search

google-deepmind / mujoco_menagerie

A collection of high-quality models for the MuJoCo physics engine, curated by Google DeepMind.
Other
1.45k stars 202 forks source link

Model collisions #99

Closed Narsimha-swamy closed 3 weeks ago

Narsimha-swamy commented 1 month ago

Which model is the issue affecting? ALOHA2 What is the issue? I am using Aloha 2 models for real-to-sim teleoperation, and I am able to send data from real robots to robots in the sim and control them, but collisions between objects seem to be off as robots are passing through objects in the scene. Is there a way to know if collisions or contact physics, in general, were implemented in the ALOHA2 XML files?. I uploaded a video of the simulation, can you let me know what seems to be the problem.

https://github.com/user-attachments/assets/c31de7ff-7134-42fe-858d-80cd26da2efc

Is there any additional context you can provide (e.g., a spec sheet or a URDF to show a value mismatch)?

btaba commented 1 month ago

Hi @Narsimha-swamy , the video is empty. AFAIK the base Aloha XML files have good defaults, we have been using these files for some time. I would expect the objects you are adding need their physics parameters tuned. Remember to turn on MultiCCD.

Narsimha-swamy commented 1 month ago

For some reason the videos I upload are not showing, I am uploading a drive link in case.

btaba commented 1 month ago

Which scene are you loading? How to repro?

Narsimha-swamy commented 1 month ago

Hi, I included the scene.xml from aloha into new scene and added a cube to it <mujoco> <include file="scene.xml"/> <worldbody> <body name="box" pos="0 0 0.02"> <joint name="box_joint" type="free" /> <inertial pos="0 0 0" mass="0.05" diaginertia="0.002 0.002 0.002" /> <geom condim="4" solimp="2 1 0.01" solref="0.01 1" friction="1 0.005 0.0001" pos="0 0 0" size="0.02 0.02 0.02" type="box" name="red_box" rgba="1 0 0 1" /> </body> </worldbody>

<option gravity="0 0 -9.8" timestep="0.02" /> </mujoco>

btaba commented 3 weeks ago

@Narsimha-swamy please lower your timestep to say 0.002. 0.02 is too high

Narsimha-swamy commented 3 weeks ago

Hi, I have already tried this with timestep = 0.002, but it still gives out similar results.

btaba commented 2 weeks ago

@Narsimha-swamy I can't reproduce your issue, send over a full XML example if you would like us to take a second look

Narsimha-swamy commented 2 weeks ago

Hi, I made a git repo for the same, please let me know if you find anything wrong with the implementation.