openai / gym

A toolkit for developing and comparing reinforcement learning algorithms.
https://www.gymlibrary.dev
Other
34.66k stars 8.6k forks source link

Maze environments #279

Closed danijar closed 3 years ago

danijar commented 8 years ago

It would be cool to have grid world environments as common in RL literature. Is that something you are thinking about?

ppwwyyxx commented 8 years ago

I did have some maze wad files, as well as code to randomly produce them. They look like this: maze

danijar commented 8 years ago

Is this for Doom? That's very interesting but for testing and comparing new algorithms, it would be nice to test them on fully-observed 2D grid world tasks, where the four actions move the current position to the four adjacent cells, respectively.

ppwwyyxx commented 8 years ago

Oh I forgot to mention it is for doom. Yes having a 2D fully-observed maze would be nice. This is a partially-observed 3D one and it would be much more challenging.

sanjinzhi commented 6 years ago

Is there 3D grid environment for gym?

stale[bot] commented 5 years ago

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

mryellow commented 5 years ago

It's not fully polished and a little complex agents to solve, but generates mazes for episodes and wraps them in Gym.

https://github.com/mryellow/gym-mazeexplorer
https://github.com/mryellow/maze_explorer

jkterry1 commented 3 years ago

I'm not open to accepting new environments in Gym at this time. However, @mryellow and others, I welcome PRs to add more environments to the third part environments list.