reasoning, instead of just search. (Need to use Pigeon Hole principle.)
Aside: Factored representation does not even allow one to ask
the question. Knowledge is build in.)
Alternative question: With N queens is there a solution with queen
in bottom right hand corner on a N x N board?

7
Bart Selman
CS4700
 How do we search for a solution?
Start with empty variable assignment (no vars assigned). Then, build
up partial assignments until all vars assigned.

Action: “assign a variable a value.”

Goal test: “all vars assigned and no constraint violation.”

What is the search space? (n vars, each with d possible values)

Top level branching: n . d

Hmm. But “only” n^d distinct value
Next branching: (n-1) . d
assignments! Different var ordering can lead to
Next branching: (n-2) . d the same assignment! Wasteful…
… Just “fix” a variable ordering:
Bottom level: d Backtrack search.
Check only n^d full var-value assignments.
(one var remains to be set)

So, tree with n! d^n leaves.

8
Bart Selman
CS4700
 Backtrack search
Aside: “legal” and “feasible”
Already assumes a bit of “reasoning.” (Next.)

There are many improvements on intuitive idea…

Intuitive:
1) Fix some ordering on the variables. Eg x1, x2, x3 …
2) Fix some ordering on possible values. Eg 1, 2, 3, …
3) Assign first variable, first (legal) value.
4) Assign next variable, its first (legal) value.
5) Etc.
6) Until no remaining (feasible) value exist for variable x_i,
backtrack to previous var setting of x_(i-1), try next possible
setting. If none exists, move back another level. Etc.
Visually, very intuitive on the N-Queens board (“the obvious strategy”).
See figure 6.5 book. Recursive implementation. Practice: Iterative with
stack.
9
Bart Selman
CS4700
 Reasoning, inference or “propagation.”

Message:
CSP propagation techniques can dramatically reduce search.
Sometimes to no search at all! Eg. Sudoku puzzles.

After placing the first queen, what would you do for the 2nd?

10
Bart Selman
CS4700
 General Search vs.
Constraint satisfaction problems (CSPs)
Standard search problem:
– state is a "black box“ – can be accessed only in limited way:
successor function; heuristic function; and goal test.

What is needed for CSP:

Not just a successor function and goal test. Also a means of
propagating the constraints (e.g. imposed by one queen on
the others and an early failure test).

 Explicit representation of constraints and constraint

manipulation algorithms
 Constraint Satisfaction Problems (CSP)

11
Bart Selman
CS4700
 Constraint satisfaction problems (CSPs)

States and goal test have a standard representation.

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
Interesting tradeoff:

Example of a (restricted) formal representation language.

Allows useful general-purpose algorithms more powerful than standard search

algorithms that have to resort to problem specific heuristics
to enable solution of large problems.

12
Bart Selman
CS4700
 Constraint Satisfaction Problem

Set of variables {X1, X2, …, Xn}

Each variable Xi has a domain Di of possible values
Usually Di is discrete and finite
Set of constraints {C1, C2, …, Cp}
Each constraint Ck involves a subset of variables and
specifies the allowable combinations of values of
these variables
Goal:
Assign a value to every variable such
that all constraints are satisfied

13
Bart Selman
CS4700
 Motivational Example:
8-Queens
l + k = 10 l–k=1
l – k =0

How do we represent 8-Queens

… as a CSP:
l – k= -5 Variables?
l=3
Constraints?
l=2
Note: More than one option.
l=1
k=1 k=3
k=2 …

14
Bart Selman
CS4700
 Example: 8-Queens Problem

Xi – column for queen in row i

8 variables Xi, i = 1 to 8 (one per row)
Domain for each variable {1,2,…,8}
Constraints are of the form:
– Xi  Xj when ji (i.e. no two in the same column)
– No queens in same diagonal:
1) Xi – Xj  i – j
2) Xi – Xj  j – i
(check that this works!)

