beta tinypilot image 2026 support pi zero 2, pi 5

[seandepagnier]

https://pypilot.org/images/tinypilot_2026_07_05.img.xz

• rebuilt entirely on top of tinycore 16,  supports pi 4/5 zero 2.
  Still intended mostly for the original pi zero where it uses about 30% of the cpu.
  on the multicore pi's it boots much faster but otherwise mostly just consumes more power
• integrated the latest improvements to pypilot, too many to list here
  
• did sailing in april/may testing the new image in a range of conditions, from 0-30 knots and up to 15 knots boat speed.
  

There are some additional pilots that actually are tested and working.

The "absolute" pilot requires a rudder feedback sensor and uses it to calculate rudder movements.  Before with the "basic" pilot (which is still the most tested/proven one), the rudder feedback was only to stop the motor from using current limiting to detect end of travel and potentially report the position.  So far in waves or running off the wind this "absolute" pilot is not really better for me, at least not yet, but in light winds if you have a drive unit that can efficiently move at slower speeds (ball screw)  it had a very nice behavior, barely making adjustments, and also can quickly apply course corrections with less overshoot. It needs more testing!

The "wind" pilot requires a wind sensor and essentially reverses the wind and compass sensor roles.  The "basic" pilot is really always a compass pilot and adjusts the compass course it actually follows to make other modes like wind or gps work. This still is probably better in very light wind and/or with a wind sensor that is moving a lot.   With the "wind" pilot in wind mode, it now completely ignores the compass sensor using only wind and gyro inputs.  This has some interesting behavior following wind shifts very closely especially useful in light winds going close to the wind. In "compass" mode it adjusts the wind angle it follows to try to keep a compass course... I only tried a few times and it did hold the course, but can't really do extensive comparisons.

I hopefully can post a video and more details soon.

Feedback is greatly appreciated as I am now in Colorado and unable to do sailing tests for some time. I would like to fix any potential bugs quickly.

[Moeritsen]

Hi Sean,
that’s really good news, thank you for the new image.
I hope to read about more of the new features soon.

One question: Are there new sensors for gyro and compass to be used after icm20948?

cheers
Michael

[seandepagnier]

I did not plan any new sensors yet, but they can certainly be incorporated if there is something out there.

[Moeritsen]

HI Sean,

I’ve now installed the beta on my boat using my existing Pypilot hardware.
To do this, I first transferred the existing calibration data to the beta image.
The image booted up fine, but the calibration data no longer made sense. The compass, roll, and pitch readings were way beyond realistic values.

Then I tried a new calibration for acceleration. After 15 minutes of patiently rotating and turning my ICM20948, I was unable to achieve a valid calibration. 

The compass calibration hasn’t worked yet either, but that’s okay, because the approximately 20-cm-long loose cable on the sensor makes calibration nearly impossible. I’ll have to do that at sea.
One big step forward is the export and import of profiles—it works great right on the computer running the web interface.
I’ll do some more test runs and then decide whether to switch back to the older, tried-and-true image.

Cheers Michael


Update: Just went to my boat again to check the calibration responses, I noticed, that both calibration sites are ok, no issues, no bad responses. Seems that a reboot after calibration procedure is fixing a lot….