"Over Water" mode for drones

[Zbip57]

The problem arises when there is a sudden change in barometric pressure and the drone can suddenly descend or ascend by a couple of meters. Now normally over land the downward sensors will kick in and avoid a mishap. However over a smooth surface like water we all know that they do not work.

The downward sensors do work, whether over water or over a smooth surface. It's just important to understand HOW they work.

A perfectly smooth surface of water only means the downward facing camera sensor might not be able to fix on a particular feature for accurately holding position when in a hover. But if you're flying over a lake, presumably the drone will have an unobstructed view of GPS satellites, so won't need the downward vision sensor for position holding.

In any case, that sensor is only doing anything useful while the drone is in a hover. It cannot in any way drive the drone into the water while it is in forward flight.

Where the position sensor might get confused is if there are waves or ripples moving across the surface of the water. While in a hover and fixating on a moving surface feature, the drone may try to follow that moving feature. Again, that cannot cause the drone to sink into the water as the vision sensor has nothing to do with holding altitude.

The barometric sensor is the primary sensor used for altitude-hold, and it is sensitive to changes in atmospheric pressure. Your takeoff altitude is initially recorded as zero, but will often register something completely different when you eventually return to land at the same spot.

And yes, air temperature and pressure does change during the day, with land usually warming and cooling faster than large bodies of water. When I sailed on Lake Ontario winds might be dead calm early in the morning. But as the land heated up, warm air rose from the farm fields drawing cooler air onshore from the lake. Later in the day, as the water surface eventually warmed up, the wind would die down completely. Even later, as the land cooled off more quickly, warmer air rising off the lake would pull the cooler air from the land blowing offshore. But that process is very gradual.

I guess it's not impossible, but any sudden temperature/pressure gradient, sufficiently affecting your barometric sensor to drive your drone into the water, is also certain to create a significant sudden wind shear.

In any case, it's not the downward facing sensors that's causing that issue. Whether the surface is perfectly smooth, or rippled and wavy, the infrared height sensors (when in range) bounce a signal off the surface. If the drone senses it is getting too low, too close to the surface, it will automatically increase height. It won't cause the drone to sink into the lake.

That infrared sensor is not effective at high rates of forward speed because the forward pitched angle of the drone's body essentially means the sensor is looking at ground that's already passed behind it. So if you're flying forward at high speed low over water, you risk getting slapped by a wave.

I've seen people theorize that the infrared sensor might be seeing through the surface of clear water and registering its height as measured from the bottom of the lake, rather than from the surface of the water. So if the barometric sensor is somehow causing the drone to sink, the theory is that infrared sensor thinks it's still plenty high enough off the bottom of the lake and doesn't see the surface of the water. Even in that case, it's not the infrared sensor that's pulling the drone down.

I suspect that's nonsense anyway. Infrared light won't penetrate a pane of window glass, so I'm pretty sure it doesn't go through water either.

 

[Zbip57]

I've seen people theorize that the infrared sensor might be seeing through the surface of clear water and registering its height as measured from the bottom of the lake, rather than from the surface of the water. [...]

I suspect that's nonsense anyway. Infrared light won't penetrate a pane of window glass, so I'm pretty sure it doesn't go through water either.

Now that I think about it, that should be relatively easy to test out. The phantomhelp.com log viewer shows the "IMU Altitude" and the "VPS Altitude". If the infrared sensor can see through the surface of the water all the way to lake bottom, when flying low and level out over the lake, that should show up as an obvious difference when comparing the two altitude numbers. The IMU (barometric) altitude would stay relatively constant while the VPS altitude grows as the lake bottom falls away. But, if the infrared signal only bounces off the water's surface, then there should be no significant divergence between the two numbers.

 

[AZDave]

The downward sensors do work, whether over water or over a smooth surface. It's just important to understand HOW they work.

A perfectly smooth surface of water only means the downward facing camera sensor might not be able to fix on a particular feature for accurately holding position when in a hover. But if you're flying over a lake, presumably the drone will have an unobstructed view of GPS satellites, so won't need the downward vision sensor for position holding.