Alternative? Boolean vars

15
Bart Selman
CS4700
 Boolean Encoding
64 variables Xij, i = 1 to 8, j = 1 to 8
Domain for each variable {0,1} (or {False, True})
Xij = 1 iff “there is a
Constraints are of the form:
Row and columns
queen on location (i,j).”
– If (Xij = 1) then (Xik = 0) for all k = 1 to 8, kj (logical constraint)
– Xij = 1  Xkj = 0 for all k = 1 to 8, ki
Diagonals
– Xij = 1  Xi+l,j+l = 0 l = 1 to 7, i+l 8; j+l8 (right and up)
– Xij = 1  Xi-l,j+l = 0 l = 1 to 7, i-l 1; j+l8 (right and down)
– Xij = 1  Xi-l,j-l = 0 l = 1 to 7, i-l 1; j-l1 (left and down)
– Xij = 1  Xi+l,j-l = 0 l = 1 to 7, i+l 8; j-l1 (left and up)

What’s missing? Need N (= 8) queens on board!

3 options:
1) Maximize sum X_ij (optimization formulation)
2) Sum X_ij = N (CSP; bit cumbersome in Boolean logic)
3) For each row i: (X_i1 OR X_i2 OR X_i3 … X_iN)
16
Bart Selman
CS4700
 Logical equivalence

Two sentences p an q are logically equivalent ( or ) iff p  q is a tautology

(and therefore p and q have the same truth value for all truth assignments)

 

  

17
Bart Selman
CS4700
 Propositional Satisfiability problem

Satifiability (SAT): Given a formula in propositional calculus, is there a model

(i.e., a satisfying interpretation, an assignment to its variables) making it true?

We consider clausal form, e.g.:

( a  b   c ) AND ( b   c) AND ( a  c)

n possible assignments
2

SAT: prototypical hard combinatorial search and reasoning

problem. Problem is NP-Complete. (Cook 1971)
Surprising “power” of SAT for encoding computational problems.
18
Bart Selman
CS4700
 Significant progress in
Satisfiability Methods

Software and hardware verification – complete Many Applications:

methods are critical - e.g. for verifying the Hardware and
correctness of chip design, using SAT Software Verification
encodings Planning,
Protocol Design,

Going from 50 variable, 200 constraints

Scheduling, Materials
to 1,000,000 variables and 5,000,000 constraints Discovery etc.
in the last 10 years

Current methods can verify automatically the

correctness of large portions of a chip

19
19
Bart Selman
CS4700
 Turing Award: Model Checking

Source: Slashdot CS4700 20

Bart Selman
A “real world” example

21
21
Bart Selman
CS4700
Bounded Model Checking instance:

i.e. ((not x1) or x7)

and ((not x1) or x6)
and … etc.

22
Bart Selman
CS4700
File format
Dimacs Format for CNF
The benchmark file format will be in a simplified version of the DIMACS format: c
c start with comments
c
c
p cnf 5 3
1 -5 4 0
-1 5 3 4 0
-3 -4 0
The file can start with comments, that is lines begining with the character c.
Right after the comments, there is the line p cnf nbvar nbclauses indicating that the
instance is in CNF format; nbvar is the exact number of variables appearing in the
file; nbclauses is the exact number of clauses contained in the file.
Then the clauses follow. Each clause is a sequence of distinct non-null numbers
between -nbvar and nbvar ending with 0 on the same line; it cannot contain the
opposite literals i and -i simultaneously. Positive numbers denote the corresponding
variables. Negative numbers denote the negations of the corresponding variables.

