Is it possible to correctly simulate a crouching motion to reduce energy consumption?

[H-Hisamichi]

Hello everyone,

I have a question. Is it possible to simulate a phenomenon where the robot's self-weight is supported directly by its body parts when a humanoid robot crouches in Isaac Sim?
For walking robots, energy consumption at rest is a big challenge. Therefore, I thought about a posture that supports self-weight without using joint torque by crouching.

Will this kind of motion be simulated correctly in Isaac Sim?
Thank you!

[PeterL-NV]

Yes, you can simulate a humanoid that passively supports its own weight in crouched poses in Isaac Sim/Isaac Lab, but you must rely on rigid-body contacts, joint torques, and controller design rather than any special “self‑weight support” feature.

In Isaac Sim/Isaac Lab, the robot is a rigid‑body articulation; gravity acts on each link, and joint torques plus contact forces must balance those loads. When a humanoid crouches, any body part with a collider (feet, shins, thighs, hands, etc.) can legally contact the ground or other objects and share the load through those contacts, just as in standing or walking. So if your humanoid’s thighs or hands touch the ground while crouching and you model contact properly (colliders + material properties), the physics engine will route part of the weight through those contacts automatically, and joints see reduced torque where leverage allows it.

[H-Hisamichi]

Hello @PeterL-NV,

Thank you for the advice.
I have a follow-up question.

Based on your explanation, I understand that the actor parts need to make contact in the simulation. In this case, would it be inappropriate to set the joint limits exactly at the contact point (i.e., limit = contact)?
I am simplifying the robot model to reduce the simulation load. Therefore, I have set the joint limit values to the angles where the actual robot arms make contact.

[PeterL-NV]

Setting joint limits exactly at the “self‑contact” angle is possible, but it is usually not the best way to model that contact in Isaac Sim / Isaac Lab.

Why using limits as contact is risky

Joint limits in PhysX are implemented as constraint “collisions” with a tolerance (contactDistance), and enforcement starts before the numeric limit is hit.

If you put the limit exactly at the real contact angle, the solver may:

• Engage the limit slightly early or late depending on velocities and tolerances.
• Fight with geometric contact (if you also have colliders there), sometimes causing jitter or extra constraint load.
• Produce behavior that does not match the real robot if the real joint can slightly go beyond that angle but is stopped by link–link contact, not the actuator hard stop.

For humanoids trained or controlled near their limits, starting in or operating close to joint limits can also lead to instability or solver stress.

When is “limit = contact angle” acceptable?

It can be acceptable if:

• You intentionally disable self‑collisions between those links and rely solely on the limit to approximate that stop.
• You are running a coarse, low‑cost sim where you only care about not exceeding a motion range, not the detailed contact forces or resting load distribution.
• You keep a small safety margin and tolerance (e.g., set the limit slightly inside the real contact angle and accept that the sim is conservative).

For energy‑at‑rest studies in crouched postures, I would recommend using collisions for weight‑supporting contacts (thigh–torso, arm–thigh, etc.) and keep joint limits slightly outside those angles, so that the load is carried by contacts rather than by limit constraints.

[H-Hisamichi]

Hello @PeterL-NV,

Thank you for the detailed advice!
I will modify my robot model based on your suggestions.