In any case, that sensor is only doing anything useful while the drone is in a hover. It cannot in any way drive the drone into the water while it is in forward flight.

Where the position sensor might get confused is if there are waves or ripples moving across the surface of the water. While in a hover and fixating on a moving surface feature, the drone may try to follow that moving feature. Again, that cannot cause the drone to sink into the water as the vision sensor has nothing to do with holding altitude.

The barometric sensor is the primary sensor used for altitude-hold, and it is sensitive to changes in atmospheric pressure. Your takeoff altitude is initially recorded as zero, but will often register something completely different when you eventually return to land at the same spot.

And yes, air temperature and pressure does change during the day, with land usually warming and cooling faster than large bodies of water. When I sailed on Lake Ontario winds might be dead calm early in the morning. But as the land heated up, warm air rose from the farm fields drawing cooler air onshore from the lake. Later in the day, as the water surface eventually warmed up, the wind would die down completely. Even later, as the land cooled off more quickly, warmer air rising off the lake would pull the cooler air from the land blowing offshore. But that process is very gradual.

I guess it's not impossible, but any sudden temperature/pressure gradient, sufficiently affecting your barometric sensor to drive your drone into the water, is also certain to create a significant sudden wind shear.

In any case, it's not the downward facing sensors that's causing that issue. Whether the surface is perfectly smooth, or rippled and wavy, the infrared height sensors (when in range) bounce a signal off the surface. If the drone senses it is getting too low, too close to the surface, it will automatically increase height. It won't cause the drone to sink into the lake.

That infrared sensor is not effective at high rates of forward speed because the forward pitched angle of the drone's body essentially means the sensor is looking at ground that's already passed behind it. So if you're flying forward at high speed low over water, you risk getting slapped by a wave.

I've seen people theorize that the infrared sensor might be seeing through the surface of clear water and registering its height as measured from the bottom of the lake, rather than from the surface of the water. So if the barometric sensor is somehow causing the drone to sink, the theory is that infrared sensor thinks it's still plenty high enough off the bottom of the lake and doesn't see the surface of the water. Even in that case, it's not the infrared sensor that's pulling the drone down.

I suspect that's nonsense anyway. Infrared light won't penetrate a pane of window glass, so I'm pretty sure it doesn't go through water either.

Much of what you said there is accurate, but some is most definitely not.

1. The downward vision sensors do NOT only "do something useful" while hovering. Even over land if I try to cruise too low to the ground the sensors tell me so. As you say, the sensors won't drive the drone into the water if you fly too close to it, but neither will they reliably keep you from doing so.

2. The ripples or waves in the water won't necessarily cause the drone to try to follow the change, but depending upon the algorithm used by the drone's internal software the rapid variation in feedback could cause the software to essentially ignore the feedback as being unreliable, at least temporarily while it tries to sort things out. This is often the case in feedback systems.

3. I agree with you in that I don't think that the infrared sensors are seeing the bottom of the lake rather than the surface of the water, since the surface of the water is almost never that smooth. But it is untrue that infrared light cannot penetrate water to some degree, and it is completely untrue that infrared light cannot penetrate a pane of glass. Every infrared camera has a glass plate covering the front, and if it also has infrared LEDs to light up the area the infrared energy has to go out through the glass and then come back in. Not sure how you got a different idea about that. Even far infrared (long wavelength ... essentially heat) energy will pass through a pane of glass to some degree, but the near infrared light used for LEDs and the sensors on our drones passes through easily. Even the sensors on the drone have a glass covering ... more than one, actually.

For our drones, this all means that the infrared sensors are getting responses over water (especially if you are close to it since further away the water looks more like a planar surface) that can't be counted on as being accurate. If you are flying close enough to the water that you depend upon the downward infrared vision sensors to avoid a splash down caused by other unforeseen variables (wind shifts, changes in barametric pressure, operator erro, etc) you are asking for trouble. The sensors don't cause the splash down, but they won't reliably stop it either.

