geo-fixability — When is a geophysical map good enough to navigate on?
Geophysical map matching works — sometimes. The open question is when. This project builds a predictive model relating vehicle trajectory characteristics and geophysical map properties to achievable navigation accuracy, so the question becomes answerable before you fly the mission rather than after: if I traverse this area at this speed and heading, how much will my navigation solution actually improve?
The approach:
- Generate synthetic scalar fields — terrain, magnetic, gravity — with controllable and classifiable information content.
- Simulate INS trajectories across a range of speed and heading profiles.
- Estimate with a Rao-Blackwellized particle filter aided by the anomaly field.
- Compute theoretical bounds — the Fisher information matrix and the posterior Cramér-Rao bound — and measure filter efficiency against them.
- Extract the trajectory and map features that predict performance, and regress (trajectory, map) → uncertainty reduction.
- Validate against real data: SRTM15+, WDMAM, KITTI.
The useful output is a fixability metric that behaves like a sensor specification: given these trajectory parameters over this map, I know my position to within some bound. That turns geophysical aiding from something you try into something you can budget for.
The resulting paper is under review at IEEE Transactions on Aerospace and Electronic Systems. The work was supported by a $50,000 Penn State Applied Research Laboratory internal research and development grant, for which I was principal investigator.