Skip to content

Latest commit

 

History

History
26 lines (15 loc) · 478 Bytes

plan.md

File metadata and controls

26 lines (15 loc) · 478 Bytes
Raw

Catching a flying ball with reinforcement leaning

2019.08.29.

  1. revise the ur5 env, similar with fetch robot, ignore the gripper fist.
  2. train the ur5 to reach the ball.
  3. try different reward function

Structure

  1. create robot environment

    • mujoco env
    • ROS env
    • real robot

  2. apply rl algorithms to train the model

    • train the model in mujoco
    • test the model in mujoco
    • test the model in ros
    • test the model in real robot