Drake: RigidBodyTree::inverseDynamics bug when acceleration is nonzero

[tkoolen]

I recently looked over the RigidBodyTree::inverseDynamics code again and found that I had made a mistake that causes the torques returned by inverseDynamics to be wrong when vd (the desired acceleration vector) is nonzero. Luckily we've never used it with nonzero vd in the rest of the Drake codebase, but it still needs to be fixed of course. Just wanted to report it for now.

[liangfok]

Thanks for reporting this. Just wondering what's the logic behind using the name "vd" for "acceleration vector"?

[sherm1]

Wild guess: vd = time derivative of velocity?

[tkoolen]

Yep.

[amcastro-tri]

That's interesting that is never called with non-zero acceleration. What are the uses of this method then?

[tkoolen]

Computing the Coriolis, gravity, and external force terms in the equations
of motion.

On Thu, Apr 28, 2016 at 5:51 PM, Alejandro Castro [email protected]
wrote:

That's interesting that is never called with non-zero acceleration. What
are the uses of this method then?

—
You are receiving this because you authored the thread.
Reply to this email directly or view it on GitHub
https://github.com/RobotLocomotion/drake/issues/2201#issuecomment-215573734

[tkoolen]

I have the time to fix this now and write the corresponding tests. @sherm1, is that OK, or would you prefer to do it yourself? @liangfok, are you actively working on this code?

[liangfok]

I am not working on this at this time. Would be great if you can fix this!

[sherm1]

I have the time to fix this now and write the corresponding tests.

That would be great, Twan!