SLAM and LiDAR Cave Mapping

[wetb4igetinthewater]

P/S: We hope that soon we would be able to map some shipwrecks and potentially caves using fully autonomous robots. The technology, from our side, is there for the past 2 years, but we had major issues with hardware.

And I look forward to collaborating on that with laser scanners!

 

[mariosx]

And I look forward to collaborating on that with laser scanners!

More is always better.

 

[Addison Snyder]

This looks very unreliable for creating the high details map we aim for

That's what I figured, I just had to ask. I know some fancier DPVs can estimate relative speed-to-flow.

It's a 2D sonar.

I don't remember getting an answer, but what component is that and how much do they cost? What kind of polling speed do you get? I looked at cheaper sonar solutions like the Ping360 but they had unsuitable ping times (~15 seconds for a 360 sweep) and I dropped interest in them after that.

checking with Dr Ioannis Rekleitis

Sent him a few emails, and just got off the phone with him. Pretty awesome dude! I'm really stoked to hear what he's looked into. He confirmed what I learned the hard way - GoPros use software magic to create their super-wide views, and no conventional calibration can correct for it. Took me a few months on-and-off to figure that one out. Nice to know what things I've gotten right, too. No image stabilization, extremely high FPS to minimize rolling shutter effect and more finely sync up frames in a stereo camera system. Stuff like that.

 

[Addison Snyder]

More is always better.

Lol not with $. If laser scanners were cheap, I'd love to explore them. I've yet to find one under $2000, much less something I can adapt for underwater use (blue/green light). Please correct me if I'm wrong though. I have been able to find 360 LiDAR that can be built for around a hundred bucks. Sure, it's slapped together and runs on red light, but add a few hundred bucks to up the quality, and you'd have yourself a system you can fit into a $700 tube that also conveniently holds high-end IMUs and an embedded computer to record (not process!) it all.

 

[wetb4igetinthewater]

Lol not with $. If laser scanners were cheap, I'd love to explore them. I've yet to find one under $2000, much less something I can adapt for underwater use (blue/green light). Please correct me if I'm wrong though. I have been able to find 360 LiDAR that can be built for around a hundred bucks. Sure, it's slapped together and runs on red light, but add a few hundred bucks to up the quality, and you'd have yourself a system you can fit into a $700 tube that also conveniently holds high-end IMUs and an embedded computer to record (not process!) it all.

The ROV/laser scanner package that we will be using in September goes for about $450K USD. It is practical for universities and institutions with benefactors. Not individuals.

 

[Addison Snyder]

So it's about as practical as draining the cave to get measurements. Maybe in 10 years .

 

[mariosx]

That's what I figured, I just had to ask. I know some fancier DPVs can estimate relative speed-to-flow.

I have no idea regarding their accuracy. I would trust more feature matching, IMUs, or potenztially DVLs.

I don't remember getting an answer, but what component is that and how much do they cost? What kind of polling speed do you get? I looked at cheaper sonar solutions like the Ping360 but they had unsuitable ping times (~15 seconds for a 360 sweep) and I dropped interest in them after that.

It's the IMAGENEX 831L. More details on the configuration are found here. Search for "A Modular Sensor Suite for Underwater Reconstruction" and download the pdf.

Sent him a few emails, and just got off the phone with him. Pretty awesome dude! I'm really stoked to hear what he's looked into. He confirmed what I learned the hard way - GoPros use software magic to create their super-wide views, and no conventional calibration can correct for it. Took me a few months on-and-off to figure that one out. Nice to know what things I've gotten right, too. No image stabilization, extremely high FPS to minimize rolling shutter effect and more finely sync up frames in a stereo camera system. Stuff like that.

Great. Yeap. Yannis is awesome. I was extremely lucky to be his student. There was a lot of suffering in our lab with issues such as this one, but that's field robotics y'all...

Not sure if you discussed about sharing data etc.

Lol not with $.

Well, I am not the one dealing with buying they toys. I am the one playing with them.

If laser scanners were cheap, I'd love to explore them. I've yet to find one under $2000, much less something I can adapt for underwater use (blue/green light). Please correct me if I'm wrong though. I have been able to find 360 LiDAR that can be built for around a hundred bucks. Sure, it's slapped together and runs on red light, but add a few hundred bucks to up the quality, and you'd have yourself a system you can fit into a $700 tube that also conveniently holds high-end IMUs and an embedded computer to record (not process!) it all.

My hardware skills are minimal I am afraid. But if you are able to build an udnerwater lidar for 1-2K$, please let me know, and I am confident that the underwater robotics community can make you a millionaire.

 

[Addison Snyder]

It's the IMAGENEX 831L.