 

[Zbip57]

Much of what you said there is accurate, but some is most definitely not.
1. The downward vision sensors do NOT only "do something useful" while hovering. Even over land if I try to cruise too low to the ground the sensors tell me so.

The downward facing sensors provide two completely separate functions. On my Phantom 3 Pro and on my Mavic Mini one component, the vision component, is a single camera lens which is only used to hold position while hovering. It has nothing to do with measuring height. The other component is the two infrared sensors on the Mini, which have nothing to do with vision, but function to measure height above ground. On the Phantom ultrasonic sensors perform the same function. The height sensors do function while the drone is in motion, but may not react quickly enough at high speed.

2. The ripples or waves in the water won't necessarily cause the drone to try to follow the change,

Not necessarily, correct yes. It depends on the circumstances and it depends on the algorithm being used.

With a good GPS lock telling the drone that it is in fact in a stationary hover, it'll probably ignore any conflicting input from the VPS camera seeing waves moving below it. But if you're flying indoors, for example, without GPS lock and hovering over a patterned rug, the drone will most definitely follow the rug if you try to drag the rug out from under it while it's hovering.

Similarly, for example, if you are flying under a wide enough and solid enough bridge or in a tunnel sufficient to block the GPS signal, then the drone will instead rely on the input from its downward facing VPS vision camera to hold position in a hover. If it sees ripples moving on the water surface below it, it will try to follow those ripples. Of course you still have control in ATTI mode to manoeuvre the drone, but if you were expecting it to stop and hold position when the control sticks are centred for hover, you might be surprised if it instead wanders off following the wave ripples.

3. [...] it is untrue that infrared light cannot penetrate water to some degree, and it is completely untrue that infrared light cannot penetrate a pane of glass. Every infrared camera has a glass plate covering the front, [...]

Hmm, I'll admit I could be wrong about that. I had an old Sony Handycam video camera that had a night-mode IFR switch. It had an IFR lamp to illuminate the scene, and recorded a ghostly black&white image. I seem to remember though that it couldn't "see" through windows. If I'm standing inside the house trying to video something outside the window, all the IFR camera saw was a black plate of glass. The IFR light just bounced off the glass without illuminating anything outside. Similarly, it couldn't see anything inside the house when filming from outside in night-mode.

If you are flying close enough to the water that you depend upon the downward infrared vision sensors to avoid a splash down caused by other unforeseen variables (wind shifts, changes in barometric pressure, operator error, etc) you are asking for trouble. The sensors don't cause the splash down, but they won't reliably stop it either.

That I do agree with. Especially when flying at high speed, flying low to the ground or low over water is more risky than flying at safer altitudes. The VPS sensors cannot be relied upon to save you under every circumstance.

But I think it's ridiculous when people suggest you should disable VPS whenever flying over water as if the VPS sensors are somehow responsible for causing the drone to crash.

 

[AZDave]

But I think it's ridiculous when people suggest you should disable VPS whenever flying over water as if the VPS sensors are somehow responsible for causing the drone to crash.

I agree with you 100% on that. Somehow I missed that erroneous suggestion from others ... probably because it's so ridiculous that it didn't properly register with me. As you say, the sensors don't cause the splashes.

 

[Gringorio]

A lot of the problem is sensors get tricked with reflections over water. It is best just not to fly too low. I stay at least 50' over water and on an exceptional sunny day even higher

I have hundreds of flights over water, primarily open ocean and some coastal areas. I always fly with sensors off. No issues flying as low as a few feet over water, just have to pay attention. The sensors just get in the way of flying IMO.

 

[BossBob]

I’ve hovered my MP1 at around 6 feet above a beach with full gps lock and had rock solid position keeping even in a breeze. I’ve then moved the drone to above the breaking waves and when hovering the drone moved up and down the beach following the breakers. The height stayed constant.
Hovering over a still clear water surface is another matter. The height of the drone has to be constantly watched and corrected.

 

[Zbip57]

Hovering over a still clear water surface is another matter. The height of the drone has to be constantly watched and corrected.

Is that a personal observation, or just a theory?

Why would a still clear water surface affect the height of the drone? What mechanism is at work then?

