Difficulty Achieving Elastic Contact with Constant Energy in Robot Jump Simulation

[caba0404]

Intro

Hi!

I am a graduate student at UCI, I use MuJoCo for my research.

My setup

I'm using Python to run the MuJoCo simulations.

My question

I'm reaching out because I'm experiencing an issue with contact dynamics in my simulation. My robot needs to jump on the floor, but I'm having trouble achieving a clean, elastic-like contact. Specifically, I'm trying to get a jump where the energy during contact is fully conserved — in other words, a perfectly elastic interaction.

The jump occurs at around 0.15 seconds into the simulation.

When I use the following solref values:
solref = [2e-5, 1]
I get these results:

However, when I change the values to:
solref = [-2e-5, 0]

I observe a different behavior:

I’d appreciate any guidance on how to tune the contact parameters to achieve constant energy during impacts, or if there's a better approach to simulate perfectly elastic contact.

Minimal model and/or code that explain my question

Confirmations

• I searched the latest documentation thoroughly before posting.
• I searched previous Issues and Discussions, I am certain this has not been raised before.

[yuvaltassa]

You need to use the negative semantics of solref, as explained here https://mujoco.readthedocs.io/en/latest/modeling.html#reference

[caba0404]

Yes, that’s exactly what I’m doing. There was a typo in my previous message — I’m using solref values like [-1000, 0]. However, when I use these negative values, the simulation becomes unstable. On the other hand, using positive values causes a significant energy loss — around 40% — when the robot touches the ground without damping. This is a major issue, and I’m not sure how to fix it.

I’m facing a similar problem with joint limits, especially with hinge joints. When I define an angular limit, the joint doesn’t block movement in the restricted direction upon reaching the limit. I’ve tried adjusting the solref values as suggested in the documentation, but I still see significant violations of the limits. I’m not sure if I’m missing something or if there's another parameter I need to tune to solve this.

[yuvaltassa]

1. Unstable how? Can you make an MRE? E.g. a bouncing sphere on a plane.
2. Limits are never hard, you can never get perfect joint locking. However low solref[0] (positive semantics) and small timestep can get you as close as you want to rigid limits.