Open bsutherland333 opened 6 months ago
I also don't know if I'm going to have time to finish this. The approach we were using before of a gradient based optimization is a flawed approach I think, as it requires too many step iterations to converge. A system identification type method would be better. However, much of the work we did here was simply creating the functionality we needed to have the autopilot perform an automated sequence to measure the system response. That portion of this work would probably still be useful for a system identification type approach.
An autotuner for ROSplane would make it much easier for people to use, so this is a gradient-based optimization approach for an autotuner.