Introduction to Artificial Intelligence

(Comp551)
1

JIMMA UNIVERSITY
JIMMA INSTITUTE OF TECHNOLOGY
DEPARTMENT OF COMPUTING

CHAPTER FIVE
CONSTRAINT SATISFACTION PROBLEMS
 Topics we will cover
2

Constraint Satisfaction Problems (CSPs)

Backtracking Search for CSPs
Local Search for CSPs
 Constraint Satisfaction Problems (CSPs)
3

 In Standard search problem:

 State is any data structure that supports:
 Successor function,
 Heuristic function, and
 Goal test.
 CSPs are a special kind of problem:
 States defined by values of a fixed set of variables.
 Goal test defined by set of constraints on variable values.
 In Constraint satisfaction problems(CSPs):
 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.
 Allows useful general-purpose algorithms with more
power than standard search algorithms.
 Example: Map-Coloring (1)
4
 As an example we will consider the map-colouring problem.
 No two adjacent regions have the same colour.

Figure: A map showing the regions of Australia

 Variables: WA, NT, Q, NSW, V, SA, T

 Domains: D = {red,green,blue}
i
 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)}
 Example: Map-Coloring (2)
5
 Here we introduce two pieces of important terminology:
 Complete: a complete state is one in which all variables have
been assigned a value.
 Consistent: a consistent state is one that does not violate any
of the specified constraints.

 Solutions are complete and consistent assignments, e.g.,

WA=red, NT=green, Q=red, NSW=green, V=red, SA=blue,
T=green
 Example: Map Coloring
6
 Color a map with three colors so that adjacent countries have different
colors ({red, green, blue}.
 Assume that initially region B is colored with Red and Region C with
Blue.
A C variables:
G A, B, C, D, E, F, G
B

?
D
constraints: “no
neighboring regions
?

F
E have the same
?

color”
?

?
 Exercise: Coloring Problem
7

Use at most three colors (red, green and blue) to

color the given map?
 Constraint Graph
8

 Another way of visualising CSPs is by using a

constraint graph.
 Constraint graph: nodes are variables, arcs are constraints.
 Binary CSP: each constraint relates(links) only two variables.

Figure - The map-coloring problem represented as a constraint graph.

 Varieties of CSPs
9
 It is possible to categorise CSPs based on the type of variable
they use and on the size of their domains:
1. Discrete variables
 In a discrete CSP, the variables take on discrete values.
 There are two types of discrete CSP:
 Finite domains: means can take on a limited number of
discrete values.
 For example, the {red,green,blue} domain
 Infinite domains: Some discrete variables can have infinite
domains.
 integers, natural numbers,strings, etc.

2. Continuous variables
 In a continuous CSP, variables can take one from a continuous
range of values.
 For example, real numbers are continuous.
 e.g., start/end times for Hubble Space Telescope observations
 Varieties of CSPs
10

 We can also categorise CSPs based on the type of

their constraints.
 There are three types of constraint we can have:
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
 The more variables that are involved in a constraint, the
harder the problem is to solve.
 Problems with higher-order constraints are the hardest
class of problems.
 Standard Search Formulations
11

In fact, there are two ways of formulating any

search problem:
 Incremental formulation:
 Start with empty state, iteratively assign values.
 Backtracking = depth-first search with one variable assigned per node.
 Complete-state formulation:
 Start with complete state, iteratively change values.
 Local search with min-conflicts is usually effective in practice.

Search Techniques for CSPs

 The search technique that we use for CSPs depends on
what formulation we use:
 Incremental formulation → Backtracking search
 Complete-state formulation → Local search
 Backtracking Search
12

 Backtracking Search:
 The most common technique for solving incremental
formulations of CSPs.
 Basically a depth-first search with a single variable assigned at
each level of the tree.
 If no legal assignments are possible then we backtrack to try
other branches of the tree.
 Variable assignments are commutative, i.e., Commutative
simply means that the order of operations does not
matter.
[ WA = red followed by NT = green ] is the same as
[ NT = green followed by WA = red ]
 Only need to consider assignments to a single variable at
each node
 b = d and there are dn leaves
(b=branching factor, d=size of domain, n = number of variables)
Backtracking example (1)
13
Backtracking example (2)
14
Backtracking example (3)
15
Backtracking example (4)
16
Improving Efficiency of Backtracking Search
17

In the description of backtracking search, we did not

address the following questions:
 Which variable to pick at each level?

 In what order should we consider child nodes?

 What are implications of current assignment
for other unassigned variables?
It turns out that:
 These factors all have a big influence on the efficiency of
backtracking search, and
 By making the right choices of variable/value we can make
dramatic improvements in efficiency.
 Rules for Improving Efficiency
18

The rules can be summarised as follows:

 MRV (Minimum Remaining Values)
 Most constraining variable
 Least constraining value

MRV (Minimum Remaining Values)

 Choose the variable with the fewest remaining legal
values.
 Most Constraining Variable
19

Tie-breaker among most constrained variables.

Most constraining variable:
 Choose the variable with the most constraints on
remaining variables.
 Least Constraining Value
20

Given a variable, choose the least constraining

value:
 The one that rules out the fewest values in the
remaining unassigned variables.
 Forward Checking
21

Another way to improve the efficiency of

backtracking search is to use a technique
known as forward checking.
With forward checking,
 Each time a variable X is assigned,
 We check each variable Y that is linked by a
constraint to X, and
 Delete from Y’s domain any value that is
inconsistent with the value chosen for X.
 Forward Checking
22

 For example, Table below we see the progress of the map-

colouring search for Australia using forward checking.

Table: The forward checking algorithm applied to the Australia map-colouring problem
 Finally we reach a state where we can see that there are no legal values
for SA, which would cause the algorithm to backtrack immediately.
 Local search for CSPs
23

 For complete-state formulation we start with complete

states , i.e. all variables assigned.
 Local search techniques typically work with complete states.

 To apply to CSPs:
 Allow states with unsatisfied constraints – start with random state.
 Operators reassign variable values.

 Variable selection: randomly select any conflicted

variable.
 Value selection by min-conflicts heuristic:
 Choose value that violates the fewest constraints.
 i.e., hill-climb with h(n) = total number of violated constraints.
 Local Search Example: 4-Queens
24

 States: 4 queens in 4 columns (44 = 256 states)

 Actions: move queen in column
 Goal test: no attacks
 Evaluation: h(n) = number of attacks

 Given random initial state, can solve n-queens in almost

constant time for arbitrary n with high probability
 8-Queens Example
1 2
• Use min-conflicts
2 2
heuristic and
2 2
rearrange the
1 1 queens such that
3 3 they are not
2 2 attacking each
2 3 other
1 1
1 1
3 2
2 3
3 3
1 0
2 2
1 1
2 1
 Summary
26
 CSPs are a special kind of problem:
 States defined by values of a fixed set of variables.
 Goal test defined by set of constraints on variable values.

 Incremental formulation:
 Start with empty state, iteratively assign values.
 Backtracking = depth-first search with one variable assigned per node.
 Heuristics help significantly:
 MRV (Minimum Remaining Values)
 Most constraining variable
 Least constraining value

 Complete-state formulation
 Start with complete state, iteratively change values.
 Local search with min-conflicts is usually effective in practice.