Vilitus Engine- by Travis Clark

Omniverse Isaac Sim
1

Hi NVIDIA team,

I’m working on Vilitus, a spatial grounding layer for embodied AI and robotics. The current design turns raw sensor input into a scene graph, persistent spatial anchors, and ontology layers for material, physics, and affordance context, with a ROS-oriented bridge and a core runtime based on confidence, geometric divergence, and semantic coherence.

I’m trying to map this cleanly onto Isaac Lab / Isaac Sim workflows. My main question is: where would a spatial grounding / world-model layer like this fit best in the Isaac Lab stack, and what integration pattern would you recommend for robot learning, sim-to-real, or synthetic data workflows?

If useful, I can share a strict manifest separating verified, inferred, and unverified pieces so the scope stays clear.

Vilitus_Engine__The_Evolution_of_Spatial_Intelligence.pdf (898.6 KB)

image
image2304×1728 443 KB

Thanks.

3

Hi @travislc179,

Thanks for sharing your work on Vilitus! Your spatial grounding layer concept sounds interesting for embodied AI and robotics workflows.

For support and integration questions about Isaac Lab, please redirect to the Isaac Lab GitHub repository: