Constraint Satisfaction Problems (PowerPoint download)

							    Constraint Satisfaction Problems

               Chapter 5




1
    Outline
     Constraint Satisfaction Problems (CSP)
     Backtracking search for CSPs
     Local search for CSPs




2
    Constraint satisfaction problems (CSPs)
     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


     Allows useful general-purpose algorithms with more power than
      standard search algorithms

3
Formulation as Standard Search Problem
 States
   Set of value assignments to some or all of the variables
 Initial state
   The empty assignment {}
 Successor function
   Value assignment to any unassigned variable without conflicts with
    previously assigned variables
 Goal test
   Finding out the current assignment is complete
 Path Cost
   A constant cost
Benefits from treating a problem as a CSP

 The successor function and goal test can be written in a
  generic way that applies to all CSPs.

 We can develop effective, generic heuristics that require
  no additional domain-specific expertise.

 The structure of the constraint graph can be used to
  simplify the solution process.
Example of CSP : 8-queen problem


                  Variables
                     {Q1,Q2,Q3,Q4,Q5,Q6,Q7,Q8}

                     Domain

                       •   Q1, …, Q8 ∊ {(1,1), (1,2), …, (8,8)}
                  Constraint set
                     No queen attacks any other
Example: Street Puzzle

     1        2         3         4         5




Ni = {English, Spaniard, Japanese, Italian, Norwegian}
Ci = {Red, Green, White, Yellow, Blue}
Di = {Tea, Coffee, Milk, Fruit-juice, Water}
Ji = {Painter, Sculptor, Diplomat, Violonist, Doctor}
Ai = {Dog, Snails, Fox, Horse, Zebra}
Example: Street Puzzle
 1      2        3         4        5
Ni = {English, Spaniard, Japanese, Italian, Norwegian}
Ci = {Red, Green, White, Yellow, Blue}
Di = {Tea, Coffee, Milk, Fruit-juice, Water}
Ji = {Painter, Sculptor, Diplomat, Violonist, Doctor}
Ai = {Dog, Snails, Fox, Horse, Zebra}

 The Englishman lives in the Red house
 The Spaniard has a Dog
                                             Who owns the Zebra?
 The Japanese is a Painter                   Who drinks Water?
 The Italian drinks Tea
 The Norwegian lives in the first house on the left
 The owner of the Green house drinks Coffee
 The Green house is on the right of the White house
 The Sculptor breeds Snails
 The Diplomat lives in the Yellow house
 The owner of the middle house drinks Milk
 The Norwegian lives next door to the Blue house
 The Violonist drinks Fruit juice
 The Fox is in the house next to the Doctor’s
 The Horse is next to the Diplomat’s
Example: Task Scheduling
                          T1

                 T2                   T4

                               T3

T1   must   be done during T3
T2   must   be achieved before T1 starts
T2   must   overlap with T3
T4   must   start after T1 is complete

• Are the constraints compatible?
• Find the temporal relation between every two tasks
        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)}
 

10
       CSP : Terminologies
 Consistent (or legal) assignment
    An assignment that does not violate any
     constraints
 Complete assignment
    An assignment in which every variable is
     mentioned
 Solution
    A complete and consistent assignment
    A complete assignment that satisfies all the
     constraints
 Constraint graph
    Node : variable
    Arc : constraint
                                            Has different color
                                            VAR: WA,NT,SA,Q,NSW,V,T
       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

12
     Constraint graph
      Binary CSP: each constraint relates two variables


      Constraint graph: nodes are variables, arcs are constraints




13
     Varieties of CSPs
      Discrete variables

        finite domains:
           n variables, domain size d  O(dn) complete assignments
           e.g., Boolean CSPs, incl.~Boolean satisfiability (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



14
     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



     Preference Constraint
           ex) Prof. X might prefer teaching in the morning.
           ex) red is better than yellow
           A cost for each variable assignment
           Many real-world CSP
           Constrained optimization problems


15
        Example: Cryptarithmetic




      Variables: F T UW
       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
         X2 + T + T = O + 10 · X3
         X3 = F, T ≠ 0, F ≠ 0
16
        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 real-valued variables




17
     Standard search formulation (incremental)
     Let's start with the straightforward approach, then fix it

     States are defined by the values assigned so far

         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

     1. This is the same for all CSPs
     2. Every solution appears at depth n with n variables
         use depth-first search
     3. Path is irrelevant, so can also use complete-state formulation
     4. b = (n - l )d at depth l, hence n! · dn leaves


18
    Solving CSP by search : Backtracking Search
 BFS vs. DFS
   BFS  terrible!
     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 : Backtracking search
          Chooses values for one variable at a time
          Backtracts when failed even before reaching a leaf.
     Better than BFS due to backtracking, but still inefficient!!
     Backtracking 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
         b = d and there are dn leaves


      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



20
     Backtracking search




21
     Backtracking example




22
     Backtracking example




23
     Backtracking example




24
     Backtracking example




25
     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?




