Problem Reduction

Search
AND/OR Graphs (AO* Algorithm)
 Problem Reduction Search
 Planning how best to solve a problem that can
be recursively decomposed into sub problems
in multiple ways.
 Formulations
 AND/OR Graphs
 An OR node represents a choice between possible
decompositions
 An AND node represents a given decomposition

 Game Trees
 Max nodes represent the choice of my opponent
 Min nodes represent my choice
 OR
A1A2A3

AND
(A1A2)A3 A1(A2A3)
AND

A3 A1
OR
A1A2 OR OR
(A2A3) OR

A1 = 3 X 4
A2 = 4 X 10
A3 = 10 X 1
 OR
A1A2A3

AND
(A1A2)A3 A1(A2A3)
AND

A3 A1
OR
A1A2 OR OR
(A2A3) OR

40 0
A1 = 3 X 4 [ 4 X 1] [ 3 X 4]
A2 = 4 X 10
A3 = 10 X 1
 OR
A1A2A3

(A1A2)A3 A1(A2A3)
AND 12 + 40 AND
=52 [3 X 1]

A3 A1
OR
A1A2 OR OR
(A2A3) OR

40 0
A1 = 3 X 4 [4 X 1] [3 X 4]
A2 = 4 X 10
A3 = 10 X 1
 OR
A1A2A3

(A1A2)A3 A1(A2A3)
AND 12 + 40 AND
=52 [3 X 1]

A3 A1
OR
A1A2 OR OR
(A2A3) OR
[3 X 10] [10 X 1]
120 0 40 0
A1 = 3 X 4 [4 X 1] [3 X 4]
A2 = 4 X 10
A3 = 10 X 1
 OR
A1A2A3

(A1A2)A3 A1(A2A3)
30 +120 AND 12 + 40 AND
=150 [3 X 1] =52 [3 X 1]

A3 A1
OR
A1A2 OR OR
(A2A3) OR
[3 X 10] [10 X 1]
120 0 40 0
A1 = 3 X 4 [4 X 1] [3 X 4]
A2 = 4 X 10
A3 = 10 X 1
 OR
A1A2A3

(A1A2)A3 A1(A2A3)
30 +120 AND 12 + 40 AND
=150 [3 X 1] =52 [3 X 1]

A3 A1
T
A1A2 T T
(A2A3) T
[3 X 10] [10 X 1]
120 0 40 0
A1 = 3 X 4 [4 X 1] [3 X 4]
A2 = 4 X 10 This is solved in the absence of
A3 = 10 X 1
heuristic function
AND/OR Graph Search Problem
 Problem definition:
 Given: [G, s, T, h] where:
 G: implicitly specified AND/OR graph
 s: start node of the AND/OR graph
 T: set of terminal nodes

 h(n): heuristic function estimating the cost of solving

2 1

OR
h=4 OR
h=3
 AND
h=10

2 1
h=4 h=3
OR OR
1
1
h=8
T
h=6 T
 AND
h=10 14

2 1
h=4 h=3 7
OR OR
1
1
h=8
T
h=6 T
 AND
h=14

2 1
h=4 h= 7
OR OR

2 1 1
1
h=8
h=2 h=5 T
AND AND h=6 T
 AND
h=14

2 1
h=4 h= 7
OR OR

2 1 1
1
h=8
h=2 14 h=5 T
AND AND h=6 T
0 h=10
1
OR
OR
h=3
 AND
h=14

2 1
h=4 16 h= 7
OR OR

2 1 1
1
h=8
h=14 h=5 T
AND AND h=6 T
0 h=10
1
OR
OR
h=3
 AND
h=14

2 1
h=16 h= 7
OR OR

2 1 1
1
h=8
h=14 h=5 T
AND AND h=6 T
0 h=10
1
OR
OR
h=3
 AND
h=14

2 1
h=16 6 h= 7
OR OR

2 1 1
1
h=8
h=14 h=5 T
AND AND h=6 T
0 h=10
1
OR
OR
h=3
 AND
h=14 16

2 1
h=6 h= 7
OR OR

2 1 1
1
h=8
h=14 h=5 T
AND AND h=6 T
0 h=10
1
OR
OR
h=3
 AND
h=16

2 1
h=6 h= 7
OR OR

2 1 1
1
h=8
h=14 h=5 T
AND AND h=6 T
0 1
h=10
1
T 1 h=3
T
OR h=2
h=3 T
 AND
h=16

2 1
h=6 h= 7
OR OR

2 1 1
1
h=8
h=14 h=5 7 T
AND AND h=6 T
0 1
h=10
1
T 1 h=3
T
OR h=2
h=3 T
 AND
h=16

2 1
h=6 8 h= 7
OR OR

2 1 1
1
h=8
h=14 h=7 T
AND AND h=6 T
0 1
h=10
1
T 1 h=3
T
OR h=2
h=3 T
 AND
h=16 18

2 1
h=8 h= 7
OR OR

2 1 1
1
h=8
h=14 h=7 T
AND AND h=6 T
0 1
h=10
1
T 1 h=3
T
OR h=2
h=3 T
 AND
h=18
2 1
h=8 h= 7
OR OR

2 1 1
1
h=8
h=14 h=7 T
AND AND h=6 T
0 1
h=10
1
T 1 h=3
T
OR h=2
h=3 T
 AND
h=18
2 1
h=8 h= 7
OR OR

2 1 1
1
h=8
h=14 h=7 T
AND AND h=6 T
0 1
h=10
1
T 1 h=3
T
OR h=2
h=3 T Now all the leaf nodes of the marked
tree are solved. Thus solve each sub
problem(node) at marked sub tree.
 AO* Algorithm
1. Initialize: Set G* = {s}, f(s) = h(s)
if s  T, label s as SOLVED
2. Terminate: if s is SOLVED, then Terminate
3. Select: A non-terminal leaf node n from
the marked sub tree.
4. Expand: Mark explicit the successors of n
For each new successor m:
set f(m) = h(m)
if m  T, label m as SOLVED
 AO* Algorithm
5. Cost Revision: Call cost-revise(n)
6. Loop: Goto step 2
 Cost Revisoin in AO*
cost-revise(n)
1. Create Z={n}
2. If Z = { } then return
3. Select the node m from Z such that m has no
descendants in Z
4. If m is an AND node with successors R1, R2,
R3, …….. Rk
set f(m) =  [f(Ri) + C(m,Ri)]
Mark the edge to each successor of m
if each successor is labeled SOLVED then
label m as SOLVED
 Cost-revise(n)
5. If m is an OR node with successors R1,
R2, R3, …….. Rk
set f(m) = min [f(Ri) + C(m,Ri)]
Mark the edge to best successor of m
if marked successor is labeled SOLVED
then label m as SOLVED
6. If the cost or label of m has changed, then
insert those parents of m into Z for which
m is a marked successor
7. Goto step 2
 Searching OR Graph
 How would AO* behave when we have only
OR nodes ????