M3 descent into the sea (near-miss)

[Mavic 3 by the Sea]

Hi guys,
Recently, my M3 almost committed suicide in the sea. I have not managed to figure out why, so your help would be much appreciated. Thanks!

I was flying a straight line out to sea at an altitude of 1.5m above the surface.
Wind and waves were minimal, no gusts, no wind warnings, no physical impediments such as boats etc. Beautiful, bright, peaceful morning.

RC-N1 with iPhone SE2020 running
DJI Fly App: 23 satellites, obstacle avoidance on & set to bypass & normal (not nifty), flight in normal mode.

At a distance of 1000 meters from me / home point, I slowed down and then hovered to adjust camera settings. No physical obstacles anywhere nearby.

Just after starting to hover, the M3 autonomously began to descend from 1.5m above sea level downwards. Just before contact with the water, I left camera settings, pulled up to 11m and narrowly avoided my M3’s (autonomous) suicide in the sea.

So there are 2 main questions here:

Q1: why did the M3 fail to recognise the water surface? (This has worked perfectly during hours of low altitude flight above the water during the past months & in same conditions, before this incident).

Q2: why did the M3 autonomously commence descent?

I did not rth after this but flew & filmed at up to a distance of 1.6 km from hp, until battery commanded rth 30 min later.

During this further flight, 3 more things were odd:

1. Drone wasn’t steady but wabbled badly in Master Shots; at the same time, however, it was steady in standard Photo or Video settings.

2. (permanent issue): flight in N-mode was/keeps being displayed as „P-mode“ in top left corner of Fly App‘s camera view. This has been going on since purchase. Going by tutorial videos for M3, this is an anomaly.

3. Auto RTH at 60m altitude worked fine, drone descended in normal mode perfectly above the home point. But: it came blasting down as if (again?) it did not recognise the surface below (distinctive rocky terrain).
I paused RTH at ca. 5m altitude & completed manual palm landing instead.

During the past months and hours of flying just above the water, and with the general exception of the constant „P-mode“ issue, I never experienced problems like these.

Can anyone make sense of these these issues?

Many thanks!

 

[Meta4]

Can anyone make sense of these these issues?

Not without seeing the recorded flight data, which will probably explain the incident..

Go to DJI Flight Log Viewer | Phantom Help
Follow the instructions there to upload your flight record from your phone or tablet.
That will give you a detailed report of the flight.
Come back and post a link to the report it gives you.
Or .. just post the txt file here.

Flying 1.5 metres above the sea is always going to be risky.
The sea surface is always rising and falling to some degree and with the Mavic 3 rate of descent of 6 metres/sec, you don't have much safety margin anyway.

 

[JooB]

Flying so close to the water surface is always tricky. Especially 1000m away where you can't see if the drone moves up or down. DJI recommends a minimum height above water of 2m, if you go lower you should turn off the VPS (How to Fly Safely Over Water).

Due to light reflections on the water surface the sensors might not work as designed, I suppose that's what happened.

Or, like I did some weeks ago without noticing it, you moved your stick minimal down and the drone started to decline slowly.
Look at my video starting at 1:25. Just before I pulled the drone up at 1:30 I noticed it was declining before without my intention and very close to the waves already. I'm sure it was my own stick mismanagement

To view this content we will need your consent to set third party cookies.
For more detailed information, see our cookies page.

Accept third party cookies

In any case the flight log will show you what really happened.

 

[hiflyer201]

Due to the curvature of our planet, I’m not sure I would try to fly at that altitude. To ad to that thought I often get erroneous altitude indications at low altitudes. I often see posts that say those erroneous altitude indications are due to barometric changes after launch, but as a long time barometer watcher I say that’s a crock. At any rate that is too low to fly at that distance in my opinion my very humble opinion

 

[vindibona1]

Other replies have noted the issue with the vision system above water and the minimum heights that one should observe when flying over water. I think there are couple more lessons to be aware of because water, after all, is a "fluid" situation .

