We start our in-depth tour of general game
playing by looking at single player games.
Such as Sudoku, or sliding tile puzzles,
Rubik's cube, and so forth.
In the game playing community, these are
often called puzzles, rather than games.
And the process of solving such puzzles is
often called problem solving rather than
game playing.
Puzzles are simpler
than multiple player games because
everything's
under the control of a single player.
World static, except when the player acts,
and changes to the world
are determined entirely by the current
state and the actions of the player.
In this lesson, as in most of the course
we
assume that the player has complete
information abut the puzzle.
Assume that it knows the initial state.
You know that, it knows its legal actions
in every state.
It knows the effects of its actions in
every state.
For every state, it knows its reward, and
for
every state, it knows whether or not it's
terminal.
In this lesson, we also assume the games
are small.
That is the player has sufficient space
and time to search the entire game tree.
This guarantees that the player can find
optimal actions to perform.
Now that said, as we'll see, it's
sometimes possible to
find optimal actions even without
searching the entire game tree.
Despite these strong assumptions, just one
player, complete information,
availability of adequate time to search
the game tree.
The study of single player games is a good
place to start our look at general game
playing.
First of all, many real world problems can
be
cast as single player games with these
same restrictions.
Su, such as finding possible protein
foldings as suggested by the illustration
here.
[COUGH]
More importantly for us, as we'll see.
The techniques we examine later can be
viewed as
more elaborate versions of the basic
techniques introduced here.