This is accompanying code for the (research-lightning)[https://github.com/jhejna/research-lightning] repository. It provides a number of different robot controllers which interface with the robot environments in research lightning.
On the desktop computer, first create a conda environment with the desired dependencies for policy inference. Afterwards, we will setup the dependencies for robot-lightning.
First, install the library dependencies via pip:
pip install -r requirements.txt
Or if using conda, you may want to add these to your conda environment file.
If you want to use VR, install oculus_reader.
- Install
git lfsvia pip instead of following the sudo setup instructions. This can be done withpip install git-lfs. - Install adb devices:
sudo apt install android-tools-adb. - After doing this, you can clone the oculus reader repository with
git clone https://github.com/rail-berkeley/oculus_reader. - Then, install the package from the oculus reader repo with
pip install -e .
To collect demos, run
python scripts/collect_demos.py --path path/to/save/ --config path/to/robot/config
If the oculus isn't working due to adb permissions, run the following:
sudo adb kill-server
sudo adb start-server
On the NUC, first make sure to install the intel realtime kernel. Then, follow the setup instructions for individual packages.
We use the monometis repo found here.
- Create the monometis conda environment by running
conda env create -f environments/nuc_monometis_env.yaml. - Clone the monometis repo:
git clone https://github.com/hengyuan-hu/monometis/tree/main. - Compile the library, following these instructions:
./scripts/build_libfranka.sh
mkdir -p ./polymetis/build
cd ./polymetis/build
cmake .. -DCMAKE_BUILD_TYPE=Release -DBUILD_FRANKA=ON
make -j
cd ../..
- Install polymetis:
pip install -e ./polymetis
Then, you can start the polymetis server as you would normally.
If you want to use ZeroRPC to wrap polymetis, you need to additionally install the robots package, then run
python scripts/start_nuc_remote.py
from the the NUC.