Robot-Learning-Library

Robot control with learning methods is hard in general, we open-source the code here for boosting our research and development.

Contents:

What we will provide in this repository:

• Robot learning environments:

  • Based on MuJoCo;
  • Based on Isaac Sim;

• Basic robotic control:

  • Inverse kinematics;
  • Trajectory generation;
  • System identification;

• Learning methods:

  • Domain randomization；
  • A deep RL library with PyTorch (or JAX, to be determined);
  • Imitation learning;
  • Meta-learning across tasks;

• Real robot control:

  • ROS;

• Additional:

  • Tactile sensory;

Projects:

There is a list of projects using these repository:

• Sim-to-Real Transfer for Robotic Manipulation with Tactile Sensory. Zihan Ding, Ya-Yen Tsai, Wang Wei Lee, Bidan Huang (IROS 2021)
• Not Only Domain Randomization: Universal Policy with Embedding System Identification. Zihan Ding