The first thing is that the altitude one shows from the launch point vs the altitude over water isn't always the same and I will presume usually different as water levels rise and fall, often second to second, so IMO it is important to have visual contact with the drone because the altitude on the controller often doesn't tell you exactly what you need to know. Second, I think one lesson of this anecdote is, that if you have to hover for whatever reason, take the drone up a few feet/meters for insurance, especially if you're focused on settings rather than flying.

It would also appear that the stronger the forward motion of the drone the less likely it's going to just fall down and into the water. I confess I'm always nervous when flying close over water, having heard so many stories of drowned drones where the vision sensors confused it or waves just came up and swallowed it.

 

[Garatshay]

Hi guys,
Recently, my M3 almost committed suicide in the sea. I have not managed to figure out why, so your help would be much appreciated. Thanks!

I was flying a straight line out to sea at an altitude of 1.5m above the surface.
Wind and waves were minimal, no gusts, no wind warnings, no physical impediments such as boats etc. Beautiful, bright, peaceful morning.

RC-N1 with iPhone SE2020 running
DJI Fly App: 23 satellites, obstacle avoidance on & set to bypass & normal (not nifty), flight in normal mode.

At a distance of 1000 meters from me / home point, I slowed down and then hovered to adjust camera settings. No physical obstacles anywhere nearby.

Just after starting to hover, the M3 autonomously began to descend from 1.5m above sea level downwards. Just before contact with the water, I left camera settings, pulled up to 11m and narrowly avoided my M3’s (autonomous) suicide in the sea.

So there are 2 main questions here:

Q1: why did the M3 fail to recognise the water surface? (This has worked perfectly during hours of low altitude flight above the water during the past months & in same conditions, before this incident).

Q2: why did the M3 autonomously commence descent?

I did not rth after this but flew & filmed at up to a distance of 1.6 km from hp, until battery commanded rth 30 min later.

During this further flight, 3 more things were odd:

1. Drone wasn’t steady but wabbled badly in Master Shots; at the same time, however, it was steady in standard Photo or Video settings.

2. (permanent issue): flight in N-mode was/keeps being displayed as „P-mode“ in top left corner of Fly App‘s camera view. This has been going on since purchase. Going by tutorial videos for M3, this is an anomaly.

3. Auto RTH at 60m altitude worked fine, drone descended in normal mode perfectly above the home point. But: it came blasting down as if (again?) it did not recognise the surface below (distinctive rocky terrain).
I paused RTH at ca. 5m altitude & completed manual palm landing instead.

During the past months and hours of flying just above the water, and with the general exception of the constant „P-mode“ issue, I never experienced problems like these.

Can anyone make sense of these these issues?

Many thanks!

There will be many answers to your query - but only an examination of the flight record will provide information on what the drone 'thought' it was doing. I had a similar experience with a Mavic 2 Pro not long ago and I formed the opinion that, if you are descending carefully close to any surface (including water, if it's not crystal clear), the drone believes that the intention is to land and it goes into that mode.

 

[MS Coast]

This is an old thread, but I'll add one thought to consider when flying over water.

The OP states that he was flying 1.5 meter above the water. But the water isn't always a flat plane. The Bay of Galway has substantial swells rolling in from the North Atlantic Oceans, even on days when the local winds are light. I suspect that the drone just went into landing mode when it reached swells that reached upward enough to trigger it.

Wind Forecast: wind speed & gusts — Windy.app

Wind direction is North, wind speed varies between 6.7 and 6.7 mph - get wind forecast: detailed wind speed and wind direction information with a leading pro weather app - Windy.app

windy.app

 

[Mavic 3 by the Sea]

1/3

Hi guys, a big thank you to all who took the time to respond! A lot of truly helpful experiences and information.

Since the OP, I looked at the flight log as suggested here, did some experiments and clocked up more hours of flying. Based on that, I’ll try to provide the most accurate reply in relation to the specific flight in question and my experience since.

Re stick inputs:

— Both the post flight visualisation on DJI Flight Log Viewer | Phantom Help and the ‘raw’ log data showed no stick inputs during the period in question at all.

Re signal loss:

— no signal loss at the time & RTH on signal loss was set

— homepoint was correctly set (point of departure)

— 27 satellites / GPS functional

Re wind / waves / swell:

— review of the footage confirmed it was a beautiful calm morning with minimal wind, waves and swell. The drone autonomously commenced & continued a linear descent into the water.

