Low cost DVL for subsea navigation and merge with INS

thanks for the reply. I agree with everything you state here. I am thinking we could get into the $1K or even less range with design reductions. I think maybe 50ft is high enough. i would think maybe a foot minimum distance. I havent done any study on the min acceptable velocity accuracy to help with an IMU, but it would be dependent on the heading drift, etc. but something that can sure be simulated with some time, I think even 100m is adequate max depth. I was thinking keep the IMU separate since there are so many available at a vsariety of price points and accuracies. probably 5V or 12V. USB is ok but even a simple rs-232 would work. lots of the IMUs do have 485 I/Fs so that could work as well.

1 Like

Most of the DVL’s I’ve used cutout around 80ft off seabed, but it depends on your frequency, operating window, power, and signal to noise ratio (SNR). The more the maximum altitude can be reduced, the better because it decreases the drive power and/or SNR required. another benefit is increased update rate. Minimum range is a function of a few components, but mostly acoustic frequency. The higher the frequency, the shorter the minimum range. 1ft is good even at 600kHz. Higher frequency of course equals better resolution but reduced range.

What sort of power requirements are people sensitive to? They don’t draw much, but for some AUVs even 10W can be a lot in their power budget.

An easy way to think of a DVL in your navigation EKF is that it “caps” your mid-term integrated velocity error to the absolute accuracy of the DVL ground speed. Heading isn’t much of a problem because the compass can keep the gyro zeroed, but on the X/Y accelerometers, even a bias of a thousandth of a g in offsets, temperature drift, or non-symetrical noise will quickly integrate into the maximum velocity of ROV (or more if you don’t set bounds on it). But if you have a DVL, let’s say +/-1cm/s absolute accuracy over the expected ground speed range of the ROV for simplicity sake, then your worst case scenario wont result in more than 3m error after 5min and 36m after an hour. If that seems bad, with an IMU offset of just 0.0001g, your position could be off by 44m in just 5min.

Without processing the data near realtime, and with tilt/heading from IMU/compass, the DVL won’t really aid in navigation, but could still be used in post processing. However, this won’t take much processing power so it could potentially be done in the raspberry pi or pixhawk (or about $50 worth of cpu/imu) and people could post process with a better IMU if they had one. What would your preference be?

1 Like

Has anyone experimented with an optical flow camera?

Any news with this integration?

There has been a lot of progress towards a cheaper ($2,000ish) DVL, but it won’t be ready this year. Although the optical flow technology holds a lot of promise in clear water, the reality is that most of the places you’d really want a DVL don’t have much visibility/or ambient light. That’s why we’re headed the acoustic route. However, if anyone wants to investigate the optical, there’s good info here: PX4FLOW · PX4 v1.9.0 User Guide

1 Like

Peter, so sorry for delay in response here.
Max altitude we’d hope for would be 150ft. (Aligns with some of the cheaper MBES 100+/- 50 ft. above the seafloor)
Power - as low as humanly possible. Agree 10W would be high and difficult to balance in the application we’re looking at.
Would be amazing to have a 300m & 6000m rated versions. Use usbl for decent, and DVL at depth for an extended mission (or another clever combo).
I think having the ability to choose between realtime processing, and post-processing correction is a pretty large architectural decision, and it would be great to have the option to do either. However, I think having it + IMU/compass and real-time processing would be the best. If there was a purpose built chip, designed for the low power consumption (like what we see on your camera board), that would aid even further and reduce the load on the central compute - not that it’s that much anyways).
Curious how a few configs would change the power/price of the offering.

Looking forward to see how this pans out!

Sorry for the late supply. The Nortek DVL1000 outputs a proprietary data string and we didn’t have time to write drivers for Ardusub to read the data, or to write code to convert it to NMEA. We’re hoping to get funding to buy one and continue development.

I’d like to integrate a DVL into this vehicle that runs on Ardusub that we built for catching lionfish. I don’t need accurate positioning, but good station keeping will make catching the fish much more efficient.

