Over water flight

[Mavic19963]

Any height where it's hard to tell if the Mavic is going to accidentally touch the surface of the water.

That means it would be safest if you can clearly see the Mavic with your eyes when flying very close to the surface (if you must for some reason) and should keep it high enough so there's no question as to if the Mavic is going to touch the surface of the water. The Mavic doesn't make quick downward movements, so it's not going to just drop into the water.
I read in the Dji manual to not fly over reflective surfaces, and that includes water. I have read way too many threads of MP pilots losing their drones in the drink. Downward sensors issue? For savvy pilots, use caution. For newbies, stay away from water.

Yes.

 

[Jwelz]

That's werid, because when it gets dark, my mavic will tell me that there isn't enough light for the vision position system to work.

That would likely be horizontal position, not altitude. The manual fails to clearly distinguish this from vertical position. Mavic can in fact hear in the dark provided the substate below is acousticaly reflective and not angled so as to reflect the return away from the ultrasonic sensor or otherwise disperse the echo.

 

[f3honda4me]

Yes. A stationary IMU relative to an observer will provide the same reading as one travelling at a constant
velocity. Accelerometers detect change in speed or direction, which is change in velocity, aka acceleration. Not movement, or change in position.

I was more interested in the claim of prioritization of ultrasonic sensors below 30 feet when the manual stipulates vertical detection over a range of 1 to 43 feet but if the movement detection claim is from the same source...

Well... if the drone was stationary, then suddenly goes down or up, that’s a change in position (through acceleration) Hence... IMU... So the IMU movement detection claim is accurate. You even just said so yourself.

 

[f3honda4me]

I found the post. It’s within range of ultrasonic, whatever the range may be, which was limited to roughly 30 ft originally on the p4p. (33ft is listed as the range today) The Mavic Pro is 43 feet.

Per the post:

“The ultrasonic when within range does control altitude. It's much more precise than the barometer. It does this in all modes not just in terrain follow.

The barometer would not be able to hold the craft in a hover as stable as the VPS ultrasonic can. Once you see the VPS icon deactivate that's when the height is handed off to the barometer.“

 

[GlenT]

Yes. A stationary IMU relative to an observer will provide the same reading as one travelling at a constant
velocity. Accelerometers detect change in speed or direction, which is change in velocity, aka acceleration. Not movement, or change in position.

I was more interested in the claim of prioritization of ultrasonic sensors below 30 feet when the manual stipulates vertical detection over a range of 1 to 43 feet but if the movement detection claim is from the same source...

I find this topic really interesting.
I know in the aviation industry dead rekoning is used as a back up for GPS based systems to give pilots a pretty good idea of where they are in a 3 dimensional space. It was my understanding this was achieved through use of gyroscopes and accelerometers. Measurements were recorded and compared along the way from the starting point or "fix", such as how much the aircraft accelerated in each direction along with barometric pressures which can then be computed into a usable outcome of where it thinks you are.

 

[Jwelz]

I have no dispute with what you say. However, we are discussing altitude over a surface which may be an elevated roof surface or at the bottom of a gully. We are concerned with a position relative to the Mavic rather than an absolute position relative to a staring point.

I don't think we can expect to compare the guidance systems of nuclear submarines with a drone flight controller IMU when an optical stereo range finder or ultrasonic transceiver pair will provide needed relative position information at a very tiny percentage of the cost. Particularly when absolute positioning information is of no value for practical obstacle avoidance.

Consider what happens when a Mavic loses GPS lock and the Mavic switches to Atti in a breeze. The IMU goes along for a ride.

 

[f3honda4me]

By the way, something else to note. In the drone DAT file there is an IMU altitude, derived from a combination of IMU and barometer.

 

[CamoGreg]

Did another flight over water, approx 1km range, all sensors on, no issue

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

No wind, no wave

 

[Jwelz]

By the way, something else to note. In the drone DAT file there is an IMU altitude, derived from a combination of IMU and barometer.

I will need to look at a DAT file to see what it says. Chinglish is often more amusing than interesting or illustrative.

No doubt it is possible to determine the time of day using carbon dating, but the accuracy may be somewhat in question. If one were to use the record of a trip of a person driving across town in various directions and speeds with stops and goes, one could determine the time from the position of the person by measuring the instantaneous speed and acceleration throughout the trip and performing calculations. If there were no other way of determining the time, this might be a valid scheme to use what you can measure to derive something that cannot directly be measured. However, it might be more accurate to do a single measurement with one range of error rather than a few million measurements, each with their own errors compounded through calculation. I did make reference to inertial sensors being a potential input in the altitude maintenance system of the Mavic in post 26 of this thread but took issue with your usage of the terminology. I also doubt, given the drift of the Mavic in Atti mode and the simple fact that these quads frequently land at different elevations than they take off that there is any meaningful contribution to the determination of altitude by the IMU. It can obviously detect a change in acceleration on the vertical axis, but not the absolute position or even vertical speed with any accuracy. In the absence of documentation, such claims are conjecture. Inertial navigation systems make sense as a backup when more reliable systems fail or are not available.

I'd like to know whether you believe that the Mavic IMUs initialized on a moving train/plane/platform can determine that it is moving? Hopefully if it had a GPS lock it should throw some warning. I would also like to know what happens when you take off from a roof top. A lot of people seem to believe that the Mavic will autoland when it determines a negative altitude relative to takeoff height but I have never experienced an autoland in a negative altitude situation. I'm also curious how a negative altitude works with the vertical position system activated in an environment where it can clearly detect the distance to the landing surface?

It is interesting to note that while the DAT file makes reference to a barometrically determined altitude there is no mention in the manual. There is no suggestion as to how accurate it might be nor how drastically it could possibly be out with changes in pressure and temperature over the course of a flight.

I find the entire discussion somewhat amusing as there seems to be clear evidence that there are circumstances where the altitude does not hold well above water surfaces in certain conditions and this is generally attributed to the 'sensors' being misled. It seems to me that it is necessary to activate terrain follow mode in order to prevent the Mavic from losing height over a rising surface, but it will presumably also maintain altitude in that mode when travelling over a falling surface. That implies to me that the downward altitude sensors or 'Vision' system, comprised of ultrasonic and optical range finding, are ignored or inactive during flight except when in terrain follow mode. If ignored, then they cannot be the cause of the fluctuations in altitude frequently reported when flying low over water. If the IMU plays any part in stabilizing altitude, why are there reports of abrupt excursions in altitude over water, transparent or possibly moving surfaces that may have been fooled by misinterpreted inputs? The vertical acceleration that goes with abrupt changes in height is precisely what the IMU detects yet it seems totally to disregard this data.

Notwithstanding numerous people, like CamoGreg, who say they fly without issue over water, the manual warns 'DO NOT guide the aircraft to fly towards.... transparent objects (e.g. glass or water)'; and 'TapFly mode may not work properly... over water...'. It says to use great caution when 'flying over water or transparent surfaces' and 'moving surfaces'. Maybe all those explanatory posts about those sensors possibly being confused are just drivel because the downward sensors are only activated in terrain follow mode and for positioning when returning to home or for positioning when the GPS is unavailable. But then why the warnings? Why the videos of near misses or dunkings in canals and streams? If the Mavic sees its reflection over a mirror surface, does it think it is twice as far away as the ultrasonic sensor indicates?

The manual is contradictory, unclear and confusing. For me this is a real issue as I have flown over a local picturesque creek and would like to do so again and be able to predict potential issues so I can proceed with minimal risk. I also spend a lot of time on small watercraft in the summer and wish to use the Mavic in these circumstances without having to maintain an altitude of more than 20 feet.

 

[Hamltnlbue]

Did another flight over water, approx 1km range, all sensors on, no issue

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

No wind, no wave

Maybe color of water has something to do with it.
Color would correspond to particulate concentration and could affect reflections.

