Reinforcement-Learning-Frozen-Lake-maze

Using First-visit Monte Carlo method, Q-learning, Sarsa algorithms to solve the Frozen Lake problem

Requirements:

numpy 1.18.5, pandas1.1.2, matplotlib3.3.2, tkinter

agent_brain4_4.py:

Include three algorithms (MenteCarolo, Q-learning, Sarsa) of the RL model for 4*4 frozen lake.

agent_brain10_10.py:

Include three algorithms (MenteCarolo, Q-learning, Sarsa) of the RL model for 10*10 frozen lake.

env4_4.py:

4*4 frozen lake environment with 4 fixed holes.

env10_10.py:

10*10 frozen lake environment with 25 random holes.

To test the algorithm and train the model please run the belows respectively:

Monte_Carlo_train4_4.py

Monte_Carlo_train10_10.py

Qlearning_train4_4.py

Qlearning_train10_10.py

Sarsa_train4_4.py

Sarsa_train10_10.py