una-auxme/

ROS2SmolVLA_ur10e_no_joints_crop_pick_place

This is our pick and place model trained for the ROS2SmolVLA project. It has been trained on this dataset: ROS2SmolVLA_ur10e_no_joints_crop_pick_place

The top camera image has been cropped to 720x720 and the observations only contain the cartesian pose of the endeffector.

The action space is:

"linear_x.vel",
"linear_y.vel",
"linear_z.vel",
"angular_x.vel",
"angular_y.vel",
"angular_z.vel",
"gripper.pos"