Maintained by the ALR team at KIT, this Python framework can be used to simulate robotic tasks for scientific research and teaching. Our default environment consists of a Franka Panda robot model.
The Franka Panda robot, is a robotic arm with 7 degrees of freedom, 1KHz control, torque-sensing in all joints, access to multiple research apps and full ROS capability. Franka Panda is therefore widely used in scientific research, including robot manipulation, machine learning etc. For more information, please visit: Franka Panda
We currently support the following physic engines, which can be used interchangebly:
Additionally we provide some environments optimized for reinforcement learning. They follow the OpenAI Gym interface and support the deployment of Stable Baselines models.
This simulation framework has been tested in the following operating systems:
- Ubuntu 18.04, 20.04
- Mac OS
Ubuntu 22.04 currently only works in Xorg Mode, for a small tutorial on how to switch to Xorg from Wayland look here
Due to the various dependencies of the physics engines, we recommend installing the ALR Simulationframework in a Anaconda/Miniconda environment. Please follow the steps as described in our detailed Installation Guide
Please make sure that your code conforms to flake8 conventions. You can automatically check and correct all requirements by using pre-commit. Follow the installation instructions and run pre-commit run --all-files to verify your installation. Before each commit, all necessary style guides are checked and, if possible, automatically corrected.
We strongly recommend you familiarize yourself with some of the ALR Simulationframework's core concepts. Please refer the Guide Chapters marked as [recommended].
Physic engines:
- Pybullet
- Mujoco
Gyms & Reinforcement Learning: