ARISE-Initiative / robosuite

robosuite: A Modular Simulation Framework and Benchmark for Robot Learning
https://robosuite.ai
Other
1.24k stars 397 forks source link

Whether the model reduce accuracy in robosuite? #489

Closed super-dashuaibi closed 3 months ago

super-dashuaibi commented 3 months ago

Excuse me,actually I don't know whether I should ask this question here or on mujoco.

I was assembling the peg in hole based on robosuite, and then I found that the peg could not enter the hole. Then I modified robosuite/demos/demo_gripper_interaction.py to verify if there was a problem with my model(peg and hole model), and I found that the peg could not enter the hole.

I found that my model became much rougher during simulation interaction, so I wondered if the model accuracy was reduced during the import process, causing the peg to not fit into the hole.

2024-06-03 16-38-03 的屏幕截图 2024-06-03 16-38-12 的屏幕截图

And when I used mujoco210 to simulate my model directly, and it looked perfect. When designing, the gap between the peg and the hole is 2mm. I am not sure if the model gap is too small.

2024-06-03 16-24-25 的屏幕截图 2024-06-03 16-39-48 的屏幕截图

super-dashuaibi commented 3 months ago

oh, i understand it.it is a setting in mujoco.

(https://mujoco.readthedocs.io/en/stable/computation/index.html)Physical realism and soft contacts Many of the advantages of our contact model can be traced to the fact that we drop the strict complementarity constraint at the heart of the LCP formulation. We will call this family of models convex; see References for related work. For frictionless contacts dropping the explicit complementarity constraint makes no difference, because the Karush-Kuhn-Tucker (KKT) optimality conditions for the resulting convex quadratic program are equivalent to an LCP. But for frictional contacts there are differences.

we can cd ~/.mujoco/mujocoXXX/bin/ ./simulate YOUR_MODEL.xml and click the Rendeing ,then click convex hull to see the collsion model