Crash! Who's Fault

[halley]

Guy, it's simple. The DJI Mavic Pro firmware has absolutely NO IDEA about terrain elevation. It has even less information about other hazards that are not entered into any database.

This is one reason that DJI has not (until recently) even offered a home-edit waypoint method. They want you to fly the mission yourself, see the hazards yourself, and record the waypoints yourself WITH your drone.

 

[CyberdyneSystems]

Ok, the engineer in me just can’t help himself. I’ve seen a few threads concerning the repeatability of the Mavic altitude determination. First off independent of knowing where you are on a topo map GPS is the only way to get an absolute elevation reading but is going to be +-15ft. From this you can set your barometric pressure as a starting point for relative altitude but even this is subject to measurement error. Barometric pressure changes about ¼ psi for every 500ft, or about 0.005 psi for 10ft. That requires a pretty precise pressure measurement. A quick google search shows a $1600 hand held barometer has on error rating of 0.015% (0.00015) over a span of 38 psi (38*0.00015=0.0057). Measuring a 10ft difference in elevation is right at the limit for this device. Frankly I’m amazed how well the Mavic holds level altitude in flight: I imagine it a combination of pressure, GPS and inertial measurements. But expecting it to miss a 75’ tree with a 95’ altitude setting over two mile course is just asking too much of the available instrumentation.

 

[Alex Baxter]

Ok, the engineer in me just can’t help himself. I’ve seen a few threads concerning the repeatability of the Mavic altitude determination. First off independent of knowing where you are on a topo map GPS is the only way to get an absolute elevation reading but is going to be +-15ft. From this you can set your barometric pressure as a starting point for relative altitude but even this is subject to measurement error. Barometric pressure changes about ¼ psi for every 500ft, or about 0.005 psi for 10ft. That requires a pretty precise pressure measurement. A quick google search shows a $1600 hand held barometer has on error rating of 0.015% (0.00015) over a span of 38 psi (38*0.00015=0.0057). Measuring a 10ft difference in elevation is right at the limit for this device. Frankly I’m amazed how well the Mavic holds level altitude in flight: I imagine it a combination of pressure, GPS and inertial measurements. But expecting it to miss a 75’ tree with a 95’ altitude setting over two mile course is just asking too much of the available instrumentation.

Kudos to the Engineer..
It would make sense DJI to use GPS for altitude as well as barometric pressure, I know that GPS triangulation is not as good vertically as it is horizontally, but I reckon it would do it to maybe 10 feet. I don't know enough to know if this is used or not. Would a gust of wind (which is after all, a change of pressure gradient) have a big effect of altitude if only the barometer was used..
I'm blissfully ignorant of these things BTW, my engineering is in the marine field, not much call for altitude there...

 

[andynormancx]

Kudos to the Engineer..
GPS triangulation is not as good vertically as it is horizontally, but I reckon it would do it to maybe 10 feet

You will not get 10 feet accuracy in altitude from real-time civilian GPS. 10 feet is pushing it a bit even for horizontal accuracy, it is typically a fair bit worse than that. The accuracy in the vertical can easily be twice that in horizontal.

 

[Alex Baxter]

You will not get 10 feet accuracy in altitude from real-time civilian GPS. 10 feet is pushing it a bit even for horizontal accuracy, it is typically a fair bit worse than that. The accuracy in the vertical can easily be twice that in horizontal.

Like I said, I'm blissfully ignorant, but Litchi missions are much more repeatable than 10', using on board GPS. Try setting a waypoint over a marker on the ground (out of visual positioning range),and see how repeatable it is, better than 10', you betcha....

 

[andynormancx]

Like I said, I'm blissfully ignorant, but Litchi missions are much more repeatable than 10', using on board GPS.

It isn't only the GPS that is being used.

 

[Alex Baxter]

Tell me more....

 

[malume]

I have had my Mavic for a few weeks and think its a great drone. I have been using Litchi almost exclusively and have been been working with the waypoints. After several small test runs with no issues I sent it on a 9000ft mission that would take it out of range and back. I have a lot of trees along the route that Max out at 75ft. I set my mission altitude to 95ft and thought that would be OK. There is a ten ft. elevation change along the route which honestly I didn't think of when planning the mission but my Mavic hit a 70ft tree and crashed terribly. I have noticed large discrepancies in elevation with the Mavic.
Is it my fault or a warranty issue?

Sent from my SM-G920V using MavicPilots mobile app

Sorry to hear that. I had a similar accident like you happened 2 weeks ago. I had it flied over 150 ft high that I thought it is Ok with Litchi waypoint. But unfortunately, it hits a tree that is over 150 ft tall ! Luckily I found it in the neighborhood with gimbal camera damaged. I took it to the DJI experience store in NYC but they told me it was my fault it give me a quote of repair for $350 including the parts and labor. Now I have no choice and it is under repairing but need 3 weeks to finish.

 

[tpop24]

I have refresh care and it's on its way to DJI. I anticipate using up one of my refreshes. The more I think of the incident the more things I find went wrong. First off, I had my waypoint mission set to follow a specific route and had set the direction directly towards the next WP, I wanted the MP to point towards the next WP then change direction direction once there. Instead it began panning about half way between points causing it to fly sideways negating the OA sensors. I do not want it to pan at all, I want to follow a route and point forward at all times and only change direction at the WP.(Anyone know how to do this?) Second, flight logs show 100' altitude when it collided with a 75' tree. There was an 8' elevation change from point of take off which makes the tree 83' I placed too much trust in the altitude reading of the MP. In the future I will drastically increase altitude on missions.

Sent from my SM-G920V using MavicPilots mobile app

 

[tpop24]

Original mission

Sent from my SM-G920V using MavicPilots mobile app

 

[tpop24]

It crashed between 11 & 12

Sent from my SM-G920V using MavicPilots mobile app

 

[andynormancx]

Tell me more....

The barometer and accelerometers can also be used to track whether it is gaining or losing height. They can't help with absolute altitude like the GPS can, but they can give the Mavic data about changes in altitude.

And it is changes in altitude rather than absolute altitude from an initial starting altitude that you are dealing with when pre programming a flight.

 

[Alex Baxter]

The barometer and accelerometers can also be used to track whether it is gaining or losing height. They can't help with absolute altitude like the GPS can, but they can give the Mavic data about changes in altitude.

And it is changes in altitude rather than absolute altitude from an initial starting altitude that you are dealing with when pre programming a flight.

Pardon me being pedantic, but when out of visual positioning range, the only horizontal can be GPS/Glonass, and it's MUCH better than 10', possibly 3 times better or more, see below. If you fly off on a mission that's a mile long, then pause at a waypoint, it'll be very repeatable, and providing it's out of visual positioning range the entire time, it can only use GPS/Glonass for horizontal positioning at that waypoint. My point being that it's very accurate. In any case I don't imagine it would use visual positioning when on a mission for positioning even at low altitude, only when hovering at low altitude.
From the manual..
GPS Mode GPS / GLONASS
Hover Accuracy Vertical:
+/- 0.1 m (when Vision Positioning is active) or +/-0.5 m
Horizontal:
+/- 0.3 m (when Vision Positioning is active) or +/-1.5 m

 

[maikm]

GPS/Glonass is very accurate for relative positions. With GPS you can tell very well in which direction you are moving and how fast, but the absolute position is not nearly as accurate.