Air / drones & GPS-denied navigation

Problem

GPS is the crutch that most outdoor autonomy leans on. The problem this work is pointed at: what does a drone do when the satellite signal is gone — indoors, under cover, in a contested environment? GPS-denied navigation is the question of how a machine keeps knowing where it is when the easy answer is removed.

Difficulty

Without GPS, the vehicle has to estimate its own state from its own sensors — and every sensor drifts. The difficulty is the same one underwater: the environment does not care how elegant the code looks, and the sky is as unforgiving as the water.

Built

This entry is not yet public — in progress. The work covers drones, avionics, guidance/navigation/control (GNC), and GPS-denied navigation. There is no public repository yet, so the honest description is: a machine being built, not yet shown.

Owned

The GNC and navigation work — the part that has to keep the machine knowing where it is when the easy answers are gone.

Failed

The failures here are the ones that GPS-denied navigation is built to expose: state estimates that drift, sensors that disagree, and the moment the machine has to act on a position it is not sure about. The specifics are not yet public.

Changed

The work is changing what I expect from navigation: that a machine should be able to know where it is without being told. The direction is the same one that runs through the underwater work — perception and state estimation that have to work when the environment stops cooperating.

Machine-now

In progress, not yet public. The machine is being built; when there is something to show, this entry will be updated with the real record.

Lesson

The lesson is still being written. What is already clear: GPS-denied navigation is the air version of the underwater problem — a machine that has to know where it is when the environment refuses to say.