A-maze-ing

Chloe, LEDDAR and distance PIDs from Andy Baker on Vimeo.

My distance / direction PIDs are now at nearly perfect tuning, and along the way, I’ve also learned a lot about how to tune the three tiers of PIDs in the horizontal motion hierarchy; it’s been slow, hence the lack of posts.

As far as the PIDs are concerned, the key is that the I-gain on all but the vertical velocity and yaw PIDs is now set to 1/100th of the P-gain; from the still existing wobbles it could probably do with going lower yet, though I think I may still need some to manage windier conditions.

The remaining horizontal drift is likely to be unfixable without the Scanse Sweep or PX4FLOW; here it’s just using double integration of (acceleration – gravity) which frankly, I’m amazed it performs as well as it does!

The yaw still needs some attention; currently, it too has two PIDs (angle and rotation), and I think that may be one too many and the latter can be removed.  At the same time, since the input is the integration of the gyro, then it’s never going to be much better until I get the compass working – it’s continuing to not play ball.

So for now, focus is on minor PID tuning, sorting out the yaw PIDs and getting the compass working, but before then, I think she’s ready to finally try some intentional horizontal motion; I have a flight plan put together than has her tracing a horizontal square at fixed height – 2 second segments each of take-off, forward, left, back, right and finally landing.  Forecast for today is good, so I’ll try that later, and post the results unless they are shamefully embarrassing!

Oh, and the reason behind the title of this post?  It’s now confirmed that the distance PIDs are working, so that a combination of PX4FLOW for 3 axis velocities, along with Scanse Sweep object avoidance could be used successfully to wander around a simple maze successfully  The premature but cunning plan to build a back-garden maze is already underway.

 

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.