Inverse kinematics with Vortex and Newton

Typically: "How do I... ", "How can I... " questions
Post Reply
Jendker
Posts: 2
Joined: 23 Jan 2019, 15:32

Inverse kinematics with Vortex and Newton

Post by Jendker » 08 Feb 2019, 16:40

Hello everyone, this is my first post on this forum.

I am working on the dynamic hand model for grasping imitation, so the precision of the simulation is important.

I created the hand model myself from Collada model. I have created the convex dynamic responsive shapes for each of the visible part, calculated the mass & inertia in dynamic properties it by setting the density to 2700 (value for aluminium).

Currently I am setting up the inverse kinematics (in hybrid operations mode) and everything works pretty fine with Newton engine, but with Vortex either the target is not fully reached or sometimes the model collapses into pieces and then is "reassembled" during simulation.

Vortex simulation (you can see target dummies)
https://www.youtube.com/watch?v=K67A705FmZQ

Newton simulation
https://www.youtube.com/watch?v=TeoRvR91dJg

So far I had impression, that Vortex engine is more accurate than Newton, for example after trying out motionPlanningAndGraspingDemo.ttt, so I am struggling to set up the simulation in proper way, so that the Vortex engine works properly with fluent motion. I am wondering if I am doing something wrong so that (as I assume) better engine does not work as it should. I would rather use Vortex if it is more precise, maybe Newton avoids some important parameters here which are just not set properly...

With Newton torque of 25 Nm in each joints is enough, but to achieve the current results with Vortex I had to set the max torque on last joint to 100 Nm to make it reach the target... Otherwise two last joints don't bend further than just couple of degrees. I observed, that if I set the torque in each of the finger joint to very high value like 200Nm makes the motion with Vortex much faster, but such values are unreasonable for a finger of weight of ~100g.

Other things which I tried out:
- increasing P parameter in PID controller - the simulation became very unstable, finger started falling apart constantly.
- setting Max. iterations for IK to very high values, no big difference
- changing to explicit handling and execution of calculations for longer time as proposed here: viewtopic.php?f=9&t=6861&p=27281 but the kinematics behaved in the same way as with automatic execution

Is there anything I may be missing here or is it just the way Vortex engine actually works?

coppelia
Site Admin
Posts: 6952
Joined: 14 Dec 2012, 00:25

Re: Inverse kinematics with Vortex and Newton

Post by coppelia » 12 Feb 2019, 09:24

Hello,

I am not sure what is happening with the Vortex engine... this does not look right. Are you sure that there are no internal collisions, or something similar? What happens if you adjust the Upper velocity limit up or down? And if you adjust the joint max. step size? Or if you use a damped resolution method for the Ik?

Maybe you can share the scene so that we can have a closer look at it?

Cheers

Post Reply