23
Bart Selman
CS4700
 Example of Sat Solver

SAT Solver : Lingeling

http://fmv.jku.at/lingeling/

Nqueen4-v1.cnf

Nquuens4-v2.cnf

24
Bart Selman
CS4700
p cnf 16 84
1234 0
-1 -2 0 1 5 9 13 0
-1 -6 0
-1 -5 0
-1 -3 0 -1 -11 0
-1 -9 0 -1 -16 0
-1 -4 0 -1 -13 0
-2 -3 0 -6 -11 0
-5 -9 0
-6 -16 0
-2 -4 0 -5 -13 0
-11 -16 0
-3 -4 0 -9 -13 0 -2 -7 0
5678 0 2 6 10 14 0
-2 -12 0
-5 -6 0
-2 -6 0
-2 -10 0
-7 -12 0 1 2 3 4
-3 -8 0
-5 -7 0
-5 -8 0
-2 -14 0
-6 -10 0
-5 -10 0 5 6 7 8
-5 -15 0
-6 -7 0
-6 -8 0
-6 -14 0
-10 -14 0
-10 -15 0 9 10 11 12
-9 -14 0
-7 -8 0 3 7 11 15 0
-3 -7 0
-4 -7 0 13 14 15 16
9 10 11 12 0 -4 -10 0
-3 -11 0 -4 -13 0
-9 -10 0 -3 -15 0
-9 -11 0 -7 -10 0
-7 -11 0
-7 -13 0
-9 -12 0 -7 -15 0
-10 -13 0
-10 -11 0 -11 -15 0 -3 -6 0
-10 -12 0 4 8 12 16 0
-3 -9 0
-4 -8 0
-11 -12 0 -6 -9 0
-4 -12 0 -2 -5 0
13 14 15 16 0 -4 -16 0
-13 -14 0 -8 -11 0
-8 -12 0 -8 -14 0
-13 -15 0 -8 -16 0 -11 -14 0
-13 -16 0 -12 -16 0 -12 -15 0
-14 -15 0
-14 -16 0
-15 -16 0
25
Bart Selman
CS4700
 How Large are the Problems?
A bounded model checking problem:

Source: IBM

26
CS4700 26
Bart Selman
SAT Encoding
(automatically generated from problem specification)

i.e., ((not x1) or x7)

((not x1) or x6)
etc.
x1, x2, x3, etc. are our Boolean variables
(to be set to True or False)

Should x1 be set to False??

27
CS4700 27
Bart Selman
10 Pages Later:

…
i.e., (x177 or x169 or x161 or x153 …
x33 or x25 or x17 or x9 or x1 or (not x185))

clauses / constraints are getting more interesting…

Note x1 …

28
CS4700 28
Bart Selman
4,000 Pages Later:

29
CS4700 29
Bart Selman
Finally, 15,000 Pages Later:

Search space of truth assignments:

Current SAT solvers solve this instance in

under 10 seconds!

30
CS4700 30
Bart Selman
Example of a Boolean Satisfiability (SAT) encoding.
Very “simple” but effective representation.
Example of a logical knowledge representation language.

For propositional logic, see R&N 7.4.1 & 7.4.2.

31
Bart Selman
CS4700
Which encoding is better?Allows for faster solutions?

One would think, fewer variables is better…

Search spaces:

88 = 1.6 x 106 vs 264 = 1.8 x 1019

However, in practice SAT encodings can be surprisingly

effective, even with millions of Boolean variables. Often,
few true local minima in search space.

Demo of backtrack search and local search on Boolean

encoding of N-queens.

32
Bart Selman
CS4700
 N-Queens

The standard N by N Queen's problem asks how to place N

queens on an ordinary chess board so that they don’t attack
each other
7
6
5
4
N=8 3
2
1
0
Is this problem 0 1 2 3 4 5 6 7
NP-Complete? 33 33
Bart Selman
CS4700
 N-Queens
N=8 (another solution)

7
6
5
4
3
2
1
0
0 1 2 3 4 5 6 7

34 34
Bart Selman
CS4700
 Linear congruence equations

N = 6 ;
N = 6  + 1;
N = 6  ± 2; (not N=4;)
N = 6  + 3; (not N=9;)
N = 6  - 2; (inc. N=4)
N = 12  - 3; (inc. N=9)

35 35
Bart Selman
CS4700
N-Queens
N=8
N = 6 2, =1

S(8) = {3}.

For all c  S(8), the linear congruence equations

Let’s consider c=3.

6 x + y  3 (mod 8),
where x = 0, 1, 2, 3.
x = 0 y=3; x=1 y = 5; x=2 y= 7;x=3 y = 1;

6 x + y  6 (mod 8),
where x = 4,5,6,7,8
X=4, y = 6; x=5 y = 0; x=6 y = 2; x=7 y=4;

36 36
Bart Selman
CS4700
 Partially Filled Nqueens

So the N-queens problem is easy when we start with an

empty board.

What about if we pre-assign some quuens and ask for a

completion?

Open question

Conjecture: completing a partially filled N-

queens board is NP-complete.

