Isaac Sim Version
5.1.0
Operating System
Ubuntu 22.04
GPU Information
- Model: NVIDIA GeForce RTX 5060
- Driver Version: 580.126.09
Topic Description
Detailed Description
I am using Isaac Sim 5.1.0 to generate a synthetic dataset for soft body deformation research. My goal is to simulate a rigid sphere probe pressing into a deformable cube to capture the non-linear Force-Displacement (f-d) relationship caused by changing contact area (Hertzian contact) or shear force. However, the f-d curves are always linear after many turns of adjustments. Is my configuration wrong or it is just a PhysX 5 bug?
Steps to Reproduce
-
Create a Deformable Body cube using deformableUtils.add_physx_deformable_body.
-
Set simulation_hexahedral_resolution to 50 and collision_simplification=True.
-
Create a rigid sphere and attach it to a kinematic base via a D6 Joint (Spring drive) to act as a force sensor.
-
Move the sphere into the cube and record the displacement vs. the spring force.
-
Change collision_simplification to False and run the simulation again.
Screenshots or Videos
(If applicable, add screenshots or links to videos that demonstrate the issue)
Additional Information
What I’ve Tried
- I adjusted the sensor spring stiffness to ensure the linearity was not caused by the sensor deformation itself.
- I adjusted the mesh resolution and increased simulation substepping, but the non-linear force response did not improve.
- I modified the contactOffset and restOffset in the PhysxDeformableBodyAPI to mitigate the tunneling issue when simplification was disabled.
Related Issues
- I searched existing forums and found that tunneling after disabling collision_simplification seems a known stability issue with no clear fix.
- I compared results with other FEA solvers where Hertzian contact is a standard feature, highlighting the limitation in the default Isaac Sim setup.
Additional Context
- The experimental setup uses a kinematically controlled rigid sphere to indent a 1x1x1 hexahedral deformable body.
- This simulation is intended to generate high-fidelity synthetic datasets for research, making the physical accuracy of the f-d curve critical.
