CZ3004 MDP algorithm implemented using reinforcement learning

1. python2.7
2. if you get error messages regarding Tkinter, see this and this

1. change the path to grid in gridworld.py get_grid_from_file(<path>) to the path of your grid
2. change parameters _episodes and _training_episodes in gridworld.py to desired value
3. (optional setup)
   1. parameter _epsilon in gridworld.py to your desired exploration rate
   2. parameter __speed in gridworld.py to your desired number of movement per second
   3. parameter pauseCallback in gridworld.py
   4. function callback graphic_display in gridworld.py line366 to null_display
4. after training, remember to save weights manually in __init__() in qlearningAgents.py

You can design your own features by modifying get_features() in state.py and change the reward function in getReward() in gridworld.py respectively

http://ai.berkeley.edu/reinforcement.html