Manual controller using your Hand and Keyboard. Hand pose is recognized via web-cam.
RL demonstration can be easily obtained by this especially for multi dimensional continuous control environment.

LunarLanderContinuous-v2
(index and middle finger)

RobogymDactyl
(all fingers and F1~4 keys)

Install this package and dependencies.

git clone https://github.com/NM512/HandyController.git
cd HandyController
pip install -e .

python wrappers/lunarlander.py

Please follow the offcial guide to install Robogym environment.
Don't forget to install Mujoco.

python wrappers/robogym_dactyl.py

Please make a wrapper for your own environment with reference to scripts inside wrappers folder.
Basically, only thing you have to consider is a correspondence between action space and input interface.
Currentry hand pose and keyboard input are supported as input interface.