AeroMap3D uses synthetic data to push UAV-to-map localization to 99.2% success
ducha_aiki · x · 2026-07-23
AeroMap3D proposes a monocular UAV 6-DoF localization pipeline anchored to visual-geometric-semantic map priors.
- It uses satellite/OSM/DEM priors, semantic rejection, RANSAC-based PnP, and an EKF that fuses map updates with relative odometry.
- The authors say synthetic training data alone can train the scale/yaw adapter, avoiding real UAV labels.
- In cross-domain registration, the adapter boosts Tiny-RoMa from 42.1% to 92.5% success and RoMaV2 from 62.4% to 99.2%.
- Runtime tests show the method can run on Jetson Orin Nano Super, while RoMaV2 exceeds the edge-device memory budget.
More from Embodied
- XPeng CEO says cars will increasingly become four-wheeled robots over the next decade — pstAsiatech · 2026-07-23
- Guangdong cookware plant runs lights-out with robots after replacing about 100 workers — heyshrutimishra · 2026-07-23
- Robotaxi sentiment has swung from premature optimism to premature dismissal — skorusARK · 2026-07-23
- Agentic Real2Sim rebuilds a physics simulator from one real-world video — anselm · 2026-07-23
- IIHS says Waymo’s driverless cars crash 68% less often than human drivers — SuzKP · 2026-07-23
- Musk says Tesla Semi self-driving should arrive by late this year or early 2026 — XFreeze · 2026-07-23