When I worked at Amazon Prime Air and Zipline, we had cloud infrastructure that ingested all flight log data and processed it to extract metrics like flight time, power/energy info, faults and firmware info, sensor min/max values and heaps of other stuff.
I made a system like that for my own logs as a personal project, but then I thought that it might be useful to the whole community, so I’ve spun it into a service.
demo.apisdynamics.ca
It has an API and a set of Python functions that allow you to automatically sync logs from a flight controller to the cloud using any Linux companion computer (you can also manually upload logs in a browser), and a set of example Jupyter notebooks that you can use to pull 1000s of logs from the cloud and analyse them en masse to characterise fleet performance and compare across firmware versions, hardware revisions, or geographical locations of weather conditions.
Every log is searchable by its takeoff coordinates, duration, max altitude, max roll/pitch angles and rates, firmware version, and much more including pilot anecdotes.
@tridge I feel like this would be really useful to the Ardupilot Dev team, especially to compare performance and regressions between versions. Or do you guys already have something like this?
I’m also keen to hear feedback on what would make it most useful to the community. There’s a non-zero cost to the cloud storage and compute infrastructure so I have it set up with a paid subscription, but the first month is free for anyone who wants to try it out and provide feedback.
Thanks for listening!