 

[BossBob]

The optical sensors don’t see the water surface clearly until it has been disturbed by the down wash and then that isn’t a constant so the sensors or the feedback loop they are part of gets confused. Personal observation, I often fly over a sheltered lake with a mirror smooth surface. If I’m moving the drone so that the disturbed water isn’t below the drone the drones altitude seems to depend on whatever is below the surface, plants, rocks or the lake bottom.

 

[AZDave]

Is that a personal observation, or just a theory?

Why would a still clear water surface affect the height of the drone? What mechanism is at work then?

Not sure how many times this needs to be said, but the water surface doesn't provide a clearly recognizable surface for the downward facing infrared sensors. If the water is clear and still, the infrared energy will penetrate the surface to some degree ... enough that you would lose much of the collision avoidance capability of the drone. The sensors might not cause a flying error, but they don't protect from one either.

If the water surface is wavy and the drone is far enough above the water the sensors will mostly see the surface OK, but if the drone is close enough to the water the waves/ripples will likely send confusing signals to the drone. How well the drone handles that is not clear, but it would depend upon how the software in the drone was written. In many feedback systems like that, the software is designed to at least puase when it gets confusing results so that it doesn't overreact and actually cause a problem that may or may not exist in the first place. That would mean that the drone's collision avoidance in the case of other error would again be limited.

 

[Zbip57]

The optical sensors don’t see the water surface clearly...

As per the User Manual, the Mini has a Downward Vision System and a separate Infrared Sensing System, using two different types of sensors. The "optical sensor" Vision System is a single camera lens, which it uses to fix on a patterned feature and is only used for position holding when in a hover with control sticks centred. It has nothing to do with maintaining altitude.

The other twinned sensors are infrared. One side emits a signal, the other receives the return signal bounced off of whatever is below the drone. The time difference in the echo is used to calculate the VPS height above ground.

The primary height measurement is still done with a barometric sensor, zeroed relative to your takeoff location. The infrared height sensor measures height above ground, whenever the sensors are within operating range (~30m).

The VPS height signal bounced off a smooth surface of water would register as a certain height. If, as you say, the signal penetrates through the water and instead bounces off the rocky bottom of the lake, that can only register as an even greater height. Neither should cause your drone to descend into the water.

My Mini only reacts to the infrared height sensor if it senses the drone is too low (<0.5m?). If I fly too low over a tree stump, the Mini will suddenly jump up. If I hold my hand close under the hovering drone, the Mini will rise up to avoid my hand. If I then take away my hand the Mini does not settle back down to its original height. The infrared height sensor will never cause the Mini to descend all by itself, unless when auto-landing below 0.5m while I hold the throttle stick full down.

Here's an experiment to try. Take off and hover your drone 6ft above the ground. The barometric altimeter and the infrared height should be equally registering 6ft altitude. If you slide a 1ft high box under the drone, the barometric altimeter will be unaffected and continue to read the same 6ft height above takeoff location, but the infrared height sensor would now show a reduced height of only 5ft above the surface of the 1ft box. Does the infrared system make the drone automatically rise up another 1ft to maintain its original 6ft AGL level (now above Box), or does it just stay hovering at its current 5ft above the 1ft box? It stays.

Conversely, with the drone still hovering 5ft above that 1ft high box, now pull the box out from under the drone. The barometric altimeter will remain unaffected and still show as 6ft above the takeoff point, but the infrared sensor height will show a sudden increase from 5ft above box now to 6ft above ground. This simulates the infrared signal initially bouncing off the surface of the water, versus suddenly bouncing instead off the rocky bottom of a one foot deep lake. Does your drone remain hovering where it was, or does the infrared system somehow automatically make the Mini descend 1 foot lower? It stays.

The infrared sensing system will never cause the Mini to descend. It only ever makes it rise if it senses something too close below.

The only possible scenario where an accident might occur would be if, as you say, the infrared signal is penetrating through the surface of the water and reflecting instead off the rocky bottom, and the water is more than 1m deep. Then the infrared sensing system might be fooled into believing it still has more than 1m ground clearance and it won't automatically prevent you from flying your drone into the lake. But it's never the infrared system that's pulling your Mini into the lake.