Long story short:

— For the exact moment in which the speed turned zero and the drone should have begun to hover, the log shows that

- the VPS turned from off to on

- the IMU and VPS altitude readings differed

- the VPS (12.1 ft) read an altitude 7ft above the IMU (5.2 ft)

— From this moment onward and while speed remained unchanged at 0 mph,

- the drone commenced its decent, measured by both VPS and IMU, albeit reading different values

— During descent both the absolute and relative difference between the VPS and IMU readings declined, (with the higher VPS reading obviously converging on the lower IMU reading rather than vice versa given the drone’s descent)

Interpretation based on footage as well as VPS & IMU log readings of altitude:

— the IMU reading was correct

— the VPS reading was substantially incorrect

— the drone’s autonomous descent appears like an attempt to synchronise the VPS altitude reading with the IMU’s altitude reading by way of autonomous change of position

— that attempt was increasingly ‘successful’ as reflected in the convergence of the VPS reading on the IMU reading

— unfortunately, however, & despite this convergence, the VPS reading (3ft) and IMU reading (1ft) were still not identical at 1ft above sea level (IMU reading remained correct based on footage)

— and they would not have been identical before the drone would actually have touched the water.

 

[Meta4]

Interpretation based on footage as well as VPS & IMU log readings of altitude:

— the IMU reading was correct

— the VPS reading was substantially incorrect

— the drone’s autonomous descent appears like an attempt to synchronise the VPS altitude reading with the IMU’s altitude reading by way of autonomous change of position

— that attempt was increasingly ‘successful’ as reflected in the convergence of the VPS reading on the IMU reading

— unfortunately, however, & despite this convergence, the VPS reading (3ft) and IMU reading (1ft) were still not identical at 1ft above sea level (IMU reading remained correct based on footage)

— and they would not have been identical before the drone would actually have touched the water.

You've misunderstood the relationship between IMU alt and VPS altitude.
VPS is very unlikely to be wrong and the drone will never attempt to synchronise them brcause they measure two different things.

It would be a good idea to post your flight data for comment.

 

[Mavic 3 by the Sea]

2/3

 

[Mavic 3 by the Sea]

You'misunderstood the telationship between IMU alt and VPS altitude.
VPS is very unlikely to be wrong and the drone will never attempt to sunchronise them brcause they measure two different things.

It would be a good idea to post your flight data for comment.

I beg to differ, see screenshot from M3 manual and post 3/3 to follow below.

 

[Meta4]

I beg to differ,

You might, but you would be wrong.
Post your data if you want to find out what actually happened.

A difference of 1 or 1000 feet in the VPS and IMU altitudes means nothing.

 

[Mavic 3 by the Sea]

3/3

On another calm and sunny day one month after the incident described here, I experimented over the sea by repeating the manoeuvre above with the same settings; but this time in relatively close proximity (ca. 30m) / excellent visual conditions.

The drone autonomously attempted to descend into the water once again.

Given the section of the M3 manual above and the “Flying over water” video by DJI, I then switched off the GNSS / obstacle avoidance and did a repeat of the experiment.

Et voilà: the drone remained in perfect hovering position at the same altitude over the sea under otherwise unchanged conditions.

So it’s unambiguous: it’s the VPS / GNSS / obstacle avoidance over water.

 

[Mavic 3 by the Sea]

Because I know what I'm talking about rather than making random guesses without an understanding of what the data shows.

Jesus Christ. “…Because you said so.

 

[Tufargon]

Why did you ask for help? Meta4 is about the best there is at analizing flight data.

 

[Mavic 3 by the Sea]

Not because he said so, but Meta4 is one of the BEST at reading data. As he suggested, POST YOUR LOG!!!

Read 3/3. That’s called an experiment. Everything stays the same except the treatment variable x. If the outcome differs, the cause is x. It’s called scientific methodology.

 

[Meta4]

Read 3/3. That’s called an experiment. Everything stays the same except the treatment variable x. If the outcome differs, the cause is x. It’s called scientific methodology.

