[Question] How to Set Motor Friction Parameters When Identified from No-Load Condition?​

[FrankTianTT]

I identified my motor parameters by recording the relationship between velocity and torque under speed control in a no-load condition. The relationship follows:

torque = k * vel + b

where both k and b are positive values.

In my understanding, k corresponds to the damping term in MuJoCo, and b corresponds to the frictionloss.

However, according to the IsaacLab Actuators documentation, the joint friction in IsaacLab is computed based on forces transmitted from the parent body. This does not align with my test setup, which was performed under a no-load condition.

Could you please advise on how to properly set the motor friction parameters in IsaacLab in this case? Specifically, how should I map the identified parameters k and b to the available actuator configuration settings?

Thank you for your help!

[RandomOakForest]

Thank you for posting this. This is a good question for our Discussions section. I'll move the post there for follow up. Here are a few notes to consider.

A way to map your empirically identified parameters $k$ (from the linear torque-velocity relationship) and $b$ (offset/intercept) to joint motor friction parameters in Isaac Lab is as follows:

• $k$ corresponds to viscous friction: In Isaac Lab, viscous_friction directly implements a torque proportional to velocity, $\tau_{fric,viscous} = k_{viscous} \cdot \dot{q}$. This is conceptually the same as your identified $k$ parameter from the no-load velocity-torque slope. Set the viscous_friction parameter in the actuator configuration to $k$.¹²
• $b$ corresponds to static/Coulomb friction: The offset $b$ in your identification (the y-intercept) captures constant torque required to overcome static friction. In Isaac Lab, both friction (static friction) and dynamic_friction (Coulomb friction) are available. Set the dynamic_friction parameter to $b$ as it models a velocity-independent resisting torque. If your measured offset includes only the torque needed to initiate motion (i.e., purely static friction), also set the friction parameter to $b$.

Isaac Lab Parameter Mapping

• damping: This is NOT the torque-proportional-to-velocity loss term under most configurations, but rather a stabilizing parameter for the joint (often the D gain in PD control) and should only be set to nonzero if you explicitly want additional dissipation in the controller.
• friction: This is static friction (breakaway torque). Set to $b$ if $b$ was measured as the minimum torque needed to initiate motion. However, for continuous velocity regions, use dynamic_friction instead.
• dynamic_friction: Use this for Coulomb friction, which resists all motion at a nearly constant magnitude independent of speed—set to $b$ from your measurement.
• viscous_friction: This implements a force/torque proportional to velocity, a direct match for your measured $k$ slope.

Footnotes

1. https://isaac-sim.github.io/IsaacLab/main/source/api/lab/isaaclab.actuators.html ↩

2. https://isaac-sim.github.io/IsaacLab/main/source/overview/core-concepts/actuators.html ↩

[FrankTianTT]

Thank you for your response, which has clarified my most pressing question!

Beyond that, I would still like to confirm with you: Are the units for both friction and dynamic_friction Newton-meters? In other words, do friction and dynamic_frictiononly characterize the friction behavior of the motor itself, without being affected by forces transmitted from other rigid bodies?

In the IsaacLab documentation, it states:

The joint static friction is a unitless quantity. It relates the magnitude of the spatial force transmitted from the parent body to the child body to the maximal static friction force that may be applied by the solver to resist the joint motion.

This has made me wonder whether friction and dynamic_friction are coefficients (such as the coefficient μ that determines the frictional force on an object moving on a plane) or actual frictional forces (with units of Newton-meters).