Unable to use GPU while running Isaac SDK application on Jetson AGX (CPU Maxxed out)

The CPU usage while running an Isaac application on a robot using Jetson Xavier AGX is really high on occasions which freezes the Xavier momentarily. Are there any configurations that forces the Isaac application uses the GPU as well?
My current system specifications

Board type: Jetson Xavier AGX
Jetpack: 4.4.1
CUDA: 10.2.89

Does an isaac application use GPU in run time on a Jetson Xavier AGX?
jtop stats shows the GPU usage to be zero consistently. Does having a powerful platform like Xavier AGX provide any performance enhancement in navigation applications?

@hemals @saralopez @AastaLLL @dusty_nv

Isaac SDK codelets can leverage the GPU as a resource depending on how they are implemented. Could you describe the codelets you’re using in your application graph?

@hemals My code repo also has a custom behavior tree where the isaac navigation gem is one of the NodeGroups being used. The application primarily runs the packages/navigation/apps/differential_base_navigation.subgraph.json with a number of custom sensor collection, processing modules.
So the codelets being run are planner, localization, control and sight widgets and everything else needed for a waypoint navigation task.

Navigation remains mostly CPU bound unfortunately. The localization codelet will leverage GPU for scan matching, but not planner, control, and sight for the most part.

@hemals With a collection of lidars, its data processing nodes the Xavier AGX freezes during waypoint navigation.
Are there any suggestions to restrict max CPU usage with a compromise to performance?