Dunk into the ground during low-altitude flight

[Spheric]

OK, so I've gotten around to writing up my little horror story from last week so I can share my experience and get your thoughts on exactly what happened.

Last wednesday, I went out to the Iskar lake to get some nice golden hour footage. As I do in low temperature conditions, I left the car running with the batteries inside. I went through my pre-flight checklist, checking props, battery temp and volts, GPS/homepoint, RTH altitude and calibrating compass. All was good. I depleted my first battery and switched over to the second.

The lake is frozen over and creates some interesting scenery with the sunset & mountains in the background, but the shots were a little too static, so I wanted to bring in some movement. To do that, I wanted to try a low-altitude shot with the camera facing downwards and panning up as the aircraft accelerates and starts to gain altitude.

I maneuvered to the position & orientation I wanted, panned the camera down and started recording. I accelerated the AC straight ahead and slowly started to pan (the camera) up and throttle up. Just as I did that, I looked up from the screen and watched in horror as my dearest UAV friend smacked into snow & ice-covered lake, flipped and then rolled to a stop. The RC gave a 'gimbal roll movement limit reached' error, then a 'motor obstructed' error and then disconnected.

I was INCREDIBLY lucky on many, many fronts:
- I recovered the aircraft, battery and vibration board together with the X3 without breaking through the ice and dying
- The initial contact of the AC with the ground was the X3 hitting a mini snow mound. This caused the vibration board to rip off at the rubber absorbers and the AC to flip forward, smack its propellers into the snow/ice and fall on its back. This in turn lead to the the battery flying off, but it also resulted in the AC sliding to a stop in a relatively non-turbulent manner
- In a crash that happened at FIFTY FREAKIN KM/H on snow + ice, I had the astronomical luck of only having 2 broken propellers, a vibration board that had to be re-mounted and 2 of the 4 battery data connectors on the AC bent slightly and not springing back properly until straightened.

Here are the major details of significance that I thought to share (please let me know if you would like to know anything in addition):
- altitude (relative to home/take-off point) read -1.6m at the time of crash, -1.7m at the start of the run
- approximate height above the ice cover was 1.5m
- the slope of terrain (ice cover) underneath was near flat, even slightly downhill (I know - weird, the water must be receding and leaving already-formed ice cover behind?)
- only pitch forward and throttle up commands were ever given following the start of the run
- vision positioning system was switched ON (is that the cause of the crash?)
- based on my research, GPS is of no significance, but regardless - the AC was connected to 18 sats, and was in F-GPS mode
- all records of battery operation during, before and after the crash were completely normal
- The snow mound that the AC initially hit is not very tall at all (20-30cm at most), so it cannot be the sole cause of the crash

After quick abit of research, my theories as to what happened are the following (in order of perceived plausibility):
1. The VPS, operating on an AC moving at 50 km/h and over sound-absorbing, patternless snow/ice cover, gave an incorrect reading that resulted in the AC intentionally losing altitude (if you pay attention, you can see the AC losing altitude on the camera recording below).
2. Some sudden change in atmospheric conditions, like a cold gust or something of the sort, caused a non-altitude-driven barometer reading change that mislead the AC into dropping altitude to the point of hitting the ground
3. Something else that I do not know about caused the AC to lose altitude and it didn't have the time/power to compensate before it hit the ground.

After the crash:
- I thoroughly dried the AC and didn't switch it on before 24h had passed
- I physically checked in full detail every nut and bolt on the AC
- I tested each rotor, looking for vibrations, sounds or oscillations
- I performed a fresh IMU calibration
- I carefully went over all the flight data I had available (HealthyDrones HD360 Gold incl)
- I gradually tested the functioning of AC until finally doing a full flight with no issues

I've uploaded the the flight log and camera recording for your viewing.

Any thoughts on any of the above are most welcome.

 

[Viper Pilot]

Remember the old pilot saying............'Altitude above you and runway behind you does no **** good'! Bottom line, you were just too low!

 

[Joola]

I think your first 2 theories as to why it crashed are the most plausible. I've experience the rapid altitude drop when flying rapidly over grass, concrete, water, and beach sand. I know the manual warns about this too. Luckily I've been able to notice it and correct it before anything bad happened.

Glad you got so lucky and are back in the air.

 

[Alastair]

i almost had the same thing happen to me, flying low over grass. I learned from this forum that when you're flying that close to the ground and accelerate forward, the AC will tip forward and have a tendency to loose altitude. I believe that low pressure created above the AC plus the prop wash generated that close to the ground are contributing factors. Bottom line is, when flying that low, keep the left hand ready to push up on the stick.

 

[Dobmatt]

i almost had the same thing happen to me, flying low over grass. I learned from this forum that when you're flying that close to the ground and accelerate forward, the AC will tip forward and have a tendency to loose altitude. I believe that low pressure created above the AC plus the prop wash generated that close to the ground are contributing factors. Bottom line is, when flying that low, keep the left hand ready to push up on the stick.

