|
Contents Domain Illustration Methods Conclusions
|
Domain
Simulated Human Observations and Use of Tactics SHOUT was designed to mimic human tactics during game play. The computer was to act as an independent entity, that realistically and intelligently responded to a players game moves. The beginning stages of design dealt strictly with the rules of the game and ensured first most that the game would be able to be played correctly. The second phase of design dealt with developing the computer as an agent within the game.
Illustration
and Methods
Lets begin with the fundamentals of the game. The Game itself is called Connect Three. The fundamentals are very simple.
The game is played on a 5x5 board. Two players alternate placing red and black
tiles into the game board. The first player to make 3 in a row in any direction
of their respected color wins the game. The programming of this logic in the
game had to take certain things into consideration. - First the game had to learn the essence of the board existing. Breaking the
5x5 board into individual cells and numbering them created this. Each cell had
the ability to contain either a black, red, or null. A Simple prolog line would
implement this. I.e. [ is(Cell#, Val) ]. When the game began all cells would be
given the null value. If a player attempted to insert a block in a cell that was
either red or black an error would be popped back. - Second it was understood that the blocks were stacked. Therefore it would
have to check if the bottom cell was full before it would allow a block in that
respective cell. This was as well done with a preprogrammed list of possible
combinations. I.e. before the game began all cells were programmed with below
constraints. Before adding a block to a cell it would check 3 things. First if
that cell was empty through the is clause. Second if the cell below it was full,
again with a query to the is clause. Finally it would then place the block in
that slot. - Finally on every end of a move the game would check for a possible winner.
The list of preprogrammed possible combinations would be searched through, and
the game would check to see if three of the same color matched one of those
combinations. If so the game would end. If not, and while all cells were not
filled the game would continue, if all the cells filled up the game would
restart with no winner. With the logic behind the game built, I began to focus on how a human would
attack a game like this. Not only did I need to develop a human's plan of
actions (or tactics), but rather how the computer could mimic this behavior. In
my observations I noticed that human players break the game into 4 standard
approaches. These approaches later became directives that the computer would
play by. I felt that: - If at all possible if you could win on a particular move that you would go
for that move. - Secondly, if you could not win then you were next inclined to stopping your
opponent at winning at the next move. - If this was not an option either, I felt that you would try to setup a move
where about you could win on the next move. Preferably where you could win in
more than one place. - Finally, if you could not setup a move, you would be inclined to prevent
the other player from setting up a winning move. Since the game was built within guidelines, and the computer was
preprogrammed with possible winning combinations. Every move was simply a search
through winning configurations to determine the next move. You would search if
you could possibly win on this move, if that combination didn't exist you would
search if your opponent had 2 out of the three winning combination pieces and
the third being null. If winning or blocking wasn't an option, you would check
to see if you could place a piece that would give you more that 1 winning
combination with a 3rd piece being null. Finally if winning combinations are out
of reach you would prevent your opponent from gaining one. Through searches
these directives are enforced, and it is possible to mimic intelligence in the
computer opponent. The computer actually thinks and reacts as a human would. I found that in programming this application that more complex methods needed to be created to make a real simulated game. My game had a list of preprogrammed winning combinations, and the computer would search through all of the possibilities. In all it was a cool experience in learning about approaches to AI, and the way computer programmers attack gameplay. |