37
Bart Selman
CS4700
 Example: Crypt-arithmetic Puzzle

SEND
+ MOR E
----------
MONEY

Variables: S, E, N, D, M, O, R, Y Soln.:95
67
Domains: 1085
[0..9] for S, M, E, N, D, O, R, Y
====
Search space: 1,814,400 10652
Aside: could have [1..9] for S and M

38
Bart Selman
CS4700
 Constraints
Option 1:
C1a) 1000 S + 100 E + 10 N + D +
1000 M + 100 O + 10 R + E
= 10000 M + 1000 O + 100 N + 10 E + Y SEND
Or use 5 equality constraints, using auxiliary +MORE
“carry” variables C1, …, C4 Є [0…9] ----------
Option 2: C1b)
MONEY
D+E = 10 C1 + Y Which constraint set better
C1 + N + R = 10 C2 + E for solving? C1a or C1b? Why?
C2 + E + O = 10 C3 + N
C3 + S + M = 10 C4 + O C1b, more “factored”. Smaller
C4 =M pieces. Gives more propagation!

Need two more sets of constraints:

C2) S =/= 0, M =/= 0
C3) S =/= M, S =/= O, … E =/= Y (28 not equal constraints)
Note: need to assert everything!
Alt. “All_diff(S,M,O,...Y)” for C3.
39
Bart Selman
CS4700
 Some Reflection:
Reasoning/Inference vs. Search

How do human solve this?

What is the first step?

1) M = 1, because M =/= 0 and …

the carry over of the addition of two
digits (plus previous carry) is at most 1.

Actually, a somewhat subtle

piece of mathematical background
Also, what made us focus knowledge.
on M?
Experience / intuition …

40
Bart Selman
CS4700
1) M = 1, because M =/= 0 and …
the carry over of the addition of two
digits (plus previous carry) is at most 1.

2) O = 0. Because M=1 and we have to have a carry to the next

column. S + 1 + C3 is either 10 or 11. So, O equals 0 or 1.
1 is taken. So, O = 0.

3) S = 9. There cannot be a carry to the 4th column (if

there were, N would also have to be 0 or 1. Already
taken.). So, S = 9.

A collection of “small pieces” of local reasoning,

using basic constraints from the rules of arithmetic.
A logic (SAT) based encoding will likely “get these steps.”
41
Bart Selman
CS4700
 And further it goes…

Largely, a clever chain of reasoning / inference /

propagation steps (no search) except for…
exploring 2 remaining options (i.e., search) to
find complete solution.
42
Bart Selman
CS4700
Human problem solving nicely captures key idea behind how to solve
CSPs: replace as much of search by propagation
(inference/reasoning).

One difficulty: Humans often use subtle background knowledge.

Can search be completely avoided?

A.: Most likely NO. General cryptarithmetic is NP-complete

(Eppstein 1987)

43
Bart Selman
CS4700
 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)}

44
Bart Selman
CS4700
 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

(Aside: Four colors suffice. (Appel and Haken 1977)

45
Bart Selman
CS4700
 Constraint graph:
Graph Coloring
Binary CSP: each constraint relates two variables

Constraint graph: nodes are variables, arcs are constraints

Two variables are adjacent or neighbors if they

are connected by an edge or an arc

46
Bart Selman
CS4700
 Application of Graph Coloring

Lots of applications involving scheduling and assignments.

Scheduling of final exams – nodes represent finals, edges between finals

denote that both finals have common students (and therefore they have to
have different colors, or different periods).
Time Period  courses
1 1
7 2 7 2 I (red)  1,6
II (blue)2
III (green)3,5
6 3 6 3 IV (black)

5 4 5 4
Graph of finals for 7 courses CS4700 47
Bart Selman
 Varieties of CSPs
Discrete variables

– finite domains: our focus

• n variables, domain size d  O(dn) complete assignments
(includes Boolean satisfiability 1st problem to be shown NP-complete.)

– infinite domains:
• integers, strings, etc.
• e.g., job scheduling, variables are start/end days for each job
• need a constraint language, e.g., StartJob1 + 5 ≤ StartJob3

Continuous variables

– e.g., start/end times for Hubble Space Telescope observations

– linear constraints solvable in polynomial time by linear
programming Bart Selman
CS4700 48
 Varieties of constraints

Unary constraints involve a single variable,

e.g., SA ≠ green
–

Binary constraints involve pairs of variables,

e.g., SA ≠ WA
–

Higher-order constraints involve 3 or more variables,

e.g., cryptarithmetic column constraints
–

49
Bart Selman
CS4700
 Remark

Finite CSP include 3SAT as a special case (under logical reasoning).

