EVO 2 Crash while Dynamic Tracking

[AstroNoot]

I was impressed with how a previous Evo 2 Pro test unit I had performed, but taking my new Evo 2 for a dynamic track mission, my heart skipped a beat.

This happened:

The drone crashed into a tree . Luckily, the impact zone was an arm, and it folded. The only damages I can see are two propellers beeing chipped, and some really light scratches on the drone.

Do you guys think that Autel will help me in any way?

 

[rvl]

Can you copy the log of the flight from your phone and attach it here? (Watch this video if you don’t know how to find the logs)
Anyway this case should be of interest to Autel engineers, since front sensors should have seen that tree.

 

[kenautelevo2pro]

sorry to see your drone crash like that but it sounds like you maybe do not have enough damage for autel to take a look at it? I guess it doesn't hurt. unless you have more than a few scratches and props, I would say keep flying. please note the autel is not a skydio and regardless of what you see on yt, the autel cannot fly thru this type of environment without enormous risk. even the skydio might have a little bit of trouble but at least it stands a good chance because it's very agile and it known when the pause. the autel is a very capable drone but if you want something that will pry it way thru the heavy brush at a high rate of speed, you need to look elsewhere. I think you got lucky, the autel gimbal is delicate....and expensive.

 

[AstroNoot]

sorry to see your drone crash like that but it sounds like you maybe do not have enough damage for autel to take a look at it? I guess it doesn't hurt. unless you have more than a few scratches and props, I would say keep flying. please note the autel is not a skydio and regardless of what you see on yt, the autel cannot fly thru this type of environment without enormous risk. even the skydio might have a little bit of trouble but at least it stands a good chance because it's very agile and it known when the pause. the autel is a very capable drone but if you want something that will pry it way thru the heavy brush at a high rate of speed, you need to look elsewhere. I think you got lucky, the autel gimbal is delicate....and expensive.

I previously did 3 other flights with the same drone, and it was capable of stopping when the space was not big enough for it to pass. And it avoided stuff like a breeze.

I am pretty sure the Evo 2 is a capable drone, but what happened to me was definitely a software error, and not really a limitation.

I seriously don't want a fast chasing drone. In my opinion, the current dynamic track is... too dynamic. No one forced the engineers to make the drone avoid obstacles in the last moment, at high speed, instead of making the drone slow down gradually. The Mavic 2 Pro is clearly way worse at tracking subjects, but you can notice how it's a bit more "chill" when active tracking.

I would honestly shut up about what happened if it would be some thin branch that it crashed into. But it crashed into the thickest tree out of all.

 

[AstroNoot]

Can you copy the log of the flight from your phone and attach it here? (Watch this video if you don’t know how to find the logs)
Anyway this case should be of interest to Autel engineers, since front sensors should have seen that tree.

Here you go!

I added the .txt extension so the forum would allow me to send the file.

 

[herein2021]

I would never trust any current drone to reliably fly through that, in fact I wouldn't even fly the EVO II through that manually. OA is meant to complement manual control; not replace it. To me OA is like seatbelts in cars...if you get in an accident a seatbelt might save your life, that doesn't mean you should go crashing headfirst into things because you are wearing a seatbelt. OA is the same thing....if while flying manually it happens to pick up something that you failed to catch then it could save your drone; expecting it to avoid every obstacle on a trail like that is unrealistic. I see Tesla drivers in the news all the time because they trusted Autopilot literally with their life.

There are some pretty incredible YouTube videos showcasing the EVO II's OA but most of them don't mention they are on their 2nd or 3rd+ drone or if it is their first drone that it didn't last a month. I've been flying drones since 2014 and have to date never crashed a single one that was under my control. Obviously it is your drone and your money but personally I think using it this way is guaranteed to destroy the drone within a few months if not weeks.

Dynamic Track in a safe cleared area combined with OA I have no doubt would be perfectly fine.

 

[phantomrain.org]

I previously did 3 other flights with the same drone, and it was capable of stopping when the space was not big enough for it to pass. And it avoided stuff like a breeze.

I am pretty sure the Evo 2 is a capable drone, but what happened to me was definitely a software error, and not really a limitation.

I seriously don't want a fast chasing drone. In my opinion, the current dynamic track is... too dynamic. No one forced the engineers to make the drone avoid obstacles in the last moment, at high speed, instead of making the drone slow down gradually. The Mavic 2 Pro is clearly way worse at tracking subjects, but you can notice how it's a bit more "chill" when active tracking.

I would honestly shut up about what happened if it would be some thin branch that it crashed into. But it crashed into the thickest tree out of all.

We have found that acceleration with a target so close is simply to fast for any sensors on the Market today to negotiate .
Sensors have there place , and when used with experience meaning understanding the limits of the sensors , knowing that they need time to Recalibrate from one target to the next , thus making it though that maze was not likely.

Acceleration is difficult to quantify but you see it with Motorcycles taking a turn and sliding out where others just get thru the bend just fine.

We own all the drones and the sensors all have that Limitation .
You might consider asking Autel for a Discount if they deny the claim.

Phantomrain.org
Gear to fly your EVO 2 in the Pouring Rain.

 

[AstroNoot]

I would never trust any current drone to reliably fly through that, in fact I wouldn't even fly the EVO II through that manually. OA is meant to complement manual control; not replace it. To me OA is like seatbelts in cars...if you get in an accident a seatbelt might save your life, that doesn't mean you should go crashing headfirst into things because you are wearing a seatbelt. OA is the same thing....if while flying manually it happens to pick up something that you failed to catch then it could save your drone; expecting it to avoid every obstacle on a trail like that is unrealistic. I see Tesla drivers in the news all the time because they trusted Autopilot literally with their life.