26
    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
 Can we take advantage of problem              the likelihood to fail
   structure?
    Tree-structured CSP
                                                Restructuring to reduce the
                                                problem’s complexity


           General purpose methods
Heuristics for variable & value ordering

 Variable selection
   Most constrained variables
     MRV(minimum remaining values) heuristic
     Choose the variable with the fewest legal values.
   Most containing variables
     Degree heuristic
     In first variable selection, MRV heuristic doesn’t help at all.
     Use degree heuristic (selects the variable with the largest degree.)  Tie-
      breaker !!
 Value ordering
   Least-Constraining-value heuristic
     prefer variables that rule out fewest choices of neighbors
     Most constrained variable
      Most constrained variable:
       choose the variable with the fewest legal values




      a.k.a. minimum remaining values (MRV) heuristic




29
          Heuristics for variable & value ordering
          Example for MRV


                                                             NT
                                                     NT           Q

                           SA                        SA      SA




NT ={R, G, B}        NT ={G, B}            NT =G
Q = {R, G, B}        Q = {R, G, B}         Q = {R, B}
WA = {R, G, B}       WA = R                WA = R
SA = {R, G, B}       SA = {G, B}           SA = {B}
NSW = {R, G, B}      NSW = {R, G, B}       NSW = {R, G, B}
V = {R, G, B}        V = {R, G, B}         V = {R, G, B}
T = {R, G, B}        T = {R, G, B}         T = {R, G, B}
     Most constraining variable
      Tie-breaker among most constrained variables
      Most constraining variable:
        choose the variable with the most constraints on remaining
         variables




31
             Heuristics for variable & value ordering
             Example for degree heuristic

                                                        NT       NT   Q
                              SA                        SA       SA




3   NT ={R, G, B},3       NT ={R, G}             NT = G
3   Q = {R, G, B},3       Q = {R, G}             Q = {R}
2   WA = {R, G, B},2      WA = {R, G}            WA = {R}
5   SA = {R, G, B},5      SA = B                 SA = B
3   NSW = {R, G, B},3     NSW = {R, G}           NSW = {R, G}
2   V = {R, G, B},2       V = {R, G}             V = {R, G}
0   T = {R, G, B},0       T = {R, G, B}          T = {R, G, B}
     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 queens feasible




33
             Heuristics for variable & value ordering
             Example for LCV
       Terminate search when any variable has no legal values
                                               Red in Q is least constraint value



                                                               NT
                                                          WA        Q

                                          NT
                   WA                WA
                                                               NT   Q
                                                          WA



NT ={R, G, B}      NT ={G, B}        NT =G               NT =G           NT =G
Q = {R, G, B}      Q = {R, G, B}     Q = {R, B}          Q=R             Q=B
WA = {R, G, B}     WA = R            WA = R              WA = R          WA = R
SA = {R, G, B}     SA = {G, B}       SA = {B}            SA = {B}        SA = ?
NSW = {R, G, B}    NSW = {R, G, B}   NSW = {R, G, B}     NSW = {G, B}    NSW = {R,G}
V = {R, G, B}      V = {R, G, B}     V = {R, G, B}       V = {R, G, B}   V = {R, G, B}
T = {R, G, B}      T = {R, G, B}     T = {R, G, B}       T = {R, G, B}   T = {R, G, B}
         Forwarding checking(1/4)
         Heuristics for early failure-detection

 How can we find out which variable is minimum remaining?
 What does eliminate the values from domain?



                                    NT = red, Q = green

                                           ⅹ
                                    WA = {red, green, blue}
                                          ⅹ green, blue}
                                    SA = {red, ⅹ
                                                  ⅹ
                                    NSW = {red, green, blue}
                                    V = {red, green, blue}
                                    T = {red, green, blue}
        Forwarding checking(2/4)
        Heuristics for early failure-detection

 Forward checking
   A variable X is assigned.
   Looks at each unassigned variable Y connected to X by a constraint
   Deletes any values of Y’s domain that is inconsistent with X
   If there is any Y with empty domain, undo this assigning.
       Backtrack!!
      If not, MRV !!
 Forward checking is an obvious partner of MRV heuristic !!
   MRV heuristic is activated after forward checking.
Heuristics for early failure-detection
Forwarding checking(3/4) – an example