You could have had a proper analysis of your data 4 months ago when I suggested posting data.
Instead, you waited 4 months to show your lack of understanding of the scientific method or how your drone and this forum works.

 

[Tufargon]

“Because you said so.”

Did anyone have to place credentials before responding to you? I've seen what Meta4 can do at analizing flight data for quite a while and it's quite impressive.
And insulting someone with a political jab is just, so not cool and against the forum rules.

 

[Meta4]

— the drone’s autonomous descent appears like an attempt to synchronise the VPS altitude reading with the IMU’s altitude reading by way of autonomous change of position

If you had even a basic understanding of how drones work, you'd know this is impossible.
You need to first understand what the IMU altitude and VPS altitude are measuring.
Clearly it's not what you think.

3/3

On another calm and sunny day one month after the incident described here, I experimented over the sea by repeating the manoeuvre above with the same settings; but this time in relatively close proximity (ca. 30m) / excellent visual conditions.

The drone autonomously attempted to descend into the water once again.

Which proves that something caused the descent, but doesn't give us any clues as to what it might have been.

Given the section of the M3 manual above and the “Flying over water” video by DJI,

Which is vague and misleading, it doesn't say that water will suck your drone down (because it won't).
The manual is referring to the difficulty the VPS sensors have maintaining horizontal position when over water.

I then switched off the GNSS / obstacle avoidance and did a repeat of the experiment.

You cannot switch off GNSS (do you even know what it is?)
You can switch off obstacle avoidance, but that is unrelated to the VPS sensors which are always on and cannot be disabled.

Et voilà: the drone remained in perfect hovering position at the same altitude over the sea under otherwise unchanged conditions.
So it’s unambiguous: it’s the VPS / GNSS / obstacle avoidance over water.

Not at all.
All you've observed is a coincidence, and if you understood some drone basics, you'd also understand that your assessment is is incorrect because the mechanism you've guessed caused your incident couldn't have had the effect you think it did.

You’re neither reading the posts nor the data above, nor the M3 manual, nor are you taking note of DJI’s further instructions on the matter, nor have you understood what an experiment is.

I've read and reread your posts, as well as the fragment of data you included.
I'm well acquainted with the manual and even understand the vagaries and misinformation in DJI's "further instructions".
I even have a very good idea what an actual experiment is, but haven't seen any evidence of a valid one in your posting.
I also have years of experience analysing the data from flight incidents and have several thousand kilometres experience flying drones over the sea.

I'm also familiar with the community guidelines for this forum and would suggest you acquaint yourself with them.

 

[Meta4]

If you change a single setting / treatment variable in a multivariate ‘equation’, and this coincides with a change in the outcome variable, you may have yet only shown a coincidence rather than already proven causation; but if you are able to repeat each experimental process and observe the corresponding change in outcome multiple times, you have indeed found evidence of a specific mechanism.

The flight data that you are avoiding sharing, records 184 parameters for every 1/10th of a second of the flight.
So you turned something (that couldn't have the effect you are sure it did) but have you checked any of the other potential factors?

Deliberate sarcasm, almost ad hominem. Nobody said or implied that water ‘sucked down the drone’….

VPS somehow lowering a drone towards water is a common myth and one that you suggested caused your incident.

Here we go, Narnia is in sight! The Lord be blessed!

You can't sidestep the very well known issue that it is not possible to turn off the VPS sensors - that you believe are the key to your incident.
That you are unaware of this underscores your very limited understanding of your drone.

That’s what I mean with “because you say so.” If you apply standards, apply them to yourself as well. I have seen zero evidence for your very pointed argument so far.

You've given plenty of evidence of your appalling lack of basic drone function.
A couple of members have chimed in to give you a hint about something you'd have known if you had spent any time here.

If you are/were as good as you think you are, you could, for example, have formulated hypotheses or an educated guess.

I follow the scientific method.
I look for evidence before I venture an opinion on what might have caused an incident.
And the place to look is in the recorded flight data.

I've gone over this way too much, but enough for other members to see that you have no idea what you are talking about and ignore you on this topic.
I'm done with your particularly uninformed trolling.