Modular Blue Robotics Based AUV

Build DIY AUVs
21

One option to look at is the Nortek Nucleus 1000. Its a roughly hocky puck sized pre-synchronised navigation package which combines a DVL, Altimeter, Temp and Pressure, with an on-board AHRS and INS. It has a single power and comms channel, draws an average of < 4 W and outputs a position estimate with typical performance in the region of 2% of distance travelled. Main limiting factor is heading on that system like any other budget dead-reckoning solution.

The integrated DVL as a bottom track range of 50m in typical (not just best case) conditions, and if you start your run in shallow water, the DVL will automatically measure background currents to give you a corrected watertrack measurement once you get out of bottom track range. It weighs 300g in water.

What is your typical application?

It doesn’t currently have GNSS + USBL input but will do soon.

What sort of budget could you realistically put toward the navigation solution as a whole?

22

Achieving 0.1 deg bias by calibrating a magnetometer is very impressive. How do you do that calibration?

23

Nils, thanks for the info. I believe the Nucleus 1000 is in the price vecinity of the A50? Spec wise how do you think they compare?

Would the Nucleus do the fusion with an external INS/GPS?

24

Hi @ljlukis

The RM3100 is sensitive like a flux gate rather than like a Hall effect sensor. That helps a lot.

I calibrate the mag with the 10 Parameter Magnetic Calibration Model. It’s a least squares ellipse fitting model. I used a bunch of different references for this, but the best (I think) is AN5019 from NXP Semi.

The 0.1 is probably not that good. That number is about repeatability of a measurement with some spot checks of rotations drawn on a wooden board. It’s not a complete survey. I don’t think that the local magnetic vector I use (from WMM) is really that accurate to start with and of course I have no idea how the immediate vicinity is affecting that vector.

What I mostly need is consistency because I’m navigating between GPS fixes. The position error I see is a function of drift, of speed through water and of course heading, all rolled together. As long as the heading is not wandering around I don’t care so much. I just make a new corrected heading estimate and off I go again.

I think the big difference here is that most people are wanting to use AUVs as survey platforms. That’s my secondary goal and if/when I go down that path I’ll almost certainly need a DVL or perhaps some kind of SLAM approach. My main navigation goal for now is about self deployment and recovery.

2 Likes
25

Okay that makes sense. Thanks for sharing!

26

Just a quick reminder of @EliotBR great work putting together a comparison of DVL’s out there.

It’s fantastic for the community to here from the suppliers @Nils.coe and @ljlukis and their incites

I love to hear how we as a group can improve the accuracy of underwater navigation given the real world missions eg

Scott

1 Like
27

Are you experienced in the DSP side of underwater comms? Fundamentally I understand it for static-non-multipath environments. Of course, realistically there is some synchronization requirements, FEC, doppler corrections, multipath corrections, etc. to deal with. This is why I was planning to use UNET STACK (free for personal use)
 but requires a Raspberry Pi or other Linux based SBC.

28

No, sorry. The bulk of my experience is about passive sonar for subs. I don’t know much about comms, so if there is a library to at least start with, I suggest to go that way. Once you get it working on a Pi you could port it to anything with floating point h/w.

The multi path stuff can get very complicated. You can have the same waveform received at two sites 10 metres apart and they are wildly different. It’s not quite the same problem as adaptive cancellation on a high speed cable, because you won’t be using a continuous transmission to train the filter. If the library doesn’t work out, I’d avoid modulation like GMSK and stick with something that looks more like LoRa chirps. FM is wonderfully forgiving stuff.

29

Yeah, I’m in a similar boat
 I worked on active sidescan and multibeam. They are stright foward when compared to a-comms. Thanks for the advice on modulation scheme.

30

Hi Bryan,

Full disclosure, I work for Nortek.

I believe we around the same price as an A50.

Spec-wise, its not like the systems are like for like. A50 is a DVL, whilst the Nucleus is more of a all-in-one navigation solution. That being said, the main place the A50 beats us out is on size, its just so compact. The Nucleus is larger, at 9cm diameter and 4.2cm height. However, you do have more sensors, the full INS capability and ability to repair with that extra space.

Comparing the DVL functionality alone though, we are reliably getting 50m bottom track in typical operating conditions, and Nortek tends to be conservative rather than optimistic when it comes to quoting specs. We also have temp and pressure sensors onboard the DVL which we use in combination with a manually inputted salinity to do real-time speed of sound calculations, which improves our DVL accuracy compared to DVLs that lack these sensors.

In addition, our on-board AHRS is a true AHRS in the sense that we have a gyrostabilized magnetic heading sensor onboard as apposed to just the IMU.

