Hydrus, a revolution in underwater robotic technology, was recently sent to the treacherous depths of the Rottnest ship graveyard located in the Indian Ocean, just off the coast of Western Australia. Upon unloading its data, the team was ecstatic to discover Hydrus had spotted a 64-metre shipwreck scattered across the seafloor.
...
Specialising in shipwreck photogrammetry, Curtin University HIVE was able to rebuild a high-resolution replica of the wreck, using the data gathered by Hydrus. This involved taking 4K geo-referenced imagery and video footage from Hydrus to generate a 3D digital twin of the shipwreck.
If you're surveying you ideally want to fix against a base station as often as possible.
In calm waters these are pretty good for 30 minutes and more ... but then there's the wonder of post processing as opposed to "best position guess at the time".
The secret source here is a multitude of inputs; multi axial inertia, magnetics, modern underwater not-LORAN, pivoting on features in data from frame to frame and smushing these all together in post to create an optimal (for some value of optimal) best {position + orientation} trajectory after "the game".
It's "sufficient" to coarsely move about and get coverage and then use everything known (past and future) to sharpen the 'present' position.
This is a seperate case to blasting off into the unknown as a missile and having to get position correct on the fly in a single pass and no do overs.
The back pedigree here is access to decades of geophysical and photogrammic survey - so an arsenal of tricks of the trade.