 

[GlenT]

I can say from my own experience with flying my Mavic at low altitude over calm water (approx. 3 feet) caused some small yet visibly identifiable and erratic altitude fluctuations. As soon as I saw the Mavic do this I increased it's altitude to approx. 8 feet and it flew straight as an arrow again. This happened with all sensors activated.
As for the Mavic launching from a moving vehicle such as a boat, it will not know the boat has moved on from it's home point. The home point is merely a GPS co ordinate, which can be more accurately identified with the aid of the downward vision system. I've seen videos of people launching Mavics from boats only for them to get tangled up in rigging due to the boat moving from the Mavics launch point.
A way to alleviate this would be to use the RC Home Point function, or even Dynamic Home Point. The RC Home Point function needs the connected mobile device to have a GPS module, so WiFi only tablets etc cannot use this.

 

[BudWalker]

I will need to look at a DAT file to see what it says. Chinglish is often more amusing than interesting or illustrative.

No doubt it is possible to determine the time of day using carbon dating, but the accuracy may be somewhat in question. If one were to use the record of a trip of a person driving across town in various directions and speeds with stops and goes, one could determine the time from the position of the person by measuring the instantaneous speed and acceleration throughout the trip and performing calculations. If there were no other way of determining the time, this might be a valid scheme to use what you can measure to derive something that cannot directly be measured. However, it might be more accurate to do a single measurement with one range of error rather than a few million measurements, each with their own errors compounded through calculation. I did make reference to inertial sensors being a potential input in the altitude maintenance system of the Mavic in post 26 of this thread but took issue with your usage of the terminology. I also doubt, given the drift of the Mavic in Atti mode and the simple fact that these quads frequently land at different elevations than they take off that there is any meaningful contribution to the determination of altitude by the IMU. It can obviously detect a change in acceleration on the vertical axis, but not the absolute position or even vertical speed with any accuracy. In the absence of documentation, such claims are conjecture. Inertial navigation systems make sense as a backup when more reliable systems fail or are not available.

I'd like to know whether you believe that the Mavic IMUs initialized on a moving train/plane/platform can determine that it is moving? Hopefully if it had a GPS lock it should throw some warning. I would also like to know what happens when you take off from a roof top. A lot of people seem to believe that the Mavic will autoland when it determines a negative altitude relative to takeoff height but I have never experienced an autoland in a negative altitude situation. I'm also curious how a negative altitude works with the vertical position system activated in an environment where it can clearly detect the distance to the landing surface?

It is interesting to note that while the DAT file makes reference to a barometrically determined altitude there is no mention in the manual. There is no suggestion as to how accurate it might be nor how drastically it could possibly be out with changes in pressure and temperature over the course of a flight.

I find the entire discussion somewhat amusing as there seems to be clear evidence that there are circumstances where the altitude does not hold well above water surfaces in certain conditions and this is generally attributed to the 'sensors' being misled. It seems to me that it is necessary to activate terrain follow mode in order to prevent the Mavic from losing height over a rising surface, but it will presumably also maintain altitude in that mode when travelling over a falling surface. That implies to me that the downward altitude sensors or 'Vision' system, comprised of ultrasonic and optical range finding, are ignored or inactive during flight except when in terrain follow mode. If ignored, then they cannot be the cause of the fluctuations in altitude frequently reported when flying low over water. If the IMU plays any part in stabilizing altitude, why are there reports of abrupt excursions in altitude over water, transparent or possibly moving surfaces that may have been fooled by misinterpreted inputs? The vertical acceleration that goes with abrupt changes in height is precisely what the IMU detects yet it seems totally to disregard this data.

Notwithstanding numerous people, like CamoGreg, who say they fly without issue over water, the manual warns 'DO NOT guide the aircraft to fly towards.... transparent objects (e.g. glass or water)'; and 'TapFly mode may not work properly... over water...'. It says to use great caution when 'flying over water or transparent surfaces' and 'moving surfaces'. Maybe all those explanatory posts about those sensors possibly being confused are just drivel because the downward sensors are only activated in terrain follow mode and for positioning when returning to home or for positioning when the GPS is unavailable. But then why the warnings? Why the videos of near misses or dunkings in canals and streams? If the Mavic sees its reflection over a mirror surface, does it think it is twice as far away as the ultrasonic sensor indicates?

