The program have two parts: 44 environment and 1010 environment. There are still some bugs in monte_carlo_state_transfer. Remaining the file want to illustrate it is a bad idea to use two coordinates in the program.
Numpy, time, sys, random, tkinter, matplotlib packages are needed. Please do not Change the Place File located with. The packages inside a folder will be needed and are different in different folders even their name are the same.
maze_env is the environment creation file, consisting of a class with init, build, reset and step functions.
X_brain files mainly consist of a class with init, check state, change_greedy, table calculate and table renew functions.
X_learning files consist of update function. (Special functions are added due to differences between algorithms.)
Test files are used to doing multiple learning process. result_display files are used to plot and show the results.
When doing verification, run the x_learning files. (Q learning is always the best one.) If want to see multiple results, run the test files. Initial holes generating policy is DFS policy in 'Sarsa_and_Qlearning_random_environment'. (Can be changed in maze_env.py in that folder.)
Yue Zenglin 岳增霖 2020.10 Mail Address e0575902@u.nus.edu