ARISE-Initiative / robosuite

robosuite: A Modular Simulation Framework and Benchmark for Robot Learning
https://robosuite.ai
Other
1.28k stars 410 forks source link

About the IK controller's Inputs #492

Closed super-dashuaibi closed 1 month ago

super-dashuaibi commented 3 months ago

Hello, I'm using the IK controller recently and I have a lot of questions. I wish you can help me out in your busy schedule. The question I'm most concerned about is which coordinate system the input pose increment (x_dpos, y_dpos, z_dpos, x_rot, y_rot, z_rot) is relative to. In the document,it says," Control input actions are assumed to be relative to the current position / orientation of the end effector and are taken as the array (x_dpos, y_dpos, z_dpos, x_rot, y_rot, z_rot). "

But I've run it in demo_controller.py, and my conclusion is that the (x_dpos, y_dpos, z_dpos) is relative to the base coordinate system under mujoco, and the (x_rot, y_rot, z_rot) is relative to the coordinate system of the end flange of the robot arm. I'm confused why this is relative to two coordinate systems, and why it's not relative to the real eef coordinate system of the end tool.

sarthakdas commented 3 months ago

I agree very confusing

kevin-thankyou-lin commented 2 months ago

I believe the authors implemented the inputs like so to make keyboard (and, spacemouse) teleop more intuitive. That said, it could indeed be less confusing if teleop-specific considerations are decoupled from the IK controller's input convention.