Hello Reader,

Welcome back to The Build Log! Today, I am continuing the last weeks newsletter on testing our autonomous submarine. This is the first time we have tested the submarine underwater fully autonomously. As you might know from our last week's newsletter, this comes with a lot of challenges.

Once the submarine submerges, it should dead-reckon its position based on data from DVL and compass. Unfortunately, this did not turn out to be that simple - for the first couple of runs, the submarine has been crashing into the bushes of the lake. If you want a refresher, you can read the previous release of the newsletter (access it at the bottom of this e-mail).

Submarine just before submerging and submarine stuck in the bushes

Thinking the issue is just inconsistency of the DVL/compass readings, the first thing we have tried is repeating the mission multiple times. Unfortunately, each time the submarine has crashed into the bushes in the same exact spot. That was when we figured out that the error might be a consistent DVL/compass offset instead of just measurement error.

We then came up with an ingenious idea! We can run the mission on the surface of the lake instead of underwater. That way, we can run both algorithms in parallel - the dead-reckoning algorithm, and the real-time GPS readings. Thanks to this, we have both of the data sources available - the "real" position measured by the GPS, and the "predicted" position, which just uses dead-reckoning.

Thanks to this, we were able to determine if the lack of consistency in position readings is caused by a random error, or by a systematic error. Here are the results:

The numbers in the picture above represent consecutive measurements of GPS-measured "real" position, and the position calculated by the dead-reckoning algorithm. As I mentioned before, the only reason for why this was possible to measure is because we have performed this mission with the GPS antenna at the surface instead of underwater.

Analyzing this data, you can see that these measurements are off by almost exactly the same angle from the starting point.

Applying this offset to all of the measurements, we were able to eliminate most of the DVL position error we were experiencing in the previous mission and complete the first mission of the submarine successfully. For now, it was a really simple mission, that was done at only 50 centimeters underwater. We plan for this submarine to swim very close to the bottom of lakes and oceans. We'll update you on the progress of that soon!

If you want to build an underwater project, the easiest way to get involved is to build the entire CPS 3 drone with our DIY Kit -> now available!

You don't have to buy any components separately, because they're all in the one box delivered to you.

Cheers,

Piotrek, CPSdrone

PS: You can access all of the previous releases of the newsletter here.