This project is meant to migrate core components and datasets of the C++ based UBC MOCCA software, which is called (TerrainRLSim)[https://github.com/UBCMOCCA/TerrainRLSim], into Gym. Thus this useful tool would be much easier to use, and potentially attract even more researchers join it.
- Converted MOCCA character (JSON format) into (mujoco)[http://www.mujoco.org] MJCF format.
- Handle joint limits/range.
- Add terrain data.
- Load and play motion data.
- Converting MOCCA terrain (JSON format) into mujoco format.
- Parsing MOCCA motion files with python.
- Play motion data with mujoco based environment.
- Being able to access motion state, e.g. position, quaternion, velocity etc., of every body part and end effector easily.
- Creating learning tasks (or known as env) with Gym manner.
- Creating RL learning nets, or reusing open source implementations such as (OpenAI)[https://openai.github.io] and DeepMind.
Python version 3.6+ is recommended.
Besides the requirements declared in requirements.txt, some external packages are required.
- (gym)[https://github.com/openai/gym]
- (mujoco-py)[https://github.com/openai/mujoco-py]
- (baselines)[https://github.com/openai/baselines]