There are some pretty incredible YouTube videos showcasing the EVO II's OA but most of them don't mention they are on their 2nd or 3rd+ drone or if it is their first drone that it didn't last a month. I've been flying drones since 2014 and have to date never crashed a single one that was under my control. Obviously it is your drone and your money but personally I think using it this way is guaranteed to destroy the drone within a few months if not weeks.

Dynamic Track in a safe cleared area combined with OA I have no doubt would be perfectly fine.

This kind of scenario is really rare for me, and in no way did I expect for the drone to actually start the dynamic track. Im surprised it did.
To be honest, I was impressed by the way it tracked another person on a bike, on a road with dense folliage on one side. Knowing it would do pathfinding easily on that road (it really didnt choose to fly on the clear side. It started going around and under the trees on the side of the road.), I thought it could really handle that dense forest scenario, considering that the road ahead wasn't really a maze. It was clear of obstacles/branches, like a tunnel through trees. Not gonna lie, the trailer for dynamic track that Autel posted on youtube, looked a lot more complicated than what I made my drone do.

Now I know its limits (and I'm quite sure I would never use the dynamic track again), but it would be really nice for Autel to make the dynamic tracking feature less tolerant, to avoid other unhappy users


And I hope they can help me out with one of those nice sets of quiet props

 

[phantomrain.org]

This kind of scenario is really rare for me, and in no way did I expect for the drone to actually start the dynamic track. Im surprised it did.
To be honest, I was impressed by the way it tracked another person on a bike, on a road with dense folliage on one side. Knowing it would do pathfinding easily on that road (it really didnt choose to fly on the clear side. It started going around and under the trees on the side of the road.), I thought it could really handle that dense forest scenario, considering that the road ahead wasn't really a maze. It was clear of obstacles/branches, like a tunnel through trees. Not gonna lie, the trailer for dynamic track that Autel posted on youtube, looked a lot more complicated than what I made my drone do.

Now I know its limits (and I'm quite sure I would never use the dynamic track again), but it would be really nice for Autel to make the dynamic tracking feature less tolerant, to avoid other unhappy users


And I hope they can help me out with one of those nice sets of quiet props

I hope you do get the silent props , lol

One thing I have noticed on both DJI an Autel is that there speeding up there videos in post. Its easy to do by a percentage rate to give the illusion there looking for. I have been fooled more than once on that Tracking , this is why i know oh so well.......

Phantomrain.org
Gear to fly your EVO 2 in the Rain.

 

[rvl]

Here you go!

I added the .txt extension so the forum would allow me to send the file.

The primary cause of the collision is poor GPS accuracy in dense forest. 2 seconds before the collision, the GPS signal was completely lost (for a second), resulting in a loss of speed control. Obstacle sensors worked fine, but the drone was moving too fast (~ 8m/s)

 

[AstroNoot]

The primary cause of the collision is poor GPS accuracy in dense forest. 2 seconds before the collision, the GPS signal was completely lost (for a second), resulting in a loss of speed control. Obstacle sensors worked fine, but the drone was moving too fast (~ 8m/s)

Replayed the flight record in the app and there are no GPS drops. The GPS sat count is oscillating between 18 and 20 sats

 

[rvl]

Replayed the flight record in the app and there are no GPS drops. The GPS sat count is oscillating between 18 and 20 sats

The flight log player in the app is created for a beauty and not for accuracy. Download the tool from that video description and check flight records yourself.

22th second, sat.count = 14

25th second, sat.count = 0, gps lost

 

[AstroNoot]

The flight log player in the app is created for a beauty and not for accuracy. Download the tool from that video description and check flight records yourself.

22th second, sat.count = 14
View attachment 10699
25th second, sat.count = 0, gps lost
View attachment 10700

Got it!
It's still weird. The crash happened 3 seconds later, but the drone was avoiding obstacles normally.

You can notice how it actually tried to go between the trees, that being one of the reasons the arms folded, and the crash wasn't really "frontal".

Do you think there was some sort of "conflict" between the drone trying to correct its position based on GPS and the OA system pathfinding through the trees?

 

[phantomrain.org]

Got it!
It's still weird. The crash happened 3 seconds later, but the drone was avoiding obstacles normally.

You can notice how it actually tried to go between the trees, that being one of the reasons the arms folded, and the crash wasn't really "frontal".

Do you think there was some sort of "conflict" between the drone trying to correct its position based on GPS and the OA system pathfinding through the trees?

So here is how we test: We fly across the water very low into the waves , using DJI drones systems.
Each wave is negotiated by the sensor and the drone moves up to avoid the wave and increase speed.

If two waves come in to fast , the drone get confused and need help quickly. Thus the slower we fly the more chance it has to calculate the obstacles. Objects repeating to close to each other confuse the sensors , its very apparent. Nothing to do with GPS.

Phantomrain.org
Gear to fly in the Rain and Land on the water.

 

[rvl]

Do you think there was some sort of "conflict" between the drone trying to correct its position based on GPS and the OA system pathfinding through the trees?

We do not know exactly how the autopilot works and what is important for it in this or that case. I think you should send this video and the log to Autel support and ask them to analyze what happened. Maybe they can clarified some things.

 

[AstroNoot]

Talked today with the support and they forwarded my info to their repair team. If I gotta send my drone for a repair, do you know how much time it'll take for it to get back to me?

 

[rvl]

Talked today with the support and they forwarded my info to their repair team. If I gotta send my drone for a repair, do you know how much time it'