fork of https://github.com/kngwyu/mujoco-maze with some fixes
Go to file
2020-09-28 01:16:54 +09:00
mujoco_maze Rolling 2020-09-28 01:16:54 +09:00
screenshots [Four Rooms] Change the start position 2020-09-07 17:58:29 +09:00
tests Rolling 2020-09-28 01:16:54 +09:00
.gitignore INITIAL 2020-05-21 19:05:06 +09:00
LICENSE INITIAL 2020-05-21 19:05:06 +09:00
poetry.lock Make it work 2020-05-25 17:26:57 +09:00
pyproject.toml Introduce MazeCell and refactor 2020-06-04 14:31:00 +09:00
README.md Block Maze 2020-09-26 18:38:19 +09:00
setup.cfg INITIAL 2020-05-21 19:05:06 +09:00
setup.py Use self.sim instead of physics 2020-05-29 17:04:26 +09:00

mujoco-maze

Some maze environments for reinforcement learning(RL) using mujoco-py and openai gym.

Thankfully, this project is based on the code from rllab and tensorflow/models.

Environments

  • PointUMaze/AntUmaze

    PointUMaze

    • PointUMaze-v0/AntUMaze-v0 (Distance-based Reward)
    • PointUmaze-v1/AntUMaze-v1 (Goal-based Reward i.e., 1.0 or -ε)
  • Point4Rooms/Ant4Rooms

    Point4Rooms

    • Point4Rooms-v0/Ant4Rooms-v0 (Distance-based Reward)
    • Point4Rooms-v1/Ant4Rooms-v1 (Goal-based Reward)
    • Point4Rooms-v2/Ant4Rooms-v2 (Multiple Goals (0.5 pt or 1.0 pt))
  • PointPush/AntPush

    PointPush

    • PointPush-v0/AntPush-v0 (Distance-based Reward)
    • PointPush-v1/AntPush-v1 (Goal-based Reward)
  • PointFall/AntFall

    PointFall

    • PointFall-v0/AntFall-v0 (Distance-based Reward)
    • PointFall-v1/AntFall-v1 (Goal-based Reward)

Caveats

This project has some other features (e.g., block maze and other robots) but they are work in progress.

License

This project is licensed under Apache License, Version 2.0 (LICENSE-APACHE or http://www.apache.org/licenses/LICENSE-2.0).