We can then use that proper AHRS in combination with the DVL to provide full, self-contained positioning functionality with the built-in INS. That being said though, the whole package is competitively priced with the A50, and the AHRS is fully MEMS-based so hence you are only getting 2% of distanced travelled position accuracy.

Right now, the main use cases for the Nucleus is as a budget friendly way to achieve reliable 50m bottom track as a basic repairable DVL for station keeping, or as a full navigation solution making full use of the on-board AHRS and INS. As such, the Nucleus is not designed for integration with external INS systems.

However, when we add support for USBL and GNSS, these inputs will be ran through our INS’s Kalman filter, as apposed to just resetting position if you get what I mean.

31

Nils,

Thanks for the breakdown. Seems like a good option and maybe better option once USBL and GNSS is added
 and just maybe, the cost of USBL will not be 3x the DVL :slight_smile: .

1 Like
32

So, just reframing the issue broadly we wish to run an AUV for multiple km’s (multi hour) of mission (be it Sidescan, photogrammetry, other forms of seabed scanning or @bsupine ‘s whale songs this longer runs is in part the whole point in autonomy) and need to know both where the AUV has been and where it has scanned and if the area we scanned had no holes. Additionally, being able to return back to an identified target in the data recording whist being both technology and brand agnostic. (this is not having a dig at any of the suppliers) its about an open discussion of the best techniques to achieve this

Sorry for the use of so many acronyms (also have a look @rjehangir great primer document or Wikipedia’s write up

Broadly strategies are

  • GPS position (absolute locations) on the surface and then to so surface regularly to “reset” the estimated position along the track

  • Use essentially double integrated Gyro’s and IMU’s to get estimated position (relative positions) which quickly degrades with time/distance, greater $’s can tighten the accuracy of units but it still a double integration

  • Use DVL (with greater or lesser feature integrations) to get estimated position (relative positions) which are essentially a single integration so although it degrades with distance it does this at a lesser rater the IMU’s, greater $’s can tighten the accuracy of units (and give greater depth bottom locks) (2% to 0.1% [export controlled] of distance thus for the above 7.3 km mission accuracy within 146m to 7.3m note with the above SSS mission we only had 70m swath and was just 1 hour)

  • Use Locator Beacon/s (USBL, SBL, LBL) to get a position underwater (absolute locations) the accuracy varies (using Waterlinked (US$11.4K @300m approx. ± about 6m) with limited range (100- 300m - longer with substantially more $’s) so essentially why have a AUV at 100- 300m range as this could be a ROV with a 300m tether alternatively given longer runs it may be more of a drive by “reset” of the estimated position along the track

  • Combinations of the above

  • Something else (any thoughts)

Broadly (sorry for grouping everyone I know there are multiple people and backgrounds here) most of the people on the Forum are not large corporations and so are looking for cost effective answers to the navigation issue

Thus, part of the measure really should be total cost of ownership/implementation verses accuracy at (ideally a multi hour multi km mission)

In my 1st post I sort of declared $’s for underwater navigation at around 1/3 of the total cost and greater than the Blue Robotics parts.

So simplistically is it better to live with

  • Say US$4K and use a Cerulean DVL Tracker (a little less defined accuracy) and use more surfacing location corrections (?at the end of each lane as the turn around?) (and break up longer missions),

  • At say US$8-10K get a 1 to 2% accuracy DVL (error 70-140m in the example)

  • At say US$12K get a 0.1% accuracy DVL (error 7m in the example)

  • At US$12K ditch a DVL and just use an Underwater GPS G2 BlueROV2 Kit with a mid level INS with data correction drive by’s (say 1km long lanes it could be up to 600m running on just the INS)

  • Or Hope and wish and wait for a cost-effective modem/LBL solution

Scott

1 Like
33

This seems like a good place to compare solutions like the Nucleus and the A50 to the Cerulean Tracker 650, and to elaborate on the USBL side. BTW, I have nothing to take away from the A50 or the Nucleus, I believe they are both great products.

@Nils.coe explained that the Nucleus approach is a “full, self-contained positioning functionality with the built-in INS.” Tracker 650 takes the opposite approach. At Cerulean Sonar (full disclosure: I work there), we always try to find the lowest cost approach to get the basic functionality. The Tracker 650 implements only the Doppler functionality providing velocity information relative to the device. The Tracker 650 leverages the AHRS/IMU in the vehicle as the central source of truth about attitude (fully implemented for the MAVLlink/Ardusub environment).

