FollowTarget example - calculation cost increases when target position is out of workspace

Hello, I’m trying to apply follow target examples to shadowhand.

IK was solved by applying the RMPflow descriptor only for the index finger, and it seems to work well. However, the problem is that when the target position goes out of the workspace, the FPS drops rapidly.

"standalone_examples/api/omni.isaac.franka/ " always keeps above 140 fps, but my hand example shows 30~90 fps.

One thing I suspect is that Hand’s structure is more complex than Franka’s.

Currently, I wrote a descriptor only for the index finger. That is, only the leftmost tree in the figure is solved the ik.


api_version: 1.0
    - rh_FFJ4
    - rh_FFJ3
    - rh_FFJ2
    - rh_FFJ1
root_link: world

default_q: [
    0.00, 0.00, 0.00, 0.00
cspace_to_urdf_rules: []
composite_task_spaces: []

Q1. Is there a way to ensure that my example always maintains the best fps?
Q2. Is there any example that uses hand robot?


Is it good options to change some parameters on LulaKinematicsSolver?

for example, “set_max_iterations()” or “set_default_cspace_seeds()”, etc…