Skip to content

SHINDONGMYUNG/Value_Iteration_FrozenLake

Repository files navigation

Solving FrozenLake using exact methods

Solving FrozenLake-v0 environment using value iteration and Q-value iteration.
Environment description: https://gym.openai.com/envs/FrozenLake-v0/

Results

Value iteration solved MDP at 75 iterations.
Q-value iteration solved MDP at 58 iterations.
Both methods found slightly different optimal policies,
because there were multiple optimal actions in some states.
It tooks approximately 73 steps to reach a frisbee (when not falling into a hole),
and chance to fall into a hole is approx. 25% versus total trials.

val_result q_val_result

Reference

I referred code from here:
https://github.com/realdiganta/solving_openai/tree/master/FrozenLake8x8
Thanks!

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages