So my brand new DJI Mini 3 Pro doesn't seem to want to descend beneath the launch point. This just happened to me yesterday morning. It was about 10am in Long Beach, CA, and I launched my new DJI Mini 3 Pro from the roof-top pool deck on the 17th floor of my friend's apartment building.

Got it up, ascended about 6 feet, moved forward over the railing, then pulled the stick down to descend. At that point it was hovering 17-18 stories above an empty parking lot with no obstacles anywhere nearby, but once it got down to its takeoff height, it refused to descend any further and the prompt kept saying "Landing" and I was like, "wtf, no!" and had to quickly hit the button to cancel the auto landing.

This happened over, and over, and over, and over and over. A good 20 times. Shut the drone down. Swapped batteries. Flew to a completely different area. Still, just kept hearing "Landing" when trying to descend. It would occasionally get down to maybe -3 feet from takeoff height, but that's it.

I did eventually figure out I could trick it into descending lower by popping it up, throwing it into sport mode, going forward over the railing and *immediately* down into descent. Even then, though, switching back to normal or cine modes, if I came back up above takeoff height and then tried to descend again, I'd hear "Landing" and once again, say "wtf, no!!" and quickly cancel the automatic landing.

I'm not sure what's going on with this?! Ideas?

The only thing I can remotely think of is that it was a little bit hazy with the marine layer, but it was clear enough that there was still pretty much perfect visibility all the way to the ground. So unless the water vapor particles in the air were interfering with the sensors and tricking it into thinking it was hovering directly above the ground, I'm sort of at a loss here. However, even if water vapor in the air does end up being the explanation, that's kind of unacceptable.

Attached is a frame from the video from that morning showing the building I launched from, along with the empty parking lot I was trying to descend toward.

I live in a penthouse and I have gone down with my drone several times without any problem.

Try to take off, fly a few dozen meters up/forward and then come back and try to hover lower.
I remember some comment that said that if we try to descend right after launch, the drone will interpret that we want to land.

I'm not sure what's going on with this?! Ideas?
Possibly an issue with the haze affecting the VPS sensors, but it's hard to see how that would stop exactly at the takeoff level..

But there's no point guessing when the recorded flight data will probably show the cause.
To get to your flight data, go to: https://www.phantomhelp.com/LogViewer/Upload/
... where you'll find instructions to upload the data to that site and it will give a report
For help interpreting the data you can post a link for the report here.
Or just post the .txt file.

Could it be anything to do with geo-fencing.  The building is (almost certainly) higher than the allowed UAS restriction.

I had my M3 refuse to land last week from the exact place it was launched. Of course you can override it.

fans5d3e3440 Posted at 7-11 18:49
Could it be anything to do with geo-fencing.  The building is (almost certainly) higher than the allowed UAS restriction.

No ... altitude restrictions don't prevent descent.

fans5d3e3440 Posted at 7-11 18:49
Could it be anything to do with geo-fencing.  The building is (almost certainly) higher than the allowed UAS restriction.

17 to 18 storey  would be how high above the ground? At a guess 170ft to 180ft so well below the 400ft AGL ceiling

Sean-bumble-bee Posted at 7-11 21:36
17 to 18 storey  would be how high above the ground? At a guess 170ft to 180ft so well below the 400ft AGL ceiling

UAS in the area of that building is 100 AGL

Oh blimey........., ahh I see its Long Beach CA, just as a matter of interest why is the ceiling so low?

fans5d3e3440 Posted at 7-11 22:06
UAS in the area of that building is 100 AGL

UAS in the area of that building is 100 AGL
Perhaps you mean there is a 100 ft altitude restriction ??.

But altitude restrictions for the drone are above launch height, not above ground level.

The drone has no idea where the ground level is and there's no mention of it not climbing past 100 ft.
This wasn't a case of a drone not climbing past a certain height like would happen if there was an altitude restriction.
It was a drone that wouldn't descend after launching.
Altitude restrictions don't stop you from descending.

Sean-bumble-bee Posted at 7-11 22:14
Oh blimey........., ahh I see its Long Beach CA, just as a matter of interest why is the ceiling so low?

Maybe Long Beach airport.

Labroides Posted at 7-11 23:55
UAS in the area of that building is 100 AGL
Perhaps you mean there is a 100 ft altitude restriction ??.

So, if the UAS restrictions are only above launch height, I could take off from the roof of the Theme Building in the middle of LAX, drop down 10 feet and then I'm clear to fly anywhere around me </sarcasm>

Hi,

Typical bottom sensor false reading example.

your txt:
"However, even if water vapor in the air does end up being the explanation, that's kind of unacceptable"

I agree but that is how it works for a DJI drone!  if the bottom sensor is false measuring height (about 0.5 meter) than with 100% RC stick down it will enter an autolanding.
Fastest to cancel this landing is UP stick momentarily!  nice to know if you fly fast and low and drone enters autolanding bc stick down....
But flightlog will show the bottom sensor height values.

DJI 'refuses' to change this landing protection feature, needs to be ON always and via user OPTION in the menu to ability to switch to OFF.
(or automatically OFF when flying with some speed in ay direction....)

cheers
JJB

fans5d3e3440 Posted at 7-12 00:14
So, if the UAS restrictions are only above launch height, I could take off from the roof of the Theme Building in the middle of LAX, drop down 10 feet and then I'm clear to fly anywhere around me

So, if the UAS restrictions are only above launch height, ...
The restrictions are relative to ground level.
But the drone doesn't has no way to tell where ground level is or what height the drone is at relative to the ground.
I could take off from the roof of the Theme Building in the middle of LAX, drop down 10 feet and then I'm clear to fly anywhere around me
If you did you'd be in breach of the altitude restriction.

If it was mist inducing landing, wouldn't holding the throttle closed just have produced a very slow descent?

Sean-bumble-bee Posted at 7-12 02:09
If it was mist inducing landing, wouldn't holding the throttle closed just have produced a very slow descent?

yes, once in autolanding mode with all stick zero drone will slowly descent and descent and....

Ta ...

fans5d3e3440 Posted at 7-12 00:06
Maybe Long Beach airport.

Ahh......so there's an airport nearby. Thanks

Labroides Posted at 7-11 18:05
I'm not sure what's going on with this?! Ideas?
Possibly an issue with the haze affecting the VPS sensors, but it's hard to see how that would stop exactly at the takeoff level..

Alright. Thanks for the tip and the link to analyze the flight report! Getting it off my DJI RC controller was a little bit of an adventure, having to copy it to the SD card manually etc., since USB connection to transfer the files isn't available.

But yeah, turns out the problem was indeed the VPS sensor thinking that it was close to the ground when it was actually substantially above the ground. Attached screenshot.



Thank you!

Now the question is... next time I'm flying in hazy conditions, how to avoid this happening. Thoughts/ideas?

Brian

theturtle32 Posted at 7-12 18:46
Alright. Thanks for the tip and the link to analyze the flight report! Getting it off my DJI RC controller was a little bit of an adventure, having to copy it to the SD card manually etc., since USB connection to transfer the files isn't available.

But yeah, turns out the problem was indeed the VPS sensor thinking that it was close to the ground when it was actually substantially above the ground. Attached screenshot.

Now the question is... next time I'm flying in hazy conditions, how to avoid this happening. Thoughts/ideas?
Unfortunately DJI in their infinite wisdom can't see that users would ever want or need to disable the VPS, and they don't show any interest in users comments about this..
It's part of the dumbing down that they've been doing for all drones using their DJI Fly app.

theturtle32 Posted at 7-12 18:46
Alright. Thanks for the tip and the link to analyze the flight report! Getting it off my DJI RC controller was a little bit of an adventure, having to copy it to the SD card manually etc., since USB connection to transfer the files isn't available.

But yeah, turns out the problem was indeed the VPS sensor thinking that it was close to the ground when it was actually substantially above the ground. Attached screenshot.

Tape over the bottom sensors? But remember you will have no landing protection to slow the drone's descent as it gets near the ground. If you are accustomed to landing by holding the throttle fully closed and do so it would crash land 'at speed'. You would have to manually control the descent speed.

Sean-bumble-bee Posted at 7-12 22:29
Tape over the bottom sensors? But remember you will have no landing protection to slow the drone's descent as it gets near the ground. If you are accustomed to landing by holding the throttle fully closed and do so it would crash land 'at speed'. You would have to manually control the descent speed.

I suppose in a pinch, it's good to know that's a possibility, though it would also disable the rearward and downward obstacle avoidance, so I would rather avoid that except in situations where I have no choice.

Any particular kind of tape recommended?

theturtle32 Posted at 7-12 18:46
Alright. Thanks for the tip and the link to analyze the flight report! Getting it off my DJI RC controller was a little bit of an adventure, having to copy it to the SD card manually etc., since USB connection to transfer the files isn't available.

But yeah, turns out the problem was indeed the VPS sensor thinking that it was close to the ground when it was actually substantially above the ground. Attached screenshot.

Also, just a thought, since I'm a software engineer...

While the IMU(+GPS) altitude readings, while they don't assist the drone in knowing whether it's near the ground, the change in the IMU and GPS altitude *should* always at least agree with the change in the VPS reading, and if they don't agree, an algorithm should be able to determine which one to trust more.

For example, if you are descending, and the IMU Alt. shows descent while the VPS altitude shows remaining at a constant height or climbing, that would be a pretty strong indication that the VPS altitude shouldn't be trusted.

theturtle32 Posted at 7-13 10:15
I suppose in a pinch, it's good to know that's a possibility, though it would also disable the rearward and downward obstacle avoidance, so I would rather avoid that except in situations where I have no choice.

Any particular kind of tape recommended?

I think you need to re read the manual, the rearward obstacle avoidance sensors are on the top of the drone and the downward looking sensors are, as far as I know, NOT part of the obstacle avoidance system. i.e. they will not make the drone avoid an obstacle in the same sense that the true obstacle avoidance sensors do.
As far as I know, the downward looking sensors/sensing system do two/three things
1) when with in range they measure the distance to a reflective surface beneath the drone and apply various constraints that change with height.
2) when landing they assess whether or not the landing are is suitable. The system can decide it is not and then the pilot has to manually land the drone or command it to land
3) providing the surface is suitable they will use that surface as the basis for position holding (mainly associated with hovers) and in actual fact that control is, with an appropriate surface MUCH more accurate than GPS.
2) & 3) are probably part and parcel of the same system.

Suitable tape? Any thing thick enough and opaque enough to block IR and visible light/

theturtle32 Posted at 7-13 10:19
For example, if you are descending, and the IMU Alt. shows descent while the VPS altitude shows remaining at a constant height or climbing, that would be a pretty strong indication that the VPS altitude shouldn't be trusted.

and what would you have it do then, refuse to descend, descend as commanded,........?
I am not saying that what you are saying is wrong but a proposed action would be a good idea.

Sean-bumble-bee Posted at 7-13 11:46
I think you need to re read the manual, the rearward obstacle avoidance sensors are on the top of the drone and the downward looking sensors are, as far as I know, NOT part of the obstacle avoidance system. i.e. they will not make the drone avoid an obstacle in the same sense that the true obstacle avoidance sensors do.
As far as I know, the downward looking sensors/sensing system do two/three things
1) when with in range they measure the distance to a reflective surface beneath the drone and apply various constraints that change with height.

Of course, you are right. The rearward obstacle avoidance sensors are on the top. They're (visually) so small compared to the front-facing ones, I hadn't noticed them!

As for tape, I was curious about the experience of others in that arena, should it ever become necessary. There are lots of factors at play. Adhesive strength, opacity, inclination to peel off if damp, how much residue is left behind, etc.

Sean-bumble-bee Posted at 7-13 11:57
and what would you have it do then, refuse to descend, descend as commanded,........?
I am not saying that what you are saying is wrong but a proposed action would be a good idea.

I think one could probably make a case for a variety of behaviors under such a circumstance.

I'd suggest that in a situation when several VPS altitude readings over the course of 2-3 seconds contradict the IMU+GPS altitude, especially when taking into account what the drone is presently being commanded to do, it should simply consider the VPS readings to be spurious and behave in exactly the same way as it would behave if it were too far away from the ground for VPS to be useful. Basically it should just function exactly as it would when it shows "N/A" for the "VPS Altitude" field in the flight report.

