SLAM and LiDAR Cave Mapping

[wetb4igetinthewater]

Random idea, if it were possible to time sync a stereo camera system and a DPV INS system (like the Suex one), would be allow for on the move capture of the data? Perhaps with some manually surveyed check points every say, few hundred feet for correction? Any maybe have those check points also possible check that checkpoints for the radio beacon system that I've seen KUR use.

It certainly is. The known check points can be used to recalibrate the IMU data. That takes some additional effort, but if the scientific use of the data warrant it, it can certainly be done and will greatly improve accuracy.

A lot comes down to money. What is the precision required?

 

[Manatee Diver]

It certainly is. The known check points can be used to recalibrate the IMU data. That takes some additional effort, but if the scientific use of the data warrant it, it can certainly be done and will greatly improve accuracy.

A lot comes down to money. What is the precision required?

Just thinking of ways to do it with all the shelf hardware. The Suex Drive system is like $8k I think complete, and pulls time data through a GPS sync. You can timesync GoPros using the Labs firmware, I would think it gets cellular time, but perhaps it can be coded to use GPS time.

That should be accurate to fractions of a second. So correcting it with the known checkpoints should make it pretty accurate. Well beyond the level that a manual survey would be.

I was thinking it might be a good option to get starting data for Jackson Blue, since I heard all the survey data was lost years ago.

 

[mariosx]

Just to add on this point that IMU+stereo/lidar just with measurements registration will have, as very correctly pointed out before, inevitable growing drifting errors. This is a problem beyond underwater caves etc.
It is important for IMU(+magnetometer)+stereo sensor rigs mostly to provide locally consistent results for high quality mapping. To drop some keywords for whoever is interested: in state-of-the-art SLAM frameworks, global consistency is obtained with (sub)map matching and loop-closure, which eliminate any expected drifts.

 

[Ryebrye]

Has anyone thought about what kind of sensors you would need to get accurate positioning in the cave - assuming you had a detailed map?

There's plenty of work that's been done around using Bayesian probability to take repeated measurements and use it to get a good fix if you've got a good base map. Getting a high probability fix with a rangefinder/ compass bearing / depth measurement would be doable but not really useful (if it's clear enough that a rangefinder can give you a measurement, seems like your eyes could do just as good a job)

Sonar would be interesting though... If you have a base map and a series of sonar readings over time and have it take a series of imperfect measurements and validate calculate the highest probability location in the cave based on those measurements I think it'd be doable...

Now, in a total silt out could you even see a computer screen to make it at all usable... Not quite sure, but it's interesting to think about how you could take a base map and use it to get even basic positioning data - if it runs from when you start it should be able to be pretty good because it's not like you're going to be teleporting into random parts of the cave. Even turning it on in the middle of the cave and moving around a little bit, it should be able to get a good idea of where you are.

... Now if it'd be cost effective to do it, that's another question

Edit: here's a paper from 2021 that talks about using orthogonal sonar sensors to build up an underwater map. I think positioning using two sensors if you already have a map would be a doable thing as well.

Predictive 3D Sonar Mapping of Underwater Environments via Object-specific Bayesian Inference

04/07/21 - Recent work has achieved dense 3D reconstruction with wide-aperture imaging sonar using a stereo pair of orthogonally oriented son...

deepai.org