I've spent the past 4 hours going through the logs of my tricopter crash with both MP and APM Planner trying to figure out why it went down and I'm stumped. I'm decent with log interpretation and try to help out others around here when I can, but this is apparently above my pay grade.
The tricopter in question runs an APM 2.5 with external mag and U-Blox GPS. Really low compassmot and good vibration levels. I check its logs after most flights and never see anything unusual.
Today, on its second flight of the day, I was testing a mapping camera doing mild maneuvering. I was descending in AltHold mode at a fast descent rate with moderate forward speed about 400' away from me. I then threw the switch putting it in Loiter mode. I expected to see it stop its forward motion but it didn't and its attitude appeared odd (it was hard to tell at the distance). I attempted control but it didn't seem to respond and sped up horizontally while still rapidly descending. I punched the throttle to full, but it continued down, impacting an adjacent hillside 6.8 seconds after putting it in Loiter. Unfortunately I was too surprised and flustered to remember to just switch back into Stabilize, which probably would have saved it. There was a fair amount of damage to the frame, but it looks like the expensive bits survived.
My first thought was that it was some sort of mechanical/electrical failure. This idea was bolstered when I played back the recorded FPV/OSD video and it showed the tri had pitched so far up the horizon was no longer visible in the moments before impact. However...
Review of the log shows that there were extreme pitch maneuvers, however they were requested by the APM. At line 10045 Loiter begins. Going into Loiter the tri is in a fairly level position, although descending. Immediately the APM requested a large positive pitch and the tricopter seemed happy to oblige. Within a few lines, CTUN shows DPit to hit the limit of 45 degrees and the tricopter is just behind it. For some time the tricopter is kept around a +45 degree pitch by the APM and impact occurred around line 10227. The tri also responded just fine to the DRol commands so that seems to rule out something mechanical/electrical. The tricopter did exactly what was asked of it by the APM. HDOP and the GPS sats were good, so GPS glitchiness appears unlikely. The BarAlt and RelAlt track well until just after impact, which would be expected.
The only divergence I could find was between CRate and DCRate. The DCRate was very positive, but the CRate was negative. Of course this was during the period it was pitched at 45 degrees and the throttle was at its 80% limit (due to still being in Loiter) so perhaps it can't maintain altitude in that position. I also see in the log that the GPS speed began increasing considerably while this was going on. There also may be something in the NTUN DVel and Vel values, they look odd, but I don't know enough of them to fully interpret them.
So...Can anyone expert in log analysis have a look at this and give me guidance? This seems to be a hard one. Why the hell was the APM requesting such a large pitch?? My fear is that I may have an obscure tuning value out of whack that prevents recovery while engaging Loiter in some flight conditions. Any help would be very appreciated.
Replies
Seeing as how it's pretty much been the sound of crickets in response to my query, I thought I'd offer up my possibly correct interpretation of what happened after a lot of additional thought. I suspect I got into a portion of the operating envelope I couldn't get out of in time to avoid hitting the hillside.
As the tricopter was rapidly descending in AltHold mode it had respectable forward speed when I foolishly threw it into Loiter. My transitions into Loiter usually result in strong jerks even at low speed and this was much faster than usual. In order to hold position in face of the horizontal speed, the tricopter requested full pitch up, which is limited to 45 degrees. At that angle the vertical lift component was about 71% of what the three rotors were putting out. At this time I was still maintaining the throttle setting for the previous rate of descent as I didn't see the situation I was putting the tricopter in. Given that my vertical lift had decreased due to the pitch angle, my descent rate actually increased.
Being confused about the high pitch angle I was seeing the tricopter in, I set the desired climb rate from negative to about zero, thinking it would halt its descent. It helped, but wasn't enough. Then I punched the throttle to full. However being in Loiter, the throttle out is limited to 800. If I take 0.80 x 0.707 I get about 56% of generated lift in a vertical direction. That's just barely enough to keep it hovering, much less stop a rapid descent. Hillside, meet tricopter.
Clearly my solution would have been to switch the mode to Stabilize and punch the throttle to full. With the exception of flying under a canopy of trees, it would be a way out of most problems. In fact I'd love a panic button on my RC transmitter that combined those two functions with a single press.