Yep, that's typical behavior of Inspire 1 when flying low and fast, VPS on or off doesn't matter. Lucky you!

 

[Spheric]

Thanks to everyone for their input!

@Alastair lesson learned - visual altitude tracking will be super careful in low-altitude flights hereon!

@Dobmatt VPS was still the most likely suspect in my mind up until this point - I could see how the high speed combined with sound-absorbing terrain messed up sonar readings, basically slowing returning sonar pulses and making the I1 think the ground is further away. But now that you are telling me you've observed this behaviour with VPS off (is that the case?), it must be something else!

In any case, I am one lucky SOAB to get this lesson down without killing my bird!

 

[NickU]

A couple of thoughts for you...

The barometers aren't massively accurate and readings can easily vary with local atmospheric conditions. They improve with height, but it's why they use the sonar/vps system at 33ft or less to give a more accurate reading.

Having said that, the vps/sonar is going to be most accurate with the aircraft stationary and either ascending or descending , if you're in sideways flight then the sensors are going to be less accurate and will see a slightly increased height as the 'beam' they use will be angled at the ground and not be tangent/90 degrees. In forward flight, the beams will be aimed slightly backwards and won't see any rise in ground until the aircraft is already passing over it, couple that with a few milliseconds delay in the telemetry and you've a situation where it warns you about the terrain rise too late to do anything about it.

The vertical accuracy/hold of the aircraft is given by DJI, and usually varies between 0.3 and 1m depending on the specific model. Put you aircraft in a hover an watch how much it will rise and fall without any stick input - it won't stay exactly in one place but will rise and fall - that happens in flight too, it can suddenly drop, if you're close to the ground you won't have much time to correct it.

Put all those together and you need to be really careful when flying close to the ground. Maybe one day they'll put a proper terrain following 'radar' on the quads. I think the mavic can sort of follow the terrain, but it's still very limited.

But, glad things parted company and you didn't have too much damage... and you got yourself and the inspire back off the ice safely.

 

[Spheric]

Lots of great info & analysis @NickU, many thanks for your input! Even though I still can't put my finger on exactly what caused the crash, I now feel a lot more informed about the many factors that need to be considered when flying low. And just how careful you have to be.

One new thing that I noticed is that prior to the crash there was a reading of negative vertical speed. What's interesting is that there wasn't a corresponding drop in altitude readings. I am guessing that the two values are derived from different sensors (barometer vs accelerometer) and that might hold a clue as to what happened - if the barometer didn't register a drop in altitude, but the accelerometer did, it could mean that the barometer was giving an incorrect reading that actually caused AC to intentionally decrease lift and drop to the ground. As you mentioned, that's where the the VPS should have come in, but was unable to adequately due to speed/terrain. Still only a theory though.

One thing I am still wondering about is whether it's a good idea to keep VPS off during low altitude, higher speed flights, and I am currently leaning more towards keeping it off.

 

[NickU]

Depends on speed - the VPS protection is meant to be inoperative above certain speeds (don't have the figures to hand). If the VPS is on, at least you'll have a slight clue as to your height and it's slightly more accurate than the barometer at low level.

If you want to fly so low to the ground, then I think you really need to be within very close visual sight of the aircraft (10m or so) and most probably the aircraft needs to be flying side on to you so *you* can judge the forward ground proximity. Flying nose out from you, you're not going to be able to judge relative heights and speeds very easily. and relying on the internal cameras and sensors isn't really going to save you as they're not really designed for forward flight so close to terrain. The wide angle perspective of the camera lens also leads to misperception of how close you are to objects - use that to your advantage and actually fly a bit higher . If you want really low stuff, then buy an RC truck and stick a gimbal on that! I once stuck a gopro under the number plate of my car - the footage looked frighteningly low even though it was still 8-10 inches off the road! (as it happened, that was a semi-disaster as the suction mount came undone and dunked the gopro on the road. The safety strap saved it from being lost. The lesson I took away from that was to make sure the surface the sucker was attached to was perfectly clean!)

As to the -v vert speed, that could simply be caused by wind, local pressure drop or just the aircraft slightly losing altitude. The problem is that you've so little room and time to react that neither you nor the aircraft can recover in time. If you just watch it in a steady low level hover in no wind, you'll see how much it can move around, add in a slight wind and it can move even more. flying fast and low to the ground and one of those movements will simply do what happened to you - CFIT, controlled flight into terrain. It's not that there was an error or fault with the avionics as such - just that you were operating on the edge of or outside of it's capabilities. Just add in another external factor such as the ice ridge and then Murphy's Law will take over at some point.

Hopefully now Murphy caught you out once, he won't get you again as you'll be wiser and better prepared for him

 

[spero]

I almost lost my I1 a few times flying low over water. I learned that they don't hold steady altitude when low over water. Mine always slowly decends when a few feet over any body of water. Glad you recovered it though.