 

[AZDave]

As per the User Manual, the Mini has a Downward Vision System and a separate Infrared Sensing System, using two different types of sensors. The "optical sensor" Vision System is a single camera lens, which it uses to fix on a patterned feature and is only used for position holding when in a hover with control sticks centred. It has nothing to do with maintaining altitude.

The other twinned sensors are infrared. One side emits a signal, the other receives the return signal bounced off of whatever is below the drone. The time difference in the echo is used to calculate the VPS height above ground.

The primary height measurement is still done with a barometric sensor, zeroed relative to your takeoff location. The infrared height sensor measures height above ground, whenever the sensors are within operating range (~30m).

The VPS height signal bounced off a smooth surface of water would register as a certain height. If, as you say, the signal penetrates through the water and instead bounces off the rocky bottom of the lake, that can only register as an even greater height. Neither should cause your drone to descend into the water.

My Mini only reacts to the infrared height sensor if it senses the drone is too low (<0.5m?). If I fly too low over a tree stump, the Mini will suddenly jump up. If I hold my hand close under the hovering drone, the Mini will rise up to avoid my hand. If I then take away my hand the Mini does not settle back down to its original height. The infrared height sensor will never cause the Mini to descend all by itself, unless when auto-landing below 0.5m while I hold the throttle stick full down.

Here's an experiment to try. Take off and hover your drone 6ft above the ground. The barometric altimeter and the infrared height should be equally registering 6ft altitude. If you slide a 1ft high box under the drone, the barometric altimeter will be unaffected and continue to read the same 6ft height above takeoff location, but the infrared height sensor would now show a reduced height of only 5ft above the surface of the 1ft box. Does the infrared system make the drone automatically rise up another 1ft to maintain its original 6ft AGL level (now above Box), or does it just stay hovering at its current 5ft above the 1ft box? It stays.

Conversely, with the drone still hovering 5ft above that 1ft high box, now pull the box out from under the drone. The barometric altimeter will remain unaffected and still show as 6ft above the takeoff point, but the infrared sensor height will show a sudden increase from 5ft above box now to 6ft above ground. This simulates the infrared signal initially bouncing off the surface of the water, versus suddenly bouncing instead off the rocky bottom of a one foot deep lake. Does your drone remain hovering where it was, or does the infrared system somehow automatically make the Mini descend 1 foot lower? It stays.

The infrared sensing system will never cause the Mini to descend. It only ever makes it rise if it senses something too close below.

The only possible scenario where an accident might occur would be if, as you say, the infrared signal is penetrating through the surface of the water and reflecting instead off the rocky bottom, and the water is more than 1m deep. Then the infrared sensing system might be fooled into believing it still has more than 1m ground clearance and it won't automatically prevent you from flying your drone into the lake. But it's never the infrared system that's pulling your Mini into the lake.

You need to go back and check what I actually said instead of reading what you want into it. Or possibly you're confusing something that somebody else said with what I said. In either case you're wrong about what I've posted.

I never said that the infrared bounces off the bottom of a body of water with a clear surface. I only said that the infrared penetrates the water to a degree and that the drone doesn't get an accurate reflection to tell it where the water surface is.

I never said that if the infrared sensor thinks the surface of the water is further away from the drone than it actually is that the drone would drop. It won't. The barometric pressure sensor determines that. The infrared sensor just won't accurately provide a proper collision avoidance in case of other error (pilot error, wind shift, change in barometric pressure relative to takeoff point, etc).

I've also said that if the drone is flying close enough to wavy, choppy, or rippling water that the drone is likely to get confusing indications of just where the water surface is, and that the feedback control software in the drone might well pause while it's trying to figure things out. In the absence of more information on DJI's software, I stand by that. That's standard control system practice to prevent the equipment from over reacting to mixed feedback. Even that isn't going to pitch the drone into the water, but it could mean that the drone ignores the sensors long enough to cripple the collision avoidance in case of other error.