     NT
WA        Q
     SA
          NSW
           V
          T
Heuristics for early failure-detection
Forwarding checking(4/4) - Heuristic flow

                            Forward Checking


                          Yes
                                   Tie?
                                         No

     Degree heuristic        Select the MRV


                            Forward Checking


                             Select the LCV


                                  Goal?
     Forward checking
      Idea:
         Keep track of remaining legal values for unassigned variables
         Terminate search when any variable has no legal values




39
     Forward checking
      Idea:
         Keep track of remaining legal values for unassigned variables
         Terminate search when any variable has no legal values




40
     Forward checking
      Idea:
         Keep track of remaining legal values for unassigned variables
         Terminate search when any variable has no legal values




41
     Forward checking
      Idea:
         Keep track of remaining legal values for unassigned variables
         Terminate search when any variable has no legal values




42
      Heuristics for early failure-detection
      Constraint propagation: arc consistency

 F/C propagates information from
  assigned to unassigned
  variables, but doesn’t provide
  early detection for all failures
 Simplest form of propagation
  makes each arc consistent                    ⅹ
 The (directed) arc (XY) is
  consistent                                   o
   ≡ For all x ∊ X, there is some y ∊ Y.
 If X loses a value, neighbors of X
  need to be rechecked
  Heuristics for early failure-detection
  Constraint propagation: arc consistency
 Detects an inconsistency that is not detected by pure
  F/C.
 Detect failure earlier than forward checking
 Makes CSP possibly with reduced domain.
 Provides a fast method of constraint propagation
  O(n2d3)
  n2 : the maximum number of arcs
  d : the maximum number of insertion for one node
  d2 : the number of forward checking
 Can be run as a preprocessor or after each assignment
     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!
      Constraint propagation repeatedly enforces constraints locally



45
     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 y



46
     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 y



47
     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 y
        If X loses a value, neighbors of X need to be rechecked

48
     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 y
        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



49
     Arc consistency algorithm AC-3




  Time complexity: O(n2d3)

50
           Heuristics for early failure-detection
           Special constraints – Alldiff


 Problem-dependent constraints
 Alldiff constraint
   If there are m variables involved in the
     constraint and n possible distinct
                                                WA
     values, m > n  inconsistent for Alldiff
     constraint                                      NSW
   An example
       3 variables: NT,SA,Q
       NT, SA, Q ∊   {green, blue}
       m=3 > n=2
       inconsistent
    Heuristics for early failure-detection

    Special constraints – Almost
 Resource constraint
   Atmost constraint (Use limited resource)
   An example
      Totally, no more than 10 personnel should be assigned to jobs.
      Atmost(10, PA1, PA2, PA3, PA4)
      Job1~4 ∊ {3,4,5,6} ( min value = 3 )
          3+3+3+ 3 > 10 (minimum resource 12 but available resource 10)
          inconsistent
      Job1~4 ∊ {2, 3,4,5,6}
          2+2+2+2 <10 (consistent)
          2+2+2+5(or 6) > 10(inconsistent)
           o Delete 5,6 from domain
             Heuristics for early failure-detection   - Backjumping

 The weakness of Backtracking                                           1
   If failed in SA, backtrack to T.                                 Q
      But, T is not relevant to SA.

 Backjumping                                                5
                                                                             NSW 2

   Backtracks to the most recent                                3   V
                                                  Next
    variable in the conflict set.
                                                  variable
      Conflict set : the set of variables that                      T
       caused the failure                                            4
      If failed in SA, backjump to V.
Solving a CSP

 Interweave constraint propagation, e.g.,
 • forward checking
 • AC3
 and backtracking

