/robot-learning-workspace

Workspace for robot learning experiments/prototyping using pybullet simulation

Primary LanguageJupyter NotebookMIT LicenseMIT

Pybullet playground for Robot Learning.

The robot setup mimicks the AIRO research setup with UR3e robots, Robotiq 2F85 grippers and Zed2i cameras.

Pybullet Env

Local Installation

  • create a conda environment conda env create -n robot-learning-playground and activate it with conda activate robot-learning-playground (use python3.9 for ikfast)
  • move into the ur_sim folder and run pip install -r requirements.txt
  • manually install the ur_ikfast package using the instructions here (and test using test_ikfast.py)
  • test that pybullet is working by running the push_env.py file. You should see the robot moving.

Learning Algorithms

Local Installation

  • activate the conda environment
  • from the learning folder run conda env update -f learning.yml