Clear now?

 

[Zbip57]

You need to go back and check what I actually said instead of reading what you want into it.

I was replying to BossBob, who said;

The optical sensors don’t see the water surface clearly [...] the drones altitude seems to depend on whatever is below the surface, plants, rocks or the lake bottom.

That's the part that's not being explained. Why and how would the drone's altitude ever be affected in any way by something under the surface of the water? The VPS system just doesn't work like that.

You (AZDave) and I are in agreement.

The infrared sensor just won't accurately provide a proper collision avoidance in case of other error (pilot error, wind shift, change in barometric pressure relative to takeoff point, etc).
[...] that isn't going to pitch the drone into the water

Exactly.

The original poster in this thread was worried that Uncommanded Descents were somehow caused by the VPS system when flying over water. He claimed to have seen "so much consistency in these reports". Someone in this thread recommended a safe altitude whenever over water of "Minimum of 50 feet...and typically 100 feet." Others even recommend altogether disabling the VPS system whenever flying over water, as though that will somehow prevent the drone from mysteriously crashing.

The Mini User Manual says the infrared height sensors may not work properly if;
"Flying over surfaces that reflect or absorb infrared waves (e.g. mirrors)"

A deep pile shag carpet can absorb or scatter, rather than reflect, the infrared signal. That doesn't mean the drone will automatically descend and crash into the carpet. It just means this obstacle avoidance feature may not be able to prevent you from flying your drone into the carpet.

So if you're flying over water and the infrared signal is absorbed by the water, rather than reflected, then the obstacle avoidance system may be unable to prevent a crash when fly your drone too low into the water. But the VPS system will never actively cause your drone to descend into the water.

Meta4 explained it most concisely.

That problem is just difficulty with horizontal position holding because the water surface is unsuitable for the VPS sensors. But that doesn't affect vertical position holding which is managed by a barometric sensor.

 

[AZDave]

I was replying to BossBob, who said;

That's the part that's not being explained. Why and how would the drone's altitude ever be affected in any way by something under the surface of the water? The VPS system just doesn't work like that.

You (AZDave) and I are in agreement.

Exactly.

The original poster in this thread was worried that Uncommanded Descents were somehow caused by the VPS system when flying over water. He claimed to have seen "so much consistency in these reports". Someone in this thread recommended a safe altitude whenever over water of "Minimum of 50 feet...and typically 100 feet." Others even recommend altogether disabling the VPS system whenever flying over water, as though that will somehow prevent the drone from mysteriously crashing.

The Mini User Manual says the infrared height sensors may not work properly if;
"Flying over surfaces that reflect or absorb infrared waves (e.g. mirrors)"

A deep pile shag carpet can absorb or scatter, rather than reflect, the infrared signal. That doesn't mean the drone will automatically descend and crash into the carpet. It just means this obstacle avoidance feature may not be able to prevent you from flying your drone into the carpet.

So if you're flying over water and the infrared signal is absorbed by the water, rather than reflected, then the obstacle avoidance system may be unable to prevent a crash when fly your drone too low into the water. But the VPS system will never actively cause your drone to descend into the water.

Meta4 explained it most concisely.

Wow ... how is that for iriony. I accuse you of mistaking me for somebody else and it turns out that I mistook your reply as being to me when it was to somebody else. And that's in spite of the fact that I thought I double checked who you were replying to. I'd chalk it up to senility on my part, but unfortunately I'm not there yet.

In any case, you have my sincere apologies.

 

[BossBob]

Sorry, I’m flying the Mavic Pro I acquired early 2017. It only has optical and ultrasonic downward sensors so may not react to water in the same way that your more modern (and restricted) drones do and the downward vision system only has a range of about 40ft. I fly with the sensors on in normal mode, some of the sensors turn off in sport mode. I don’t use ‘landing protection’ as I often hand catch my drone. My observations are of my experiences flying my drone. I’m pretty sure that after 4 years of flying the same drone I’m aware (or at least have a pretty good idea) of how it is going to react under most circumstances.

