Use Q-Learning and SARSA to solve maze problem generated randomly
IMPORTANT: comment the env.render() can obtain results quicker (rendering surely much slower than CPU staffssssss).
-
gym_maze: the gym library for generating experiment environment
-
How to run? Simply run .py that has a main function
-
Why it's a mess? I changed the value of a lot of parameters to do experiments.
-
Any innovation point? See the last part in the report.
-
the lr=xxx folders? Stored the experiment result.
I think you can find everything you need in my report.
For DQN version maze solver, see: https://github.com/saaries/Maze_DQN_reinforcement_learning
A screenshot for reference, check if this repo is what you're looking for:
