State-Space Representation
150040707002
What you should know
Create a state-space model
Estimate number of states
Identify goal or objective function
Identify operators
Next Lecture: how to search/use model
Everyday Problem Solving
Route Planning
Finding and navigating to a classroom seat
Replanning if someone cuts in front
Driving to school
Constant updating due to traffic
Putting the dishes away
Spatial reasoning
Goal: Generality
People are good at multiple tasks
Same model of problem solving for all
problems
Generality via abstraction and
simplification.
Toy problems as benchmarks for methods,
not goal.
AI criticism: generality is not free
State-Space Model
Initial State
Operators: maps a state into a next state
alternative: successors of state
Goal Predicate: test to see if goal achieved
Optional:
cost of operators
cost of solution
Major Simplifications
You know the world perfectly
No one tells you how to represent the world
Sensors always make mistakes
You know what operators do
Operators dont always work
You know the set of legal operators
No one tells you the operators
8-Queens Model 1
Initial State: empty 8 by 8 board
Operators:
add a queen to empty square
remove a queen
[move a queen to new empty square]
Goal: no queen attacks another queen
Eight queens on board
Good enough? Can a solution be found?
8-Queens Model 2
Initial State: empty 8 by 8 board
Operators:
add ith queen to some column (i = 1..8)
Ith queen is in row i
Goal: no queen attacks another queen
8 queens on board
Good enough?
8-Queens Model 3
Initial State:
random placement of 8 queens ( 1 per row)
Operators:
move a queen to new position (in same row)
Goal: no queen attacks another queen
8 queens on board
Minton
Million Queens problem
Cant be solved by complete methods
Easy by Local Improvement
to be covered next week
Same method works for many real-world
problems.
Traveling Salesman Problem
Given: n cities and distances
Initial State: fix a city
Operators:
add a city to current path
[move a city to new position]
[swap two cities]
[UNCROSS]
Goal: cheapest path visiting all cities once and
returning.
TSP
Clay prize: $1,000,000 if prove can be done
in polynomial time or not.
Number of paths is N!
Similar to many real-world problems.
Often content with best achievable:
bounded rationality
Sliding Tile Puzzle
8 by 8 or 15 by 15 board
Initial State:
Operators:
Goal:
Sliding Tile Puzzle
8 by 8 or 15 by 15 board
Initial State: random (nearly) of number 1..7
or 1..14.
Operators:
slide tile to adjacent free square.
Goal: All tiles in order.
Note: Any complete information puzzle fits
this model.
Cryptarithmetic
Ex: SEND+MORE = MONEY
Initial State:
Operators:
Goal:
Cryptarithmetic
SEND+MORE = MONEY
Initial State: no variable has a value
Operators:
assign a variable a digit (0..9) (no dups)
unassign a variable
Goal: arithmetic statement is true.
Example of Constraint Satisfaction Problem
Boolean Satisfiability (3-sat)
$1,000,000 problem
Problem example (a1 +~a4+a7)&(.)
Initial State:
Operators
Goal:
Boolean Satisfiability (3-sat)
Problem example (a1 +~a4+a7)&(.)
Initial State: no variables are assigned values
Operators
assign variable to true or false
negate value of variable (t->f, f->t)
Goal: boolean expression is satisfied.
$1,000,000 problem
Ratio of clauses to variables breaks problem into 3 classes:
low ratio : easy to solve
high ratio: easy to show unsolvable
mid ratio: hard
CrossWord Solving
Initial-State: empty board
Operators:
add a word that
Matches definition
Matches filled in letters
Remove a word
Goal: board filled
Most Common Word
(Misspelled) Finding
Given: word length + set of strings
Find: most common word to all strings
Warning: word may be misspelled.
length 5: hellohoutemary position 5
bargainsamhotseview
position 10
tomdogarmyprogramhomse position 17
answer: HOUSE
Misspelled Word Finding
Let pi be position of word in string i
Initial state: pi = random position
Operators: assign pi to new position
Goal state: position yielding word with
fewest misspellings
Problem derived from Bioinformatics
finds regulatory elements; these determine
whether gene are made into proteins.
