During February 2025, lllllcf advanced robot perception for the h2r/GHOST repository by building a multi-robot depth completion feature that fuses depth and color data from multiple cameras into a unified 3D map. Leveraging C#, HLSL, and Unity’s Sentis library, lllllcf integrated ONNX-based AI models into the perception pipeline, enabling real-time depth estimation from diverse viewpoints. The solution established robust data flow and processing for cross-robot sensor fusion, enhancing obstacle detection and collaborative navigation. The work demonstrated technical depth in 3D rendering, AI/ML, and computer vision, resulting in a reproducible, traceable codebase that supports safer, more intelligent robot fleets.