Allowing Safe Contact in Robotic Goal-Reaching

Introduction

This repo includes code for the paper "Allowing Safe Contact in Robotic Goal-Reaching: Planning and Tracking in Operational and Null Spaces".

This repo provides Jupyter notebooks and configuration files to run the simulation experiments in the paper. Besides the ball and wall environments, we further provide a strawberry picking environment and a PyBullet robot simulator.

Install

In the Conda env, install Jax, Brax, and other libraries with the following command

• pip install --upgrade "jax[cuda]" -f https://storage.googleapis.com/jax-releases/jax_cuda_releases.html
  • We highly recommend to use the CUDA version, results in the paper are computed with CUDA support
  • Please check the installation manual for details
• pip install brax pybullet notebook tqdm scipy

Usage

• Use Brax_ball.ipynb to play with the ball environment
• Use Brax_wall.ipynb to play with the wall environment
• Use Brax_strawberry.ipynb to pick strawberries :)
• Use PyBullet_Kuka.ipynb to see the compliance controller in PyBullet