core

Robot state, camera views, show() and finish(): what every program needs.

Layer: base
Needs: –
Requires: –
Conflicts with: –
Never dropped: yes
Mixin: skills.core:Core
Manifest: skills/core/skill.toml

Functions

`finish`

def finish(summary='')

End the episode now: the program’s own checks say the task is done. No further model call follows.

`get_state`

def get_state()

The robot state, read fresh from the robot: step, ended, joint_position (14: left 6 joints, left gripper, right 6 joints, right gripper), left_ee_pose and right_ee_pose (link6 poses), and step_limit when the connection reports it.

`get_observation`

def get_observation(cameras=None)

Current camera views as an Observation: .images[camera] (RGB uint8 HxWx3), .intrinsics[camera] (3x3), .camera_to_world[camera] (4x4, OpenCV convention) and .state. cameras is a camera name or a list of them; None means every camera.

`show`

def show(what, caption='')

Attach camera views or an RGB array to the next message you receive.

Also in the namespace: ARMS, CONTROL_HZ (constants and names the skill’s exports(session) provides).

What the model reads

From the RoboDojo prompt pack (prompts/robodojo/), with every skill and feature on:

finish(), when an unhandled event stops it, or when it raises an error; each time uses
  get_state() -> dict          step, step_limit, ended, joint_position (14: left 6 joints,
                               left gripper, right 6 joints, right gripper),
                               left_ee_pose, right_ee_pose
  get_observation(cameras=None) -> Observation
                               .images[camera] RGB uint8 HxWx3, .intrinsics[camera] 3x3,
                               .camera_to_world[camera] 4x4 (OpenCV convention), .state
show it (or reply <finish>). You are asked again when the program ends without