 

[Zbip57]

I’m pretty sure that after 4 years of flying the same drone I’m aware (or at least have a pretty good idea) of how it is going to react under most circumstances.

Well, that's the most important thing. As long as you know what it's going to do, and have practised using its various features, you're much less likely to panic in an emergency.

I would just like to understand the how and why explanations when something unexpected happens. My drone crashed while I was flying under a full moon. Ergo the moon can cause your drone to crash. That just doesn't cut it for me.

I know that the infrared height sensing portion of the VPS system on my Mavic Mini only ever makes it ascend if it senses it's too low. It will never cause the Mini to descend into water.

My Phantom 3 Pro, like your Mavic Pro, had ultrasonic sensors instead of infrared. It didn't have 'landing protection'. It wouldn't suddenly jump up to avoid your hand if you reached up to grab it out of the air, unlike the Mini which always annoyingly leaps away. There's no option to disable the Mini's sensors.

Anyway, the P3P did have a sort of terrain following capability. My driveway slopes gradually uphill. If I flew the Phantom low and slow up the driveway, it would automatically climb that slope maintaining the same constant height above the driveway. The Mini does that too, if it's within the 0.5m obstacle avoiding height.

But what I don't remember for sure is whether the Phantom would do the same in the downhill direction. I seem to remember it would also automatically decrease its height to match the downward slope of the driveway, always maintaining a constant height above the driveway. Can that be? The Mini definitely doesn't do that. If it's anywhere higher than its 0.5m ground sensing obstacle avoidance height, the Mini will just continue flying level as the driveway drops away below it.

I'd need to get the P3P out of its mothball storage to test that, but maybe you could verify with your Mavic Pro?

If your Mavic Pro automatically descends to follow a downward slope, and if that ultrasonic sensor can see through the water surface to a downward sloping rocky bottom, could that cause the drone to follow that downward slope until splashdown?? Obviously, that's preventable if you're paying attention, but... ?

 

[BossBob]

If I put the drone into terrain following mode it will automatically maintain its altitude above a rising terrain but not follow it back down unless I hold the left stick hard down. The ultrasonic sensor doesn’t seem to see water and I’m not about to risk flying above water with the drone in control of itself as regards height.
The biggest problem with the these drones is that the pilots forget that there is a piece of software between them and the drone. The software tries to interpret what you want the drone to do and then tells the drone to do it. It only takes a slightly off sensor or compass and the flight controller gets confused and sends the drone up, down or in the opposite direction to that intended by the pilot.
In four years I can probably count the number of calibrations I’ve done of any kind on one hand. I’ve got a sweet spot where the drone does what I want it to and I’m not going to spoil it with any unnecessary calibrations. The compass has been fine travelling the full length of the UK as I don’t have a drone with the later software that forces compass calibrations.

 

[AZDave]

The biggest problem with the these drones is that the pilots forget that there is a piece of software between them and the drone. The software tries to interpret what you want the drone to do and then tells the drone to do it. It only takes a slightly off sensor or compass and the flight controller gets confused and sends the drone up, down or in the opposite direction to that intended by the pilot.

This ... exactly. These drones are remarkably intelligent and built with all sorts of failure avoidance mechanisms, but they can't cover all possibilities and the sensors have limits that the operator should understand.

There was actually a guy not long ago here on these forums that claimed the drone should be able to anticipate when it was about to lose GPS so that it would prevent him from flying to the spot where that might happen. I hope he doesn't vote ....

 

[Gringorio]

Again, just turn off the sensors and fly. Argue all you want about what the sensors do or do not do in given conditions, it all comes down to flying the drone and capturing the shot you want. The sensors too often interfere with this and why it is best to turn them off for most occasions, including over water and flying from ships/boats.

 

[Meta4]

Again, just turn off the sensors and fly. .

That would be a simple remedy (if one really was needed), except that DJI have removed the ability to disable sensors in the app for their newer/smaller/cheaper drones.

I'm still not convinced that's the issue is genuine and I'd like to see some flight data to support the concept of sensors somehow causing over-water crashes.
None has caught my attention yet.