Sean-bumble-bee Posted at 7-12 22:29
Tape over the bottom sensors? But remember you will have no landing protection to slow the drone's descent as it gets near the ground. If you are accustomed to landing by holding the throttle fully closed and do so it would crash land 'at speed'. You would have to manually control the descent speed.

Tape over the bottom sensors?
No help at all for flyers who find themselves in the situation with a drone that only descends at a crawl.
The OP was lucky that the drone was close when this happened.
Others only find out after they have flown some distance and/or have climbed to a significant height.

theturtle32 Posted at 7-13 15:06
I think one could probably make a case for a variety of behaviors under such a circumstance.

I'd suggest that in a situation when several VPS altitude readings over the course of 2-3 seconds contradict the IMU+GPS altitude, especially when taking into account what the drone is presently being commanded to do, it should simply consider the VPS readings to be spurious and behave in exactly the same way as it would behave if it were too far away from the ground for VPS to be useful. Basically it should just function exactly as it would when it shows "N/A" for the "VPS Altitude" field in the flight report.

I am not arguing with you but what merely demonstrating things that are concievable. Say the pilot is flying out of VLOS and using the screen (as happens) with the camera looking horizontally, they are composing a descending shot but inadvertently descend over a tree or bush or perhaps something like pampas grass ( it's not a great scenario I admit but feasible). VPS starts to detect the 'leaves' but the down draft blows them aside, resulting in lower growth being exposed and a sensor data conflict ensues. Currently the very slow 'landing' descent might offer the pilot the chance to save the situation, a descent at the commanded rate would possibly reduce their chances and I would bet there would be those that scream "it's DJI's fault".



The above said, I wonder if it would be possible to use information coming from the main camera or the OA sensors to assess whether the 'ground' is in fact possibly mist?

Labroides Posted at 7-13 18:12
Tape over the bottom sensors? No help at all for flyers who find themselves in the situation with a drone that only descends at a crawl.
The OP was lucky that the drone was close when this happened.

My post was in response to theturtles32's question
"Now the question is... next time I'm flying in hazy conditions, how to avoid this happening. Thoughts/ideas?"

I think it would be fair to say that he would, prior to launch, be aware that conditions are hazy so I think that my answer was useful to the probable circumstances and in my post that was and is my sole concern.

Obviously it would be 'rather' difficult for someone whose drone is airborne to apply tape to the drone but, where "descending at a crawl" is due to fog or mist or cloud if the sensors had been taped over prior to launch then those pilots too would not have encountered the descend at a crawl problem.

I do think the best solution would be for DJI to make it possible to switch landing protection off, as, from memory, it is with the Mavic 2  Pro but that's up to DJI.
As far as I remember that, the ability to switch landing protection off, has been requested since the Fly app was launched.

If you put it in SPORT mode, in fog will work a Forward-Descent....
I had this issue at mountain, when a very dense fog was at 40m high....Instead of direct descent (entering Landing mode moving very slow), I put it in Sport mode, joystick full forward and down....Was very fine, the descent was fast and all fog was traveled in less than 1 minute....with direct descent was probably out of battery before landing as was very slow descending (landing mode).

Geo_Drone Posted at 7-14 03:13
If you put it in SPORT mode, in fog will work a Forward-Descent....
I had this issue at mountain, when a very dense fog was at 40m high....Instead of direct descent (entering Landing mode moving very slow), I put it in Sport mode, joystick full forward and down....Was very fine, the descent was fast and all fog was traveled in less than 1 minute....with direct descent was probably out of battery before landing as was very slow descending (landing mode).

oke...

but with forward speed (normal or sport mode) AND downwards sensor is measuring 0.5 meter and for longer period >= 85 % stick down it will enter an autolanding.

So you were lucky to get it down flying with forward speed. Does not work always!

Hope that finally DJI add an option to the FlyApp to switch off the landing protection!!

cheers
JJB

JJB* Posted at 7-14 11:19
oke...

but with forward speed (normal or sport mode) AND downwards sensor is measuring 0.5 meter and for longer period >= 85 % stick down it will enter an autolanding.

Yes, but is an emergency solution... Nobody fly in fog unintentionally...Is not like the fog appeared from nowhere...
The button can be made as Fog button, for a number of seconds to disable the down sensors, but to reenable them, as many people are not so technical and will land with sensors closed and max descent speed = crash.