ODE Joints don't support friction parameter

[osrf-migration]

Original report (archived issue) by Steve Peters (Bitbucket: Steven Peters, GitHub: scpeters).

---

At ODEJoint.cc line 113, the ODE joint friction parameter is ignored. There is a warning, but we should implement this.

[osrf-migration]

Original comment by Nate Koenig (Bitbucket: Nathan Koenig).

---

• changed state from "new" to "on hold"

Until 2.0

[osrf-migration]

Original comment by Ian Chen (Bitbucket: Ian Chen, GitHub: iche033).

---

• changed state from "on hold" to "open"

[osrf-migration]

Original comment by Steve Peters (Bitbucket: Steven Peters, GitHub: scpeters).

---

• set component to "physics::ODEPhysics"

[osrf-migration]

Original comment by Fadri Furrer (Bitbucket: ffurrer).

---

Is there already a plan to implement this?

[osrf-migration]

Original comment by Steve Peters (Bitbucket: Steven Peters, GitHub: scpeters).

---

It's not on our short-term list; please vote for the issue if it's important to you.

@skuinder You might be interested in this one too.

[osrf-migration]

Original comment by Jonathan Bohren (Bitbucket: jbohren).

---

Is this still unsupported? It's really desirable.

[osrf-migration]

Original comment by Steve Peters (Bitbucket: Steven Peters, GitHub: scpeters).

---

I am going to start on this today. I will post an outline of my approach, write a test, and then start implementing it. Please feel free to give feedback at any point if you're interested.

[osrf-migration]

Original comment by Steve Peters (Bitbucket: Steven Peters, GitHub: scpeters).

---

ODE and Bullet have "joint motors" which will try to force the joint velocity to match a specified value, with a limited maximum force (see ODE user guide, dParamVel and dParamFMax).

The ODEJoint class currently uses this functionality for the Joint::SetVelocity API, though Bullet does not. I would propose to implement ODEJoint::SetVelocity in a different way, then implement joint friction by setting dParamVel = 0 and dParamFMax = friction.

[osrf-migration]

Original comment by Jonathan Bohren (Bitbucket: jbohren).

---

@scpeters I would propose to implement ODEJoint::SetVelocity in a different way, then implement joint friction by setting dParamVel = 0 and dParamFMax = friction.

That sounds reasonable, I'd be happy to try it out, though I probably wouldn't be able to until October.

[osrf-migration]

Original comment by Silvio Traversaro (Bitbucket: traversaro).

---

@scpeters What you describe seems to implement the Coulomb friction, while in URDF [2] and SDF [3] documentation the friction joint attribute is described as :

The physical static friction value of the joint.

What do you think?

Given that terminology on Coulomb/static friction can be inconsistent across literature I am using the terminology proposed in [1], see figure 1 at beginning of page 3 for a clear explanation.

[1] : http://robotics.tch.harvard.edu/publications/pdfs/armstrong1994survey.pdf

[2] : http://wiki.ros.org/urdf/XML/joint

[3] : http://osrf-distributions.s3.amazonaws.com/sdformat/api/dev.html#friction384

[osrf-migration]

Original comment by Steve Peters (Bitbucket: Steven Peters, GitHub: scpeters).

---

@traversaro Thanks for pointing out the imprecision. Looking at Fig 1b in [1], I think it would require more than one parameter to describe static + Coulomb friction. With only one parameter, I would choose Coulomb friction. I will update the sdformat documentation accordingly.

[osrf-migration]

Original comment by Silvio Traversaro (Bitbucket: traversaro).

---

@scpeters It makes sense, the Coulomb + Viscous model is much easier for doing simulation and also friction estimation/compensation in real world.

If only the SDF definition is changed the mismatch between URDF/SDF should be clearly stated. By the way it seems that some users are already changing the URDF friction parameter semantic meaning to Coulomb friction coefficient, see for example [1].

[1] : https://github.com/RobotLocomotion/drake/blob/master/doc/drakeURDF.xsd#L539

[osrf-migration]

Original comment by Steve Peters (Bitbucket: Steven Peters, GitHub: scpeters).

---

• set assignee_account_id to "557058:5de38267-b118-494c-aa76-4fab35448816"
• set assignee to "scpeters (Bitbucket: scpeters, GitHub: scpeters)"

[osrf-migration]

Original comment by Steve Peters (Bitbucket: Steven Peters, GitHub: scpeters).

---

Created #1275 to track the implementation details for simbody.

[osrf-migration]

Original comment by Steve Peters (Bitbucket: Steven Peters, GitHub: scpeters).

---

First pass at implementation for ODE and a test is in 32def69 (branch issue_381). The test fails for bullet and dart, and is disabled for simbody.

This breaks ODEJoint::SetVelocity, so that will have to be fixed before this can be merged to a release branch.

[osrf-migration]

Original comment by Steve Peters (Bitbucket: Steven Peters, GitHub: scpeters).

---

Created #1281 to track the implementation details for dart.

[osrf-migration]

Original comment by Steve Peters (Bitbucket: Steven Peters, GitHub: scpeters).

---

see pull request #1206 for the initial ODE implementation

[osrf-migration]

Original comment by Steve Peters (Bitbucket: Steven Peters, GitHub: scpeters).

---

resubmitted as pull request #1221

[osrf-migration]

Original comment by Nate Koenig (Bitbucket: Nathan Koenig).

---

• changed state from "open" to "resolved"

Resolved in pull request #1221

[osrf-migration]

Original comment by Nate Koenig (Bitbucket: Nathan Koenig).

---

• set version to "all"

[osrf-migration]

Original comment by Nate Koenig (Bitbucket: Nathan Koenig).

---

• changed state from "resolved" to "closed"