3 Likes

Hello Anthony. Thank you for the input. The processing overhead isn’t usually as power hungry as the transducers themselves. at 150ft the drive power could be hundreds of watts each time the transducer fires, but if altitude and power requirements are what you’re after, then a lower frequency (and hence precision) transducer and less frequent update rate would be the ticket. So, at the very least we would want to include widely configurable update rates.

300m and possibly 950 would be a good marketable range. The last industrial DVL i worked with had a 6,000m depth rating that required a $9,000 external Titanium housing. If we can keep the size small (though it requires higher frequency transducers/shorter range) then the cost of a deeper unit would be reduced, but 300m is relatively cheap to attain in any size.

I have not rolled an ASIC, though I have worked in the IC industry and FPGAs for years and it’d probably never pay off to spin custom silicon. There are some cool shortcut to reduce the processing power and a smaller FPGA with careful power management should require less average power than the transducers themselves.

Gcelec. That’s an impressive way to fish. How many can you bag per day?

if you’re in relatively clear water near the bottom, optical may obviously be a good solution. What average altitude and depth would you anticipate requiring for an acoustic DVL?

1 Like

Could not agree more.

The two that I’ve found are (as stated above):

  1. Nortek DVL1000-300m https://www.nortekgroup.com/export/pdf/DVL1000%20-%20300%20m.pdf
  2. NavQuest 600 Micro DVL LinkQuest
    (Any others that I’m missing??)

These are currently the baseline for getting to the bottom, and keeping a lock. Of course, transiting larger depths you still run into problems getting to the bottom. Though still solvable with other solutions.

Configurable update rate would be fantastic if possible (don’t see why not) to give a robot the ability to adjust based on available power and conditions.

1 Like

@BlueLink_PS Until recently, the vehicle has been fishing in Bermuda where the infestation isn’t as bad as it is here on the East Coast and Gulf. In Bermuda they’d fish about 30 or 40 a day, but now the vehicle is in Florida where we hope to be shooting over one hundred per day. Ultimately though, we need to catch about 3-400 per day and that’s not possible without rock solid, economical station keeping solution.
Once we’ve achieved solid station keeping at low altitudes we’ll add machine vision to track and shoot the fish semi autonomously, and and eventually autonomously.

2 Likes

Does anyone know anything about using MBES for terrain locking to replace the DVL altogether? I’m assuming it’s just the range of the MBES’s that’s requiring another long range DVL for diving?

Another one to consider is the RDI 600 kHz Pathfinder DVL.

I am aware of a Blue Robotics vehicle owner at a University in Asia who is integrating a Pathfinder into their vehicle.

1 Like

Both MBES and DVLs can reach 6km range or more (obviously at low frequencies only). Some survey setups can do terrain matching or terrain flow but, to my knowledge, it isn’t as good a quality, at least not compared to a high end 600kHz-1.2MHz head. Also, at higher altitueds, anything but a $500k MBES may not produce a good lock on some locations. I have been in the Mediterranean and Black Sea where a Reason 7125 couldn’t make out hardly a feature on the seabed from 10m high, it was just a flat abyssal plane, but an RDI Workhorse 1200 would hold the ROV within 50cm for hours without any USBL or INS input. Some of this is probably carry over from the industry, and some of it is job requirements (end client will frequently specifi what type of survey equipment is used and so if they say FOG, 1cm accuracy MBES, and 600kHz DVL, you just have to bring $1M of survey gear to win the contract. It’s all job dependant. If I were making a shallow reef, in clear water, on a budget, visual flow or even scanning sonar and MEMS INS would be a good choice.

2 Likes

A little discussion of the upcoming Cerulean Sonar DVL has been posted here: Doppler Velocity Log (DVL) -
No pricing set yet, but it will definitely be less than a BlueROV2!

9 Likes

This is super super exciting!

Cannot agree more, super exciting!

Hi everyone! Just wanted you to know that we just launched the DVL A50. Take a look here, we’re sure you’ll like it: Water Linked launches: DVL A50 Doppler Velocity Log

3 Likes

Looks awesome. Well done.