Sent from my VS990 using InspirePilots mobile app

 

[Spheric]

Depends on speed - the VPS protection is meant to be inoperative above certain speeds (don't have the figures to hand). If the VPS is on, at least you'll have a slight clue as to your height and it's slightly more accurate than the barometer at low level.

Do you mean that the VPS switches off when beyond its operating limits? If that is the case, then that would be much better than the alternative (it not working properly but still sending commands to the ESCs) and would mean there is no point in switching it off manually from the DJI GO menu like I was considering to do hereon.

You got a point about the wide angle, in most cases the effect is dramatic enough even when not that close to the terrain, so it is not worth risking going too low in most situations. Good thing you had a safety line on that GoPro! I am actually planning on getting the OSMO suction cup, so I'll be applying both mine (on low-altitude flights) and your learnings (clean before suction + safety line) in future projects

@spero man I was lucky the ice was strong enough to take the hit from the I1, all would be lost now heh! Stay safe!

 

[spero]

You know, now that I'm thinking about my own near water death experience with the inspire I had a laugh. The only reason mine did not end up crashing into the river was as I was watching the video the landing gear started to lower. I was concentrating so hard on getting the right shot that I didn't realize the bird was only a few inches off the water. As soon as I heard that and looked up to see what was about to happen I pulled up on the stick. We both got lucky lol

Sent from my VS990 using InspirePilots mobile app

 

[Spheric]

You know, now that I'm thinking about my own near water death experience with the inspire I had a laugh. The only reason mine did not end up crashing into the river was as I was watching the video the landing gear started to lower. I was concentrating so hard on getting the right shot that I didn't realize the bird was only a few inches off the water. As soon as I heard that and looked up to see what was about to happen I pulled up on the stick. We both got lucky lol

I feel you, it's unbelievably tough splitting your attention between getting a good shot and piloting, navigating in 3 dimensions and composing a good shot simultaneously. We are finally getting our second remote next week and look forward to doing some shots with a separate camera operator.

 

[NickU]

The osmo car mount is great compared to the GoPro one - 3 suction cups vs 1 just make sure you tighten down the osmo -to-mount screw

The manuals state that the Phantom 4 and Inspire 2 VPS operates at speeds <10m/s at 2m above ground according to the specs - that seems to be for the downward facing sensors. They don't state any explicit operating speeds for the forward facing sensors. Not sure about the I1, you'd need to check its specs. I'd presume that the response of the system can't get an accurate enough fix quickly enough above those speeds. It doesn't state that that is purely for vertical speed, so I'd work on the assumption that it's the same for horizontal speed too.

Roll on the day they install forward and downward laser rangefinders and a full terrain following mode

 

[William Gaddy]

Expert in optical flow (VPS) here. Any optical flow system is going to have a real hard time giving accurate estimates with a relative homogeneous ground is moving rapidly under it. My understanding is that their VPS has an effective 1080 pixel forward resolution. If you were flying low and fast enough that the ground was moving more that 960 pixels per frame, it's going to "**** the bed" entirely. If the barometer is giving conflicting signals and you're below 10 feet, I'm going to guess the controller is going to more rely on the VPS, and WHAM.

 

[lake_flyer]

VPS was only intended to assist during low hovering and landing at places where GPS reception (like indoors) is too low, and to help when hovering low in Atti. Nothing else. It's not meant to be an extra aid for low flying. I prefer to have it off even. I never liked the sudden drop when I flew low over the lake bank to the water. Same as the auto gear. I had it coming down a couple of times during a very low flight and that's not good. Never switched that back on again after switching off 2 years ago..

I switched VPS on for an indoor shooting lately, and it did help a little bit to counter the ground effect while landing and hovering low in Atti. But more than a few feet up and it's useless, thank god...It's a kind of false 'safety' to rely on. Specially when flying outside. It behaves particularly weird over water.

 

[spero]

That sounds about right. My near death was under 10 feet, in atti mode, and full stick forward. The bird slowly lost altitude over 30 seconds or so until I almost went in. Literally saved by a couple inches.

 

[Larr2000]

There is another possible cause for loss of altitude - turbulence. If there were any trees and slight wind in the area, you could have been the victim of turbulence. Remember there is an 8 X 1 ratio from the direction the wind is blowing AND in cold conditions the air mass is heavier adding the the accelerated downdraft. What were your surroundings? I have seen these downdrafts in low flights below 10 meters. Often resulting in a crash. Glad you recovered and are back in the air.

 

[zulu]

I feel you, it's unbelievably tough splitting your attention between getting a good shot and piloting, navigating in 3 dimensions and composing a good shot simultaneously. We are finally getting our second remote next week and look forward to doing some shots with a separate camera operator.

 

[licensed pilot]

Flying so low while filming is not easy. FAA calls it "controlled flight into terrain."