AIDEN

Check Out This Project

This project is based on the game Simplexity designed by Brain Blender Games. The AI is composed by 2 classes: G02AIDEN and G02AIDENThinker. The G02AIDEN class (which extends AIPlayer) is used to instantiate the G02AIDENThinker class and to set the AI's name. The G02AIDENThinker class (which implements IThinker) has 3 main methods:

The Think() method accepts a (copy of the) Board and a CancellationToken. It starts by defining an array of int which later gets filled in a for loop that defines the AI's columns ordering preference by a static function which zig zags from the middle column. After that, Negamax() gets called to start the AI's decision making process. If too much time goes by, no move is performed and the AI forfeits the match.

The NegaMax() method returns a score based on the selected position. It has a Board, CancellationToken, and two int type numbers as parameters. The algorithm used here is Negamax with alpha beta pruning, move ordering and deepening, which means it will select a position a lot faster than the regular Negamax. Firstly, the method creates an int named bestScore with its value set to the lowest int value possible. Afterwards, the current AI's depth is compared to its max value (maxDepth) and the game verifies if there is any Winner. If either of these two conditions is true, the bestScore gets returned immediately. After that, a max value is set and with the alpha's value, gets compared to the current beta in order to help define (and control) it. Lastly in this method there is a for loop which checks all columns by the prefered order. In this loop, if a column is detected as full, then it's skipped and we move towards the next. If not, two scores are set to 0 and a new variable to hold a piece shape (PShape) is created. The AI checks if there are still pieces of a certain shape and if so we perform a move on the currently selected column with that piece shape. The depth is increased and a variable holding a score of that piece type is set as the sum of the inverted Negamax() method and the Heuristics() method, switching the alpha and beta values when the Negamax() gets called. The depth is decreased and the previous move is undone (this is done for each shape type, seeing as any player can play any shape). The AI then checks if either of the last scores are greater than the current bestScore and if so the bestScore is updated to be the greater of the two values. The best shape (playShape) is changed to represent the shape with he highest score. A check is then done to know if there are still round or square pieces available, and if not, the current used piece is replaced by its opposite. The futureMove variable is set to a new FutureMove with the current column and playShape.

Lastly, the AI checks if the bestScore is greater or equal to the current beta value and if so, its best score is returned. Then, it verifies if the bestScore is greater than alpha, and if so, the AI sets the alpha to have its value, returning the alpha.

In the Heuristics() method, when a shape or color gets found on a corridor, a counter gets increased. If the next position checked is empty, the counter is reset and if the next position checked has a shape different from the previous shape and a color different from the previous color and counter has at least a value of 2, the final score is increased, the counter gets reset and the AI saves the shape of the current shape and color of the piece. According to its depth level, the final score can be returned as inverted or normal.