3SAT is known to be NP-complete.

So, in the worst-case, we cannot expect to solve a finite CSP in less

than exponential time.

51
Bart Selman
CS4700
 Solving CSP by search : Backtrack Search
BFS vs. DFS
– BFS  not a good idea.
• A tree with n!dn leaves : (nd)*((n-1)d)*((n-2)d)*…*(1d) = n!dn
• Reduction by commutativity of CSP
– A solution is not in the permutations but in combinations.
– A tree with dn leaves
– DFS
• Used popularly
– Every solution must be a complete assignment and therefore
appears at depth n if there are n variables
– The search tree extends only to depth n.
• A variant of DFS: Backtrack search
– Chooses values for one variable at a time
– Backtracks when failed even before reaching a leaf.
• Better than BFS due to backtracking but still need more
“cleverness” (reasoning/propagation).

52
Bart Selman
CS4700
  Backtrack Search

empty assignment

1st variable

2nd variable

3rd variable

Assignment = {}
54
Bart Selman
CS4700
  Backtrack Search

empty assignment

1st variable

2nd variable

3rd variable

Assignment = {(var1=v11)}
55
Bart Selman
CS4700
  Backtrack Search

empty assignment

1st variable

2nd variable

3rd variable

Assignment = {(var1=v11),(var2=v21)}
56
Bart Selman
CS4700
  Backtrack Search

empty assignment

1st variable

2nd variable

3rd variable

Assignment = {(var1=v11),(var2=v21),(var3=v31)}
57
Bart Selman
CS4700
  Backtrack Search

empty assignment

1st variable

2nd variable

3rd variable

Assignment = {(var1=v11),(var2=v21),(var3=v32)}
58
Bart Selman
CS4700
  Backtrack Search

empty assignment

1st variable

2nd variable

3rd variable

Assignment = {(var1=v11),(var2=v22)}
59
Bart Selman
CS4700
  Backtrack Search

empty assignment

1st variable

2nd variable

3rd variable

Assignment = {(var1=v11),(var2=v22),(var3=v31)}
60
Bart Selman
CS4700
 Improving Backtracking Efficiency
Variable & value ordering to increase
 Which variable should be assigned next? the likelihood to success
 Minimum Remaining Values heuristic
 In what order should its values be tried?
 Least Constraining Values heuristic
 Can we detect inevitable failure early?
 Forward checking
 Constraint propagation (Arc Consistency)
 When a path fails, can the search avoid repeating
this failure?
 Backjumping Early failure-detection to decrease
the likelihood to fail
 Can we take advantage of problem structure?
 Tree-structured CSP Restructuring to reduce the
problem’s complexity

General purpose techniques

 Improving backtracking efficiency

function BACKTRACKING-SEARCH (csp) returns a solution, or failure

return RECURSIVE-BACKTRACKING({}, csp)

function RECURSIVE-BACKTRACKING(assignment, csp) returns a solution, or failure

if assignment is complete then return assignment
var  SELECT-UNASSIGNED-VARIABLE(VARIABLES[csp], assignment, csp)
for each value in ORDER-DOMAIN-VALUES(var, assignment, csp) do
if value is consistent with assignment according to CONSTRAINTS[csp] then
add {var=value} to assignment
result  RECURSIVE-BACKTRACKING(assignment, csp)
if result != failure then return result
remove {var = value} from assignment
return failure
 Choice of Variable
#1: Minimum Remaining Values (aka Most-constrained-variable
heuristic):

Select a variable with the fewest remaining values

64
Bart Selman
CS4700
 Choice of Variable, cont.

Tie-breaker among most constrained variables

#2 Most constraining variable:

choose the variable with the most constraints on remaining variables

–

65
Bart Selman
CS4700
 Choice of Value:
Least constraining value

#3 Given a variable, choose the least constraining value:

the one that rules out the fewest values in the remaining variables
–

66
Bart Selman
CS4700
 Constraint Propagation

The process of determining how the possible values of one

variable affect the possible values of other variables

67
Bart Selman
CS4700
 Forward Checking
After a variable X is assigned a value v, look at each unassigned
variable Y that is connected to X by a constraint and deletes
from Y’s domain any value that is inconsistent with v

Terminate branch when any variable has no

legal values & backtrack.