The manual is contradictory, unclear and confusing. For me this is a real issue as I have flown over a local picturesque creek and would like to do so again and be able to predict potential issues so I can proceed with minimal risk. I also spend a lot of time on small watercraft in the summer and wish to use the Mavic in these circumstances without having to maintain an altitude of more than 20 feet.

The Mavic uses the vision and ultrasonic systems both to know when it's close to the ground and how close. The relativeHeight isn't used for this purpose. There have been numerous incidents caused by a faulty vpsHeight indicating close to the ground when the actual height was much higher. Here's an example
Mavic Chrash

I understand your point about using the IMU to determine altitude. But, I think the claim is that altitude is computed by using both the barometer and the IMU. It's been known for a long time that Yaw is done by fusing the magnetometers and gyros - in fact, it's mostly the gyros and the magnetometers providing corrective input. Recently, it was realized that the GPS coords seen in the .DAT are actually computed by fusing IMU data with GPS data. A really good explanation of this was provided by @sar104 here
Lost all control and crashed
Using the same strategy as Yaw and the GPS:coords altitude could be a fusion of barometer and IMU data. But, why? vpsHeight is used close to the ground (or water). But, hovering would be where faster, more accurate altitude could be used.

Manual? What manual? We don't need no stinking manual. The stuff you see in the .DAT was all reverse engineered so there isn't any info about accuracy etc except what we've managed to figure out.

I've not heard anyone claim that a negative altitude will trigger an autoland. But, that's clearly not the case.

 

[Jwelz]

The Mavic uses the vision and ultrasonic systems both to know when it's close to the ground and how close. The relativeHeight isn't used for this purpose. There have been numerous incidents caused by a faulty vpsHeight indicating close to the ground when the actual height was much higher. Here's an example
Mavic Chrash

I understand your point about using the IMU to determine altitude. But, I think the claim is that altitude is computed by using both the barometer and the IMU. It's been known for a long time that Yaw is done by fusing the magnetometers and gyros - in fact, it's mostly the gyros and the magnetometers providing corrective input. Recently, it was realized that the GPS coords seen in the .DAT are actually computed by fusing IMU data with GPS data. A really good explanation of this was provided by @sar104 here
Lost all control and crashed
Using the same strategy as Yaw and the GPS:coords altitude could be a fusion of barometer and IMU data. But, why? vpsHeight is used close to the ground (or water). But, hovering would be where faster, more accurate altitude could be used.

Manual? What manual? We don't need no stinking manual. The stuff you see in the .DAT was all reverse engineered so there isn't any info about accuracy etc except what we've managed to figure out.

I've not heard anyone claim that a negative altitude will trigger an autoland. But, that's clearly not the case.

Great info. I will look into the DAT files to see what's there. I'm also going to do an optical vision system calibration and measure the angles of view displayed on the Assistant screen. I'm also working on an Arduino/stepper focus puller and will try out an 40 kHz ultrasonic range finder over flat, rippled and wavy water come spring time. Everything outside is frozen where I live right now but that should give a little more insight. Thanks for the info and link.

 

[Turn11]

GPS altitude determination

As I recall, the elevation on my "GPS Test" app is +/- 30' My personal experience has been a little closer. At sea level I have laid the phone on the sand in California, Massachusetts and Cabo Mexico and it has been under +5' from zero. this does not take in to account tidal variations.

Not too bad when you consider the trig involved.

 

[Turn11]

guidance systems of nuclear submarines

How is that done?

This entire thread has been extremely interesting and informative. Thanks to all for not evolving into a ssing contest.

 

[CamoGreg]

Hello guys! I post also here my last video as it’s also about flying low over the water !

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