Compressed Data Streaming and Pose Estimation via Oct-Tree Data Compression and Global Iterative Point Cloud (GO-ICP) Registration

UAS platforms are required to operate within urban canyons and other GPS-denied/degraded environments. Navigation within such denied environments carries with it the potential for large communication and computation overheads. More specifically, the sUAS platforms may need to download full or partial 3D maps over the communication network in real-time, and then utilize these maps for navigation via coregistering real-time sensor data with the pre-existing 3D georeferenced maps downloaded over the network. The key challenge is finding a data structure for storing, streaming, and processing 3D data sets that successfully reduces the burden on the communication network, while still enabling the data to be processed in real-time on-board the sUAS platform. OKSI proposes leveraging Octree data structures which frequently provide around 99.8% data compression in comparison to standard 3D voxel grids.