Today we are going to talk about Artificial Intelligence
(AI).
The prototype and the first version of the game.
First was made a paper prototype with some powers, with this
test the game worked, but still depended on two people;
In the first digital version in Unity of the game, still,
there was a version of the implemented one of AI. It was a two-player version.
Depending on the cleverness of the players.
 |
| Prototype |
The initial concept of AI.
First we thought of plays used more in the prototypes in
paper and in the digital version. We also thought of a hierarchy of actions.
Then the team drew a few more recurring moves and were placed in the AI code.
However, this mapping is still insufficient. With this we are using an
algorithm to map the moves and improve AI.
The technical functioning of the AI.
To decide which one to play the AI, Tic Tac Toe SP analyzes
the board for each situation, and assigns points to each house according to the
situations, if found. The house with the most points is the one in which the AI
must play.
During the analysis of the board, each house is assigned two
values: priority (number of points) and type of action: none, basic, prevent
and destroy (this list is subject to change).
None: AI can not play in that house.
Basic: The AI must play its basic piece (X or O) in that
house.
Prevent: AI must play a part (ice, plan, its normal part
even) to disrupt an opponent's move. The difference to basic is that it can use
ice and plant as well.
Destroy: The AI must destroy the piece that is in the house,
with a blowtorch or a bomb.
If the AI can not perform the action with more points (for
not having a blowtorch or bomb to destroy, for example), this is discarded and
the next one with more points is attempted.
At each function, the AI looks for a specific situation on
the board. In the above example, you would find two loose pieces of the
opponent and two of their boards. He does not currently look for an isolated
block of ice, as it does not matter - if he were to be together with other
pieces of that might matter.
The situations that the program looks for are the following:
Glossary:
Basic piece of the player: The basic piece of the player (in the case of AI) - X or O
Basic piece of the opponent: The opponent's basic piece - X or O
Priority: number of points, in the end the house with the highest priority is the one chosen to be played.
Empty house: house without any part placed
Glossary:
Basic piece of the player: The basic piece of the player (in the case of AI) - X or O
Basic piece of the opponent: The opponent's basic piece - X or O
Priority: number of points, in the end the house with the highest priority is the one chosen to be played.
Empty house: house without any part placed
CheckPlayerOneBasicPiece ();
Priority: 50
Action: Basic
Look for the basic pieces of the player, regardless of
whether they are together or not. Then assign points to all houses around these
that are empty.
CheckPlayerTwoBasicPieces ();
Priority: 75
Action: Basic
Look for basic player pieces that are adjacent to each
other. If he finds one piece adjacent to another, he assigns the points to the
houses that allow him to complete three pieces (two at most) - those that are empty, obviously.
CheckEnemyTwoBasicPieces ();
Priority: 74
Action: Prevent
Like the previous one, but look for the opponent's pieces.
If he finds one piece adjacent to another, assign the points to the houses that
the opponent could use to complete three pieces (two at the most) - those that
are obviously empty.
Obs: In the current state of the game, it seems that the
houses assigned by this function have not been chosen: there is always another
house with more points.
CheckPlayerThreeBasicPieces ();
Priority: 100
Action: Basic
Look for three basic pieces of the player that are adjacent,
and you only need to put one more piece to win the game. The priority is
maximum, and at most two houses will be assigned points (those that are empty).
CheckEnemyThreeBasicPieces ();
Priority: 95
Action: Prevent
Like the previous one, but look for the opponent's pieces.
It is very important to prevent his victory, only the AI victory on this turn is a better move than this.
CheckEnemyTwoAndAFourthBasicPieces ();
CheckEnemyOneAndThirdFourthBasicPieces ();
Priority: 95
Action: Prevent
Look for cases where the opponent has two pieces together,
an empty house, and another piece, and playing in the void would give him a
victory. Assign the points to this empty house, and have the same priority of
three pieces of it together.
CheckEnemyOneAndThirdFourthBasicPiecesAndAnIceBetween ();
CheckEnemyTwoAndAFourthBasicPiecesAndAnIceBetween ();
Priority: 90
Action: Destroy
Here we look for the situation of having two pieces, one
ice, and another piece aligned. Destroy one of the pieces.
Note: In fact, this is the only case currently in which the
AI is requested to use the blowtorch or pump.
ResetActionsBoard ();
Priority: 0
Action: Basic or None
Ensures that all boxes in the chessboard have assigned
values. It assigns zero priority to each house, and if it is occupied it is
assigned a none action, prohibiting action on it.
Note: The assigned values can be replaced later by another
of the above functions, including occupied houses, in which a destroy: action
can be assigned to destroy the part.
Note 2: The order in which the functions are executed does
not matter, because only higher priority values can replace others. If a
function tries to assign values with lower priority, it is prevented to do that.
