Understanding difference between qfrc_applied and xfrc_applied

[Hardwarize]

Intro

Hi!

I am a master student at UNED, I use MuJoCo for my research on generative human motion.

My setup

Mujoco python

My question

I'm very noob in the robotic and simulator field. I am not able to understand the difference between qfrc_applied and xfrc_applied. Reading in other places and watching some videos I understand that qfrc_applied injects torques/forces directly into the joint space, while xfrc_applied applies forces and torques in the 3D (Cartesian) space of each body. If you have a free joint both are not basically the same? and if you have a hinge join and you want to aplly torque with an actuator, I've seen that you should use qfrc_applied but don't you can get the same result with xfrc_applied?

Thanks in advanced

Minimal model and/or code that explain my question

No response

Confirmations

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

[Balint-H]

EDIT: Thanks for @preist-nvidia for pointing out a mistake in this answer, I edited to remove the error.

You are correct as far as a freejoint goes, the first three components qfrc ends up being applied as forces in the global space. The remaining three components for freejoints is interpreted in the body frame (i.e. a qfrc that makes a robot frontflip will usually still do that if the robot faces a different global direction.)

However, for all other degrees of freedom, qfrc will be applied along the current direction of the joint axes, whereas xfrc_applied will always stay in the global frame.

E.g. if you have a motorbike model where the wheel axes initially are parallel to the global x axis, applying torque via the corresponding element in the xfrc or qfrc vectors on the wheels will both result in the bike accelerating forward. But after the bike makes a 90 degree turn, qfrc would still drive it normally towards the bike's new forward direction, while xfrc would tip the bike over and tumble it down along the original forward direction.

An example case where xfrc representation is useful is the "mouse spring" perturbation vector in the interactive viewer. You can use it to intuitively pick up, rotate, and place objects in the scene. The forces necessary to do that are easiest to interpret in the global space.

'xfrc_applied' ends up being transformed to qfrc in the engine, and summed up with other sources of qfrc:

mujoco/src/engine/engine_core_smooth.c

Line 2032 in d4ca66a

mju_transformSpatial(cfrc_com, cfrc, 1, d->subtree_com+3*m->body_rootid[i], d->xipos+3*i, 0);

Depending on what you are trying to achieve, either qfrc or xfrc can be more appropriate. I'll close the question but feel free to reopen if something is still unclear, or if I got something wrong.

[preist-nvidia]

Because I just tripped over this, I think this statement is incorrect:

You are correct as far as a freejoint goes, qfrc ends up being applied as forces and torques in the global space.

From this: #691 (comment)

it follows that while the force component of the free joint qfrc is in global frame, the torque component is in body frame.

[Balint-H]

Thanks for pointing out the mistake, updated the answer!

[preist-nvidia]

Awesome thank you Balint!