Hi Kristian,
Thanks for responding. I’ve spoken to others who are trying to fuse the velocities with usbl/lbl positioning.
Note @williangalvani is prototyping a lua script to more seamlessly switch from GPS positioning to DVL positioning which would make using waterlinked components much easier. It would allow you to acquire GPS at the surface before diving to the working depth.
I’m sure it could be adopted to take in UGPS values.
Ultimately Kalman filtered positioning incorporating DVL, Usbl/lbl and INS data would be the best solution. It’s what’s used on survey spec’d ROVs that have to maintain a very high degree of accuracy in the data string.
Ardupilot EKF (Kalman filter)
I don’t know whether or not the data can be fed into ardusubs inbuilt Kalman filter, on the surface it seems like it would work, but I’m sure there are obstacles. The fact that depth is used for altitude in the ardupilot ekf, or an issue arising from “number of satellites” data and how error varies with depth. But those are all guesses. A lot of people on here know more about this stuff than me.
Processed on topside
Kalman filtering is supposed to be quite doable in an external script ran on the topside unit (though not doable enough for my skill level).
Also data can be Kalman filtered “after the fact” so could potentially be developed and tested using logs.
I guess if waterlinked had a proprietary kalman filter that ran on the topside computer, you could send corrected data down to the ROV using NMEA injector.
It would encourage people who already own a DVL to get a UGPS and vise versa 
It would certainly encourage me for what that’s worth.