The Omnitrack USBL (recently launched) is a full top side deployment system featuring integral GNSS compass for accurate position and trouble free heading input to the on board IMU. Omnitrack interfaces directly in the Blue Robotics/MAVLink/Ardusub environment, providing full fusion of USBL (position, 1 sec updates), and Tracker 650 DVL (velocity, 10Hz updates) data.

Here’s an idea, replace the standard Omnitrack ring float with a Blue Boat and you can really track your AUV where ever it goes!

2 Likes
34

I’m hoping many innovations come from this group of creators/makers that lurk the dicussion boards of BR and others. Some of this innovations shall make larger companies cut the costs just like Waterlinked did when they introduced their DVL 
 now Celurean is competing in this market too.

As you said, it is very mission oriented. I know people that never let their vehicle go further than 0.5km. For this the GPS-G2 would work. Other want to go 1-2km far, do a scan, and RTH
 these might be better suisted by the enhanced DVL
 or surfacing once or twice. Depth dependent. LBL is great but time consuming to deploy and perfect in multipath environments.

Now, can someone answer me this? When WL A50 is upgraded to 0.1%
 what does this really mean if we are using a low cost MEMS INS? Do we still get the 0.1% error? – I’m guessig not but has anyone done the math or tested it?

Maybe there can be a matrix or calculator that can help us determine that right navigation sensor based on our budget and requirements. Some of us may have to spend longer days surfacing a few times to get the needed accuracy.

35

Sly grin :thinking: wouldn’t that just sort of be a tetherless towed sidescan, you would need a U/W Modem top and bottom plus the Blue Boat so the AUV knew its location

36

If you only need speed through water, you could use an EM log or a paddle wheel.

This is only going to measure movement compared to the water body, rather than against the sea floor. If you can reliably subtract out the movement of the water mass, it’s an option.

37

Well, yeah, you’d know where the AUV is, but it wouldn’t.

38

Hi @psupine

The Electromagnetic log is an interesting thought, a pretty simple devise (although not “off the shelf” in the AUV world) but I did find this implementation on a boat so it’s been done and would not be a substantial cost. From the paper (here) there was an AUV that had a 20.5 m positional error over a 712.1 m run (2.9%) with one (not quite but in the broad area of some of the DVL’s). I can imagine calibration and characterisation runs being fairly straight forward.

Hard to get accurate estimations of the error but being that it is basically an Electromagnetic flow meter (common in industry) which easily get better than 0.2% velocity.

Big picture its output is just a single integration (velocity over time with the time then giving distance - the same integration as the DVL albeit no account for vessel drift due to currents or IMU variance)

As an aside I still sort of expect to use a DVL this isn’t about minimising cost it about being able to do job with the tech that is required to achieve

Which may offset endurance (time underwater) with surfacing (or alike) for corrections

Scott

39

That little arduino project is a paddle wheel type, not an EM log .. unfortunately.

I probably come across as a scrooge. To the various community members here who are associated with companies making this gear, please let me assure you that I do understand that these things cost money to make. They’re a niche product and just on numbers sold vs cost to keep a business alive there’s only so much to squeeze. If you squeeze too hard the business and the product vanishes and nobody wants that.

A bit over 20 years ago I led a development team on a new multibeam sonar with an 8:1 continuous swath. There were 8 beams in flight simultaneously, launched fore and aft at different angles and at different frequencies, sweeping hyperbolic arcs across the seafloor. They were all spread spectrum pulses. It all mapped into a live processing suite that surveyed directly against the WGS84 geoid. It ran live stats on the seafloor grid; the whole data cleaning chain that was usually (maybe still is?) done “offline”. The idea was that hydro ships cost a lot to run, so rate of effort was everything, and that you’d never have to go back the next day to fix a patch of seafloor you missed. We did everything we could to maximise coverage at the official international hydro survey standard. The sonar ended up being expensive as a result. We thought we had a really solid business case against the operational costs, but the market didn’t agree and the product crashed badly. (I was so sick of that project anyway and I went back to my beloved submarine sonars)

The experience taught me that the fanciest engineering solution isn’t always the right solution. I think DVLs and side scans are really cool, but I’m always looking for a way to leave components out.

1 Like
40

No mate it’s fine, it was more a generalised comment as things are moving from what I (and hopefully others) am trying to achieve of a usable AUV

I still think that the EM Log still has potential legs maybe tied with you suggested Analog Devices ADXL355 and PNI RN3100 with the 10 Parameter Magnetic Calibration Model. I have had a quick scan and there is a bit out there associated with Magflow electronics which is basically the same device

Scott

1 Like