Datasets:

---

OpenGalaxea

/

Galaxea-Open-World-Dataset

like 24

Follow

Galaxea 33

Hi, thanks for excellent work!

I have a question, I was just reading the dataset card

# action dimensions:
# 26 = 6 (left arm) + 1 (left gripper) + 6 (right arm) + 1 (right gripper) + 6 (torso) + 6 (base)
"action": tf.Tensor(26, dtype=float32),  # robot action, consists of [6x joint velocities, 1x gripper position]

and I am confused about the specific meaning of the actions

• What's the meaning of 6 dim torso? joint_position_torso is a tf.Tensor(4, dtype=float32) in card..
• What's the meaning of 6 dim base? base_velocity is a tf.Tensor(3, dtype=float32) in card..

Thanks for your interest!

In our dataset, the arms and grippers are controlled in position, while the torso and chassis are controlled in velocity.

Action: The chassis and base actions are collected via ROS geometry_msgs/TwistStamped during teleoperation. Therefore, the [6× torso, 6× base] corresponds to [velocity.x, velocity.y, velocity.z, angular_velocity.x, angular_velocity.y, angular_velocity.z] (with some constant zero values).

Observation: The proprioceptive states are positional. Specifically, joint_position_torso is a 4-dimensional vector (positions of 3 joints plus one zero-padding value), and joint_position_chassis is a 3-dimensional vector (heading of the three wheels).