This repo hosts the implementation of tutorials for "Modelling and Control of Legged Robots" course.
Clone the repo:
git clone https://github.com/Jonas-Zhang97/TUM-MCLR.git
If you are familiar with devcontainer, you can skip this steps, direct to this page or this readme to ckeck out the code.
- Enable remote development extension in your VSCode (extension ID: ms-vscode-remote.vscode-remote-extensionpack).
- It is recommended to use the docker container in a computer with NVidia GPU, to start directly with GPU, you can skip to the get started, if not available, follow this chapter to build the docker container without GPU
- For the final project, a Nvidia GPU is sufficient, and the utilizes functionalities from Isaac Sim for reinforcement training and inference, for which,
omniverse-launcher-linux.AppImage
is required, it can be downloadedhere, you should download the.AppImage
version and put it into theisaac/
folder.
Navigate to the .devcontainer/Dockerfile
, in stage 4, comment the following lines:
ENV NVIDIA_VISIBLE_DEVICES \
${NVIDIA_VISIBLE_DEVICES:-all}
ENV NVIDIA_DRIVER_CAPABILITIES \
${NVIDIA_DRIVER_CAPABILITIES:+$NVIDIA_DRIVER_CAPABILITIES,}graphics
then, to the file .devcontainer/devcontainer.json
, in runArgs
comment the following lines:
"--runtime=nvidia"
"--gpus=all"
Open the local repo in VSCode, you should see a pop-up asking if you want to reopen in devcontainer, click yes. If you don't see any pop-up, press ctrl + shift + p, type "remote" in searchbar and choose rebuild and reopen in devcontainer
.
After the container is built, you can then check this page or this readme for the code.
Final project of this course is defined as: Reinforcement Learning Based Walking Control for Humanoid Robot.
The project is based on Talos from PAL robotics and Isaac-Sim from Nvidia. Configure the Omniverse app as descriped in prerequest.
To run Omniverse app in the container, first change the mode:
chmod +x ./isaac/omniverse-launcher-linux.AppImage
then run it in the container with command:
dbus-launch ./isaac/omniverse-launcher-linux.AppImage --no-sandbox
Intsall isaac sim 4.0.0
in the omniverse app (the default install entry is a legacy version). Install the isaac lab
according to the official documentation and follow the tutorial to get a overview of isaac lab.
- Modify this file for a new terrain
- FromAsCasing issue in Dockerfile (not fetal, fix in the end)