Mavic Air - Fly Away - Does not RTH even with strong GPS signal

[Droniee7788]

Hello All,

Appreciate if someone could shed some light here. A pretty unhappy day yesterday flying my Mavic Air in a bush area. The Air set HOME POINT, flew for just about 2 minutes and unfortunately, I was looking at the mobile and it flew behind a hill. It lost signal and never RTH. I pressed the RTH button, no luck an I walked around to see if I could get signal and no luck. So, after 40 mins, I lost the Air (pretty angry with myself). I've flew in this exact location before (with my SPARK) and everytime it losses signal, it had always RTH.

Whilst, its my fault for not maintaining line of sight and letting it flew behind a hill. I came back home and view the flight log. It confirmed there was strong interference signal, but RTH was establised and there was GPS signal all the way through. So, my question was why it never tried to fly back home?

Here is the log attached and appreciate if anyone could shed some light, it would be appreciated.

Flight Log

Regards

 

[lannes]

Maybe it crashed into the hill or trees on the RTH.

What height was your RTH set at ?

 

[BadWolf1]

Hello All,

Whilst, its my fault for not maintaining line of sight and letting it flew behind a hill. I came back home and view the flight log. It confirmed there was strong interference signal, but RTH was establised and there was GPS signal all the way through. So, my question was why it never tried to fly back home?


Flight Log

Regards

There are others far more expert than me at pulling logs apart. However I downloaded it and looked at the end of the flight. It's pretty easy to spot a crash because the drone's attitude changes dramatically when at the end.

What do we see. ?
The dark blue line near the bottom is roll you've: got a lot of roll on, and full aileron - that's the khaki coloured line which comes off the bottom after 145 seconds and goes to zero when the log stops.
That roll line shows drama just before the end of the flight, so I know what I'm thinking ...
At 144.997 seconds roll is -11.8 degrees, by 145.107 it was +7.9. So it rolled 20 degrees in 0.11 of a second. With no control change.
There was a smaller change in pitch at the same moment.
The drone doesn't stop dead and rolls back. I know at a certain amount of roll there is a cut out mechanism - you can use it for hand landing, and the last data point is -14.2 degrees of roll. It's possible that triggered a shutdown (greater expertise needed!)
I'd say you flew sideways into something and looking at the plot on a satellite photo a tree would be favourite. You started at 142M and ended at 260M so a very tall tree and/or rising ground or it's some kind of mid air collision (bird strike ?). There's no drop in height in those last two seconds and the drone keeps moving in roughly the same direction, so going into the branches of a tree would fit.

 

[slup]

Nothing in the log shows any signs of impact ... the small roll & pitch change in the end of the log comes from you slightly letting off the Aileron & Elevator sticks ... probably when you notice the live feed lag.



The flight was at 120m height above the HP & the RTH height was set at 40m ... so in case of a failsafe RTH the AC would have remained at 120m. But looking into the geographical conditions the reason for that the AC never came home becomes apparent ... the AC crashed into the hill side which was right in the RTH route.

 

[Droniee7788]

Thank you Iannes, BadWolf1 and slup for your feedback and analysis. It explains that it crashed sideways (no obstacle avoideance censors) during the time it lost signal (so could not see it on the mobile) and proves the importance of flying within line of sight. Expensive mistake! But nonetheless lessons learnt.

Can I please ask how did slup get the 3D terrain map and overlay the flight path?

Thanks all for your assistance.

 

[Detritalgeo]

If you can access that hill.. I would try a scavenger hike..

 

[Meta4]

Thank you Iannes, BadWolf1 and slup for your feedback and analysis. It explains that it crashed sideways (no obstacle avoideance censors) during the time it lost signal (so could not see it on the mobile) and proves the importance of flying within line of sight. Expensive mistake! But nonetheless lessons learnt.

The drone would have gone very close to clearing the hilltop, except for the trees.

