Alternatives to Raspis

[esm]

Has anyone had success running Pypilot on any hardware other than a PI Zero 2W or PI4, like Banana PI, Orange PI, or other boards?  Can't get any Raspis as everyone knows.  Thanks for any suggestions.

[seandepagnier]

I used orange pi a while ago so there is already some support for them in places. It should in theory mostly work on other boards with only slight changes required. I will be seeking alternatives sometime next year when I run out of raspberry.

[ironman]

This guy uses an old smart phone and an ESP32 over bluetooth to run it all https://www.youtube.com/watch?v=65PKalHsWdA. He studied the pypilot code then did his own thing. Ticks many boxes in one go.

[seandepagnier]

A similar to:
https://www.youtube.com/watch?v=C9QuWKvwhN4
https://github.com/yOyOeK1/oiyshTerminal

at least oiysh has source code.


It seems interesting but there are some issues:

1)  no source code..  we can't collaborate to improve it.
2)  using internal phone sensors means the results will widely vary across device.. difficult to quantify, and many phones do not have compass sensors
3)  using android, I dont know if you can ensure realtime.    I think there must be more lag from measurement to update also with a wireless link, perhaps milliseconds difference but it matters more as boats are smaller or faster.
4)  rely on android calibration algorithm..  pypilot is optimized to calibrate the compass on a boat and calibrate and compensate other sensors as well.

Autopilots will range in performance

•   not good enough
  
•  adequate useful performance
  
• ....
  
• pypilot
  
• ...
  
• ideal performance
  

So android phone can be shown to have adequate useful performance in nice conditions which already makes it very useful, but even then delayed reaction may lead to less efficient sailing or more wear on the motor, steering and even sails.   For example,  using elastic from jib sheet to tiller also can steer most boats on most courses and conditions, but the miles traveled per day is lower and the chafe of the sail and lines will be higher.

Eventually maybe even run pypilot on the phone with the understanding of limitations, and   python can certainly run on android.   
Keep watching this space as more development occurs.

[ironman]

Agree on all points, except that it is still interesting to see how he might succeed in dealing with those issues. Even closed source developments can be inspiring to follow. And the mere idea of boat rudders having APIs that can be called by a wide range of autopilot applications is a good example of thinking against established system delineations.

[ironman]

The bluetooth code is quite easy to follow. I'm toying with the idea to tape an esp32 to a pypilot motor controller, bridging bluetooth to the serial connection, then adapting servo.py to use bluetooth. The esp32 could later also serve button clicks back to the pypilot. The idea is to open up that motor interface, standardise it at some point, to break up and thereby be able to combine these old and new efforts. I don't really know why that idea seems so appealing, though.

[seandepagnier]

I am wary of using a wireless link here, but open to testing and comparison. If I ever get it set up, I would scrutinize the latency.

I think the only way to ensure it wont have lag no matter what to use esp32 on both ends rather than the raspberry bluetooth.

It may work most of the time but occasional 50ms or more lag is not acceptable. Already know this is the case with wifi.