Study on bugs in drone software

[Daniel Ricketts]

I’m working on a study to understand the most common sources of bugs in drone software. I’d love to hear from developers about where bugs commonly arise. Here is a certainly non-exhaustive list of potential sources of problems:

• Mathematical mistakes in control laws
• Control implementation bugs
• Interactions between multiple controllers/modes
• Floating point exceptions
• Floating point rounding error
• Memory issues
• Timing issues
• Mode switch issues
• Sensor fusion problems
• Failure to notify user of a problem
• Communication failures

I’m using the loosest possible interpretation of “bug”, so a bug could be an important missing feature, a bad user interface design, etc. Another way of putting this is: what aspects of drone software are the most difficult to get right?

Just for some context, this is a part of a project to develop better verification tools and techniques for cyber-physical systems (veridrone.ucsd.edu). Our hope is to use this study to focus our work on the most common sources of issues in drone software.

 

[Motopreserve]

I'm not a developer, but I'll follow your results with baited breath.

Good luck with the study!

 

[Old Man]

One of the first "study" threads that's been started that has a foundation of research points that make sense and could do more to advance our flight controllers beyond what they already are. Not to mention what it could do for manufacturer transparency.

 

[dazzab]

I see you are already talking with the Arducopter developers on Google Groups and getting some good info from them. The PX4 team will probably be helpful as well. Good luck getting any info from DJI which is unfortunate.

 

[Daniel Ricketts]

Thanks for the positive feedback. I posted on the PX4 google group to hopefully get some good feedback from that community as well. I did email DJI, but as dazzab suggested, there's little hope of getting a useful reply.