If you want a challenge with a handheld GPS unit, swim across the creek, and it's a 530 metre climb to the loss of signal point at -33.65105° 151.27209°
Walk towards -33.65232° 151.27202° 140 metres south.
The mavic is probably halfway there but might have crashed further north if there's a high enough tree to snag it.
Or there's a sealed road about 870 metres to the west.

 

[DanMan32]

Didn't even the original Air have the feature of going over obstacles encountered on RTH? Unless it reached max altitude in which it wouldn't be able to go over the obstacle. It would in that case hover in front of the obstacle until the battery runs out.

 

[Dakarenduro]

Thank you Iannes, BadWolf1 and slup for your feedback and analysis. It explains that it crashed sideways (no obstacle avoideance censors) during the time it lost signal (so could not see it on the mobile) and proves the importance of flying within line of sight. Expensive mistake! But nonetheless lessons learnt.

Can I please ask how did slup get the 3D terrain map and overlay the flight path?

Thanks all for your assistance.

I dont think it crashed sideways due to a lack of obstacle avoidance sensors. In RTH it would be traveling forward and most likely got caught in a tree as noted.

 

[slup]

... It explains that it crashed sideways (no obstacle avoideance censors)

No it didn't crash sideways ... the AC was on it's way to the HP by means of failsafe RTH due to lost signal ... but looking at the sloped terrain & how the AC approached it, adding the vegetation there several scenarios may be possible. The highest point on that slope along the RTH route was according to the heights from GE approx. 115m above your HP, leaving 5m between the AC & the bare ground at the slope. So either some tree got right in the way & was sensed by the OD & the AC got stuck there & battery forced landed eventually. Or the AC slowly got closer & closer to the tree tops with finer branches coming in to the OD sense area from the right and those wasn't big enough to be noticed by the OD & the AC clipped them and went down.

...Can I please ask how did slup get the 3D terrain map and overlay the flight path?

If you upload the .TXT log either to PhantomHelp.com or AirData.com you have a possibility there to download .KML files which you can open in Google Earth ...

PhantomHelp.com



AirData.com




*EDIT*

Can also add below if you want to hike up to try to find it ... search off between the pin coordinates. If it's really steep there the AC may have rolled down though.

 

[Detritalgeo]

Great info, I am sure someone will take this geo-caching challenge.. But I hope it's the true owner and not a clever margin hunter.

 

[BadWolf1]

Nothing in the log shows any signs of impact ... the small roll & pitch change in the end of the log comes from you slightly letting off the Aileron & Elevator sticks ... probably when you notice the live feed lag.

Not wanting to argue, but being curious and much less experienced at looking at logs, I wonder if you can explain that conclusion.

Control time-line is at 107 seconds aileron (blue on the chart below) increases to -100% where it stays until 145.3 when it goes back to the 85% range before release at 147.2.
There is significant roll before 107 seconds and it goes to -14.3 at 112 seconds and says between -14.3 and -11.3 until 145.0. So it looks to me like the drone was flying sideways for the last third of the data.
Then there is some event at ~145.0 when roll (green, below) goes from -11.8 degrees to +3.6 between two samples 0.1 seconds apart just before the stick was [partly] released. There's a smaller "jag" in pitch (red, below) at that time.




To me, "drama" very close to the end recorded data suggests a crash. A large change in attitude (red/green) with no control change (blue) would would be a drama. I picked up the horrible term "Controlled Flight Into Terrain" (CFIT) after it applied to a friend's death. As a novice investigator it's the term I'm reaching for here. [But as I'm at pains to point out happy to learn why I've got it wrong]

 

[slup]

Not wanting to argue, but being curious and much less experienced at looking at logs, I wonder if you can explain that conclusion.

Control time-line is at 107 seconds aileron (blue on the chart below) increases to -100% where it stays until 145.3 when it goes back to the 85% range before release at 147.2.
There is significant roll before 107 seconds and it goes to -14.3 at 112 seconds and says between -14.3 and -11.3 until 145.0. So it looks to me like the drone was flying sideways for the last third of the data.
Then there is some event at ~145.0 when roll (green, below) goes from -11.8 degrees to +3.6 between two samples 0.1 seconds apart just before the stick was [partly] released. There's a smaller "jag" in pitch (red, below) at that time.


