So there I was... flying a DJI F550 with an APM 2.5 over a bluff by the Pacific Ocean. We were carrying a Gopro on the gimbal. The purpose of this flight was to be a weather ship to scout the conditions before launching a CS8. I took off with no problems. There was a little bit of gusty winds over the bluff which tapered out as the copter climbed out of the dome effect. I put it in Loiter mode and it seemed to be pretty stable. I entered a left yaw to pan around the bluff. Right around a 45 degree yaw to the left the copter entered a rapid descent. I went to max throttle to attempt to allow the copter to elevate in loiter mode. I went to max throttle position and the copter was still descending by the time I recognized an out of control situation and killed loiter mode it was too late. The copter was below the cliff out of my line of sight. Luckily the damage was minimal for crashing into a cliff. It managed to crash into a soft sandy culvert, only broke one arm, snapped two props and got sand in a few motors. The Gopro was unscathed. After reviewing the GoPro footage I here the motors still spooled up but I can here them struggling. It sounded like the copter was trapped in its own wash and the autopilot was unable to climb out of the wash. The copter stayed level throughout the entire descent until it crashed. It took about three seconds from descent to impact.
I typically fly this thing in New Mexico where it is 5500 feet higher than the pacific ocean in California. It seems to be pretty forgiving in and out of its prop wash where the air is thinner. I figure that, had I realized that I was in an out of control situation I would have taken it out of loiter and flown out of it.
The question of the day is why would the FC not sense this change and apply a rapid correction? It seemed that the autopilot allowed it to get into a uncontrolled state was lagging in its ability to correct for the change until it was too late. I am guessing that my PID settings are under-compensating in loiter mode, so what settings should I focus on to make theses functions more robust?
I am flying APM 2.5 on Arducopter 2.8 firmware
Replies
Just a few quick thoughts.
You didn't mentiton it, but were you using a SONAR and below 26 feet. if so, your result of flying over bluff would be normal the copter descends descends till it is at same height above ground level. Probably not, but worth a mention.
Sometimes for flights where actual ground level is quite variable, SONAR can be more of a liability than an asset.
However, manual override of any altitude holding mode (like Loiter) is not the same as normal throttle control and it basically just gives it a bump upwards, clearly not what you needed to escape from this situation.
Quite possibly rapid switching back to stabilize was the best move and living here on the Pacific Coast it is something I have had to employ frequently.
On the coast we usually have an onshore breeze coming in from the ocean over the bluffs, this produces an updraft at the very bluff edge, but the oncoming wind shears that and turns it into a rotor which has a downdraft component as high as the initial updraft. If SONAR was not the culprit, you quite possibly ran into the rotor and quite possibly your copter or the built in response rate was not high enough to compensate for it. Rotors are both the friend and the bane of hang glider pilots. You will often see the Sea Gulls gliding along the top of the rotor.
To compound matters, the updrafts / downdrafts produce local air pressure changes which can be interpreted by your barometer (altimeter) incorrectly leading to the copter into trying to go to where it thought it was supposed to be.
If you are going to try operating around coastal (or any) bluffs in the future try it on a truly windless day and make sure your SONAR if you have one is not enabled.
Of course if it was a low battery as Chris suggests, none of this helps.
Any chance your battery was running out? That's the usual cause for decent like that.
Can you post the video and the log file please?