68
Bart Selman
CS4700
Forward checking

69
Bart Selman
CS4700
 70
Bart Selman
CS4700
 71
Bart Selman
CS4700
 72
Bart Selman
CS4700
 Constraint propagation
NT
Q
WA
SA
NSW
V
Forward checking propagates information from assigned to unassigned
variables, but doesn't provide early detection for all failures:
What’s the problem here?
NT Q
WA
SA
NSW
V
T

≠
NT and SA cannot both be blue!
Use: constraint propagation repeatedly to enforce constraints locally.
73 Bart Selman
CS4700
 Definition (Arc consistency)

A constraint C_xy is said to be arc consistent w.r.t. x iff

for each value v of x there is an allowed value of y.

Similarly, we define that C_xy is arc consistent w.r.t. y.

A binary CSP is arc consistent iff every constraint C_xy is arc

consistent wrt x as well as wrt y.

74
Bart Selman
CS4700
When a CSP is not arc consistent, we can make it arc
consistent.

This is also called “enforcing arc consistency”.

75
Bart Selman
CS4700
 Example
Let domains be
D_x = {1, 2, 3}, D_y = {3, 4, 5, 6}
One constraint
C_xy = {(1,3), (1,5), (3,3), (3,6)} [“allowed value pairs”]

C_xy is not arc consistent w.r.t. x, neither w.r.t. y. Why?

To enforce arc consistency, we filter the domains,

removing inconsistent values.

D’_x = {1, 3}, D’_y={3, 5, 6}

76
Bart Selman
CS4700
 Arc consistency

Simplest form of propagation makes each arc consistent. NT

I.e., X Y is consistent iff Q
WA
for every value x of X there is some allowed y SA
NSW
V

77
Bart Selman
CS4700
 NT
Q
WA
SA
NSW
V

78
Bart Selman
CS4700
 NT
Q
WA
SA
NSW
V

79
Bart Selman
CS4700
If X loses a value, neighbors of X need to be rechecked.
Arc consistency detects failure earlier than forward checking.

Can be run as a preprocessor or after each assignment.

(takes polytime each time)

NT
Q
WA
SA
NSW
V

Empty domain detected! Backtrack early.

80
Bart Selman
CS4700
 Beyond Arc Consistency
{1, 2} {1, 2}
XY
X Y
Is this network arc consistent?

What is the solution? XZ YZ

Clearly arc consistency is not enough Z

to guarantee global consistency. {1, 2}
There are other forms of consistency,
such as k-consistency.

But when k = n (num vars), we are

looking at the original problem!

82
Bart Selman
CS4700
 k - Consistency
{1, 2} {1, 2}
XY
X Y
A graph is K-consistent iff the following is true:
XZ YZ
Choose values of any K-1 variables that satisfy all
the constraints among these variables and choose
any K th variable. Then, there exists a value for Z
this K th variable that satisfies all the constraints
{1, 2}
among these K variables.

A graph is strongly K-consistent if it is J-

consistsent for all J<=K.

What type of consistency would we need here to solve any constraint

problem without search?
K=N
83
Bart Selman
CS4700
 Consistency

Node consistency = strong 1- consistency

Arc consistency = strong 2- consistency
(note: arc-consistency is usually
assumed to include node-consistency as well).
See R&N sect. 6.2.3 for “path-consistency” = 3-consistency for binary CSPs.

Algorithms exist for making a constraint graph strongly K-consistent for K>2
but in practice they are rarely used because of efficiency issues.

Other consistency notions involve “global constraints,” spanning many

variables. E.g. AllDiff constraint can handle Pigeon Hole principle.

84
Bart Selman
CS4700
 Summary: Solving a CSP
Search:
– can find solutions, but may examine many non-
solutions along the way
Constraint Propagation:
– can rule out non-solutions, but but may not lead
to full solution.
Interweave constraint propagation and search
– Perform constraint propagation at each search step.
– Goal: Find the right balance between search
(backtracking) and propagation (reasoning).

Surprising efficiency (last 10 yrs):

100K + up to one million variable CSP problems
are now solvable!
See also local search. R&N 6.4

85
Bart Selman
CS4700