View attachment 103925

To me, "drama" very close to the end recorded data suggests a crash. A large change in attitude (red/green) with no control change (blue) would would be a drama. I picked up the horrible term "Controlled Flight Into Terrain" (CFIT) after it applied to a friend's death. As a novice investigator it's the term I'm reaching for here. [But as I'm at pains to point out happy to learn why I've got it wrong]

Depending on the scale you have there it can visually seems to be a "drama" ... but looking into the changes in real figures, there are a change in roll of 19,6 degrees & a slight up pitch of 7,8 degrees ... all this together with a ease off of both aileron & elevator sticks ... & a yaw angle that doesn't move at all. To me that's not a "drama", if something would have happened there the movements had been much larger & in all 3 axis.

 

[sar104]

Not wanting to argue, but being curious and much less experienced at looking at logs, I wonder if you can explain that conclusion.

Control time-line is at 107 seconds aileron (blue on the chart below) increases to -100% where it stays until 145.3 when it goes back to the 85% range before release at 147.2.
There is significant roll before 107 seconds and it goes to -14.3 at 112 seconds and says between -14.3 and -11.3 until 145.0. So it looks to me like the drone was flying sideways for the last third of the data.
Then there is some event at ~145.0 when roll (green, below) goes from -11.8 degrees to +3.6 between two samples 0.1 seconds apart just before the stick was [partly] released. There's a smaller "jag" in pitch (red, below) at that time.


View attachment 103925

To me, "drama" very close to the end recorded data suggests a crash. A large change in attitude (red/green) with no control change (blue) would would be a drama. I picked up the horrible term "Controlled Flight Into Terrain" (CFIT) after it applied to a friend's death. As a novice investigator it's the term I'm reaching for here. [But as I'm at pains to point out happy to learn why I've got it wrong]

It might be easier to visualize in the aircraft frame of reference:





It was moving forwards and left at the end of the log, due to forward elevator and left aileron. The roll excursion starting at 145 seconds was in response to a decrease in left aileron.

 

[BadWolf1]

View attachment 103947

It was moving forwards and left at the end of the log, due to forward elevator and left aileron. The roll excursion starting at 145 seconds was in response to a decrease in left aileron.

Still trying not to argue. but the changes in velocity_right, and in roll happened 0.3 seconds before the change in aileron. It's hard to see on the chart above, it's easier to see if you look at just the last 5 seconds.
Do we have to allow for the control-input log and speed/attitude log not being perfectly sync'd? I was seeing a big roll excursion happening first and picture disturbance causing the pilot to partially then fully release the controls.

 

[sar104]

Still trying not to argue. but the changes in velocity_right, and in roll happened 0.3 seconds before the change in aileron. It's hard to see on the chart above, it's easier to see if you look at just the last 5 seconds.
Do we have to allow for the control-input log and speed/attitude log not being perfectly sync'd? I was seeing a big roll excursion happening first and picture disturbance causing the pilot to partially then fully release the controls.

Good spot. And there is a pitch excursion at the same time:





The stick data would not be out of sync enough to account for that, so it is possible that there was a brief impact with something - a branch perhaps. The DAT file would be needed to explore that further.

 

[slup]

My thinking is ... depending that the AC is going with the left side first, i.e. roll is on negative side ... if then something, a branch, disturbed the left rear or front prop we would have seen a larger initial roll towards the same negative side, but we don't. Instead the roll angle moves positive from a mainly unchanged angle ... & to this not a slightest change in the yaw.

What we see as a slight time diff. between change in stick commands & AC movements I think can be explained in a rapidly deteriorating RC-AC connection. Have seen it in the logs from my own MA1 ... with full stick command, even though the log shows briefly 100% connection in between average connection, the AC shift position & speed slightly. This as if the stick command gets some latency there ... sometimes it can even be seen in the video recording.

But as said ... the DAT will for sure show what this is.