 + Take advantage of the CSP structure
4-Queens Problem

                        X1          X2
     1   2   3   4   {1,2,3,4}   {1,2,3,4}
 1
 2

 3

 4
                        X3          X4
                     {1,2,3,4}   {1,2,3,4}
4-Queens Problem

                        X1          X2
     1   2   3   4   {1,2,3,4}   {1,2,3,4}
 1
 2

 3

 4
                        X3          X4
                     {1,2,3,4}   {1,2,3,4}
4-Queens Problem

                        X1          X2
     1   2   3   4   {1,2,3,4}   {1,2,3,4}
 1
 2

 3

 4
                        X3          X4
                     {1,2,3,4}   {1,2,3,4}
4-Queens Problem

                        X1          X2
     1   2   3   4   {1,2,3,4}   {1,2,3,4}
 1
 2

 3

 4
                        X3          X4
                     {1,2,3,4}   {1,2,3,4}
4-Queens Problem

                        X1          X2
     1   2   3   4   {1,2,3,4}   {1,2,3,4}
 1
 2

 3

 4
                        X3          X4
                     {1,2,3,4}   {1,2,3,4}
4-Queens Problem

                        X1          X2
     1   2   3   4   {1,2,3,4}   {1,2,3,4}
 1
 2

 3

 4
                        X3          X4
                     {1,2,3,4}   {1,2,3,4}
4-Queens Problem

                        X1          X2
     1   2   3   4   {1,2,3,4}   {1,2,3,4}
 1
 2

 3

 4
                        X3          X4
                     {1,2,3,4}   {1,2,3,4}
4-Queens Problem

                        X1          X2
     1   2   3   4   {1,2,3,4}   {1,2,3,4}
 1
 2

 3

 4
                        X3          X4
                     {1,2,3,4}   {1,2,3,4}
4-Queens Problem

                        X1          X2
     1   2   3   4   {1,2,3,4}   {1,2,3,4}
 1
 2

 3

 4
                        X3          X4
                     {1,2,3,4}   {1,2,3,4}
4-Queens Problem

                        X1          X2
     1   2   3   4   {1,2,3,4}   {1,2,3,4}
 1
 2

 3

 4
                        X3          X4
                     {1,2,3,4}   {1,2,3,4}
      General Heuristic for complexity reduction

 Decompose into many sub-problems.
   Independent sub-problems
      Solve them respectively
      Ex) The mainland and Tasmania.
   Analysis of connected components of constraint graph
           Algorithm for Tree-structured CSP
 If constraint graph is a tree, we solve the
   CSP in time linear in the number of variables
1. Choose a variable as root
2. Order variables from root to leaves such
   that every node’s parent precedes it in the
   ordering Makes constraint graph arc-
   consistent
3. For j from n down to 2, apply arc
   consistency to the arc(parent(Xj), Xj)
   Just finds a consistent value.(no backtrack
    because of arc-consistency)
4. For j from 1 to n, assign Xj consistently with
   Parent(Xj)
 Time complexity : O(nd2)
     Heuristics for complexity reduction

     Transforming to Tree-Structured CSP
 Conditioning
   Instantiate a variable, prune its neighbor’s domains
 Cutset conditioning
   Instantiate a set of variable such that the remaining graph is
    tree
   S is called cycle cutset when constraint graph is a tree after the
    removal of S
   Time Complexity : O(dc(n-c)d2)
        c : cutset size
        dc: the # of possible cutset assignment :
        Time complexity in tree-structred CSP : (n-c)d2
        Runtime O(dc . (n-c)d2)
Heuristics for complexity reduction

Transforming to Tree-Structured CSP
Iterative algorithm with Min-conflicts
heuristic
 Initially assigns a (random) value to every variable
 reassign the value of one variable at a time
 Variable selection :
    randomly select any conflict variables
    minimum of conflicts with other variables
      Hill-climb with h(n) = total number of violated constraints
 Min-conflicts is effective for many CSPs
   particularly with reasonable initial state
 Specially important in scheduling problems
   Airline schedule (when the schedule is changed)
     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

      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




70
     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




      Given random initial state, can solve n-queens in almost constant
         time for arbitrary n with high probability (e.g., n = 10,000,000)


71
  Local Search for CSP
             1
             2                                    2
                                                  0
             3                                    2
             3                                    2
             2                                    2
             2                                    2
             3                                    2

Pick initial complete assignment (at random)
Repeat
     •     Pick a conflicted variable var (at random)
     •     Set the new value of var to minimize the number of conflicts
     •     If the new assignment is not conflicting then return it

(min-conflicts heuristics)
Remark
 Local search with min-conflict heuristic works extremely
  well for million-queen problems
 The reason: Solutions are densely distributed in the O(nn)
  space, which means that on the average a solution is a few
  steps away from a randomly picked assignment
Infinite-Domain CSP
 Variable domain is the set of the integers (discrete CSP) or of
  the real numbers (continuous CSP)
 Constraints are expressed as equalities and inequalities
 Particular case: Linear-programming problems
 Constraint Programming
     “Constraint programming
     represents one of the
     closest approaches
     computer science has yet
     made to the Holy Grail of
     programming: the user
     states the problem, the
     computer solves it.”
     Eugene C. Freuder, Constraints, April
     1997
When to Use CSP Techniques?
 When the problem can be expressed by a set of variables
  with constraints on their values
 When constraints are relatively simple (e.g., binary)
 When constraints propagate well (AC3 eliminates many
  values)
 When the solutions are “densely” distributed in the space of
  possible assignments
          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 work to
  constrain values and detect inconsistencies
 The CSP representation allows analysis of problem structure
 Tree-structured CSPs can be solved in linear time
 Iterative min-conflicts is usually effective in practice