ur_openai_gym

OpenAI Gym interface for Universal Robots with ROS Gazebo based on openai_ros

Examples

Reinforcement Learning with Soft-Actor-Critic (SAC) with the implementation from TF2RL with 2 action spaces: task-space (end-effector Cartesian space) and joint-space.

Start the simulation environment based on ur3 roslaunch ur3_gazebo ur3e_cubes_example.launch
Execute the learning session:

For task-space example: rosrun ur_rl tf2rl_sac.py -e 0

For task-space example: rosrun ur_rl tf2rl_sac.py -e 1

Install with Docker