Status: Archive (code is provided as-is, no updates expected)

rosbridge

Warning: abandoned in Oct 2016, when we changed to a different system. It might be a good starting point for using a Fetch robot from Gym, but nobody at OpenAI is maintaining it

A service implementing a bridge from Gym to ROS robots. Currently supports the Fetch Research Robot

• Runs as a ROS node (ie, start it with roslaunch rosbridge fetch_proxy.launch.

• Listens on a ZMQ socket, which an OpenAI Gym ProxyClient connects to.

• Converts the action space to ROS commands, and the ROS sensors to an observation space.

Fetch Research robot Environments

• See Docs, especially the API

FetchRobot-v0: Action and observation space

• Action is a tuple with

• an 8x1 vector of joint torques, in range [-1 .. +1] corresponding to the following joints:
  • shoulder_pan_joint
  • shoulder_lift_joint
  • upperarm_roll_joint
  • elbow_flex_joint
  • forearm_roll_joint
  • wrist_flex_joint
  • wrist_roll_joint
  • l_gripper_finger_joint

• Observation is a tuple of

• An 8x1 vector of joint angles in range [-4 .. +4]
• An 8x1 vector of joint velocities in range [-4 .. +4]
• A 480x640 array of floats representing distance in meters

FetchRobotRGB-v0: Action and observation space

• Action: (same as FetchRobot-v0)
• Observation is a tuple of

• An 8x1 vector of joint angles in range [-4 .. +4] representing radians
• An 8x1 vector of joint velocities in range [-4 .. +4] representing radians/sec
• A 480x640x3 array of uint8s representing RGB pixel values

Installation on a Fetch:

See https://github.com/openai/fetch-config for installation scripts