Volumetric Occupancy Mapping With Probabilistic Depth Completion for Robotic Navigation

In robotic applications, a key requirement for safe and efficient motion planning is the ability to map obstacle-free space in unknown, cluttered 3D environments. However, commodity-grade RGB-D cameras commonly used sensing fail register valid depth values on shiny, glossy, bright, or distant surfaces, leading missing data map. To address this issue, we propose framework leveraging probabilistic completion as an additional input spatial mapping. We introduce deep learning architecture providing uncertainty estimates of images. Our pipeline exploits inferred complement raw images improve speed quality free Evaluations synthetic show that our approach maps significantly more correct with relatively low error when compared against using alone different indoor environments; thereby producing complete can be directly navigation tasks. The performance validated real-world data.