This is a auto-learning robot for the bovo game(an open source FIR game). You can get all code in github.

FIR is short for 'Five in a Row'. This python program is running in Ubuntu14.10. By capture the shotcut of Bovo and estimate the activities of mouse, the program (robot) will be able to behavior as a human playing the FIR. Through the games between the Bovo's AI and the FIR robot, the robot can learn how to play.

The Model used by FIR learning robot.

• MODEL 1
  • P_t = W*B_t, P is 22x1 reshped chessboard probability matrix, W is linked wight matrix, B 22x1 reshped chessboard matrix. For t=N, P_N = F, F is the result of this game.
• MODEL 2
  • P_t = W*B_t + v, v is Gaussian white noise
• MODEL 3
  • X_t = W*B + X_{t-1} + v, X is the state matrix at time t.
  • P_t = U*X_t
  • For Model 1, the U works as a max function

By playing FIR game with Bovo AI, the program will store the chessboard record for training W matrix. The Model Mothods below will be applied for this aim.

• Model 1
  • Method 1
    • W = W + k*(P_{t+1} - P_t), t = 0...N-1
    • W = W + k*(F - P_{N-1})
  • Method 2
    • W = W + k*(T_t - P_t), t = 0...N-1, T_t is the step of the game ai at step t
    • W = W + k*(F - P_{N-1})

Copy the files in bovo-theme/spacy to bovo spacy folder to replace the theme.

Install the neccessary python dependencies. For Debian/Ubuntu user, you can

# apt-get install python-gtk2-dev
# apt-get install python-xlib

The 'Bovo' program has to be run before the robot running. For example, in 'Bovo' build file, execute ./bovo

run the robot, ./FIRLRobot.py

For Ubuntu user, you can install by searching 'bovo' in 'Ubuntu Software Center'.

You can also compile the code by yourself, visiting Bovo Icon Bovo - Development Information. Clone it down use the code below.

# git clone git://anongit.kde.org/bovo