What's it cost? As far as I'm concerned, if it has a five-digit cost, it's worthless for widespread use. Modest cost means practicality for the common cave diver, and something that will actually amount to more than a one-off cave mapping before a white paper is written and the entire operation left to collect dust. Maybe that's a harsh opinion, but it's one set in reality. As cool as Sunfish is (and just to be clear, I don't mean this personally or to single them out), I don't see more than what, 700ft of peacock from them? I'm hoping that there is a combination of tech out there that lets anyone say "maybe I won't buy a rebreather this year" and instead map a couple cave systems local to them. Doesn't have to be accurate to a centimeter (that's a hyperbole), just enough to navigate reasonably. Just my 2 cents.

My hardware skills are minimal I am afraid. But if you are able to build an udnerwater lidar for 1-2K$, please let me know, and I am confident that the underwater robotics community can make you a millionaire.

OpenTOFLidar/LidarTotalBOM.xlsx at develop · iliasam/OpenTOFLidar

Open Source TOF Lidar. Contribute to iliasam/OpenTOFLidar development by creating an account on GitHub.

github.com

Some of the components are more expensive now, but this is the acquisitions list. Not at all saying it'll produce professional results down to the millimeter, but hey, this guy has something you can build for in the neighborhood of 200 bucks for above water use. Switch out the laser for something blue, maybe add a narrow band filter, and it will probably work.

 

[mariosx]

What's it cost? As far as I'm concerned, if it has a five-digit cost, it's worthless for widespread use. Modest cost means practicality for the common cave diver, and something that will actually amount to more than a one-off cave mapping before a white paper is written and the entire operation left to collect dust. Maybe that's a harsh opinion, but it's one set in reality.

The most current pipeline runs only on a single gopro9 or on a pair of gopro9s. Check this link in case you missed it.

As cool as Sunfish is (and just to be clear, I don't mean this personally or to single them out), I don't see more than what, 700ft of peacock from them? I'm hoping that there is a combination of tech out there that lets anyone say "maybe I won't buy a rebreather this year" and instead map a couple cave systems local to them. Doesn't have to be accurate to a centimeter (that's a hyperbole), just enough to navigate reasonably. Just my 2 cents.

Only hardware issues are stopping as at the moment from dropping our robots in the cave. SLAM works and navigation also. I can personally guarantee the second one as the main developer of the navigation stack and enough experiments with the robot navigating safely in extremely confined environments. I am confident we can do better with a completely untethered robot in the very near future. I don't have access to the technology of Sunfish, but from few demos I have seen I speculate that they are using a bit more old-school approach that face some challenges we are able to resolve. I am doxing a bit myself but this is 4-5 years old work. You can see safe navigation in both known and unknown environments, inside simulated shipwrecks, etc. Real demonstrations were more limited due to issues with SLAM (lack of texture, features, etc) and hardware issues we faced (motherboard getting fried twice) before we run more interesting experiments. Assuming SLAM working, the average cave seems an easier environment to explore.

I believe strongly that the technology is there, it's mostly a matter of time and money. The problem is that yet there is no enough profit from people interested to map caves instead of buying RB. And especially when the primary sector driving the market is oil/gas and defense, well... they have enough money to waste. You think Sunfish (?$) or our robot... Aqua2 (150K$+) are expensive platforms? The robot I am working right now costs 22M$... If for some reason that becomes a priority, well BlueROV runs at 12K$ lets add some cheap sensors etc it can go up top 15K and give your price for the software which could be anything from 0$ if it's opensource to $$$$. It can happen. It's just that no company would survive raising those money in Shark Tank...

View attachment 775707

OpenTOFLidar/LidarTotalBOM.xlsx at develop · iliasam/OpenTOFLidar

Open Source TOF Lidar. Contribute to iliasam/OpenTOFLidar development by creating an account on GitHub.

github.com

Some of the components are more expensive now, but this is the acquisitions list. Not at all saying it'll produce professional results down to the millimeter, but hey, this guy has something you can build for in the neighborhood of 200 bucks for above water use. Switch out the laser for something blue, maybe add a narrow band filter, and it will probably work.

Robots are not simple systems, and I would say that every second of their existence "defies" the probabilities. The problem is not only on the single sensor, actuator, etc. It's also (if not mostly) on how these are combined and how they rise the complexity of the entire system. Many many many things can go wrong and if somebody would like to guarantee performance, especially in remote underwater cluttered environments, you need the best of the best of the best solutions which costs. I am in the DIY forum advocating a bit "against" of DIY, but I think what we are discussing is well above the average complexity of the other topics.

 

[Addison Snyder]

Update to the SLAM cave mapping project. Bought some new toys - a sensor tube, pressure/temp sensor, a whole new camera/IMU system, and a reaaally awesome IMU (in the progress of fabbing a breakout board for that). The days of syncing up video files, extracting IMU data, etc, are over! Gonna finally integrate depth readings in, completely eliminating one whole axis to the positional estimation.
Need to build some internal structure to hold all this stuff within the tube, but the software/wiring side was super simple.
On a sidenote, does anyone have some extra single-tank adapters they'd be willing to donate? Was hoping to use a couple of those to rig the tube in front of my DPV.