[Question] How to make a rigid object hard to move?

[Soappyooo]

Question

I am working on an RL curriculum that make object static in the beginning and gradually movable as the training goes. My current method is:

1. make object kinematic_enabled=True at first
2. make it kinematic_enabled=False, and set large linear_damping and angular_damping (e.g., 1000)
3. gradually reduce the damping value to zero.

However, I noticed that linear_damping and angular_damping cannot stop my robot from moving the object. I also tried setting small max_linear_velocity and max_angular_velocity, but none of them works.

Actually, the properties above seem to work well when the object is not in contact with another object (e.g., my robot). In Isaac Sim, if I shift+drag the mouse on the object, it can barely move. But if I shift+drag the mouse on another object in contact with it, it can still be moved. Please check the video and usd file below.

A workaround I found is adding a D6 joint to the object, with low limit and high limit all set to 0 for all DoFs, and set a large damping. But I'm not sure if this is a proper solution.

Build Info

• Isaac Lab Version: 2.3.0
• Isaac Sim Version: 5.1.0

Videos

Setting linear_damping, angular_damping, max_linear_velocity and max_angular_velocity. Further increase damping doesnt make much difference:

1.mp4

Adding a D6 joint with zero limit and large damping, better than above:

2.mp4

USD File

movable_object.zip

[RandomOakForest]

Thank you for posting this. What you see is expected: rigid body damping and velocity limits act on the object’s own state, but they do not prevent contact impulses or joint constraints from moving it, so they are not sufficient to “lock” an object under contact.¹ I'll move this post to our Discussions section for follow up. Here is a summary to consider.

Why damping and velocity limits fail here

• Linear/Angular damping apply forces proportional to velocity and only slow the object down; they do not constrain its pose or block impulses from contacts or joints.¹
• Contacts are solved as constraints; the solver enforces non‑penetration and friction first, then applies damping as an extra force, so a strong robot can still push the object even with very large damping.²¹
• max_linear_velocity and max_angular_velocity are clamps on the object’s own velocity, but the constraint solver can still move it to satisfy contact constraints, so they do not behave like a “hard lock” either.²¹

Proper ways to make an object “hard to move”

In Isaac Lab / PhysX, there are basically three robust strategies for your curriculum:

1. Use kinematic → dynamic transition

• Start: object with kinematic_enabled=True so it is fully controlled and unaffected by forces.³⁴
• Later stages: switch to kinematic_enabled=False (dynamic) when you actually want the robot to be able to move it; do not rely on extreme damping for an intermediate “almost static” state, because PhysX is not designed to treat damping as a hard constraint.¹²

2. Use joints with locked or stiff DOFs (your D6 trick)

• Adding a D6 joint between the object and the world (or a fixed base) with all DOFs locked (low limit = high limit = 0) is a standard way to fix a rigid body in place while keeping it dynamic.⁵⁶
• You can then:
  • Early curriculum: lock all 6 DOFs (the object behaves effectively fixed).⁶⁵
  • Intermediate stages: unlock or “limit + high drive stiffness/damping” only along the DOFs you want the robot to gradually influence (e.g., allow small translation or rotation with large drive damping).⁶²
  • Final stage: remove or fully loosen the joint limits so the object becomes freely movable.²⁶
• This is not just a hack; constrained DOFs with strong drive damping is a recommended way in PhysX to control how easily an articulation or rigid body moves, since joint drives and limits are handled directly in the constraint solver.⁶²

3. Temporarily increase effective “stiffness” via mass / inertia

• Another curriculum knob is to start with a large mass and/or inertia (e.g., in RigidBodyPropertiesCfg), so the same robot actions produce much smaller accelerations, then gradually reduce mass toward the real value.⁷³
• This does not prevent movement completely, but it can make “accidental” motion negligible in early phases while still being physically consistent.¹²

Recommended curriculum pattern for Isaac Lab

A practical pattern that tends to behave well in RL:

• Phase 0 (fully fixed):
  • Object: kinematic_enabled=True, no joint to world. Robot cannot move it at all.⁴³
• Phase 1 (locked but dynamic):
  • Switch object to dynamic (kinematic_enabled=False).
  • Attach a D6 joint to a world frame (or static base) with all DOFs locked (low=high=0).⁵⁶
  • Optionally add high joint drive damping/stiffness, but DOF locking alone should prevent motion.²⁶
• Phase 2 (very limited motion):
  • For selected DOFs (e.g., small translation or rotation), change from locked to limited with tight low/high limits and large drive damping.⁶²
  • Keep others locked. The robot can only cause tiny compliant motion.
• Phase 3 (heavy but free):
  • Remove tight limits (or relax them a lot) but keep mass higher than final to make motion “heavy”.³⁷
• Phase 4 (final task):
  • Realistic mass/inertia, no joint to world (or very loose limits), normal damping values.⁷³

Implementation tips

• In Isaac Lab, configure body properties via RigidBodyPropertiesCfg (linear_damping, angular_damping, mass, etc.) and joint properties via joint drive configs; use these as per the latest Isaac Lab API docs.⁴³⁷
• Avoid using extreme damping (e.g., 1000) as the main mechanism for “locking” motion, because it can hurt numerical stability and still does not block contact-driven movement.¹²
• Your D6‑joint approach is appropriate; the key is to treat it explicitly as the curriculum mechanism (progressively unlocking DOFs and relaxing limits), rather than relying on body damping to emulate a constraint.⁵²⁶

Footnotes

1. https://documentation.help/nvidia-physx-sdk-guide/RigidDynamics.html ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷

2. https://nvidia-omniverse.github.io/PhysX/physx/5.1.3/docs/Articulations.html ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰ ↩¹¹

3. https://isaac-sim.github.io/IsaacLab/main/source/api/lab/isaaclab.sim.schemas.html ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶

4. https://isaac-sim.github.io/IsaacLab/main/_modules/isaaclab/sim/schemas/schemas.html ↩ ↩² ↩³

5. https://www.youtube.com/watch?v=uZwQe_9lsuk ↩ ↩² ↩³ ↩⁴

6. https://docs.nvidia.com/gameworks/content/gameworkslibrary/physx/guide/Manual/Joints.html ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹

7. https://isaac-sim.github.io/IsaacLab/main/source/how-to/make_fixed_prim.html ↩ ↩² ↩³ ↩⁴

[Soappyooo]

Thank you!