search

Klins101 / Peg-in-a-hole-insertion-using-DRL

Simulation and training a UR5 robot to accurately insert a pin into a hole using the CoppeliaSim simulator and catalyst-rl reinforcement library
9 stars 1 forks source link

readme

Peg-in-a-hole-insertion-using-DRL

Simulation and training a UR5 robot to accurately insert a pin into a hole using the CoppeliaSim simulator and catalyst-rl reinforcement library

Setup

  1. Download the robot simulation platform, CoppeliaSim, from the official website

  2. Setup toolkit for robot learning research, PyRep, from their github repository. PyRep library is built on top of CoppeliaSim to facilitate prototyping in python.

  3. The goal is to train a robot to perform peg-in-hole insertion. The simulation environment is in CoppeliaSim. The simulator comes with various robot manipulators and grippers. UR5 robot with RG2 gripper was used.

  4. Perform training
    Run scripts/run-training.sh

  5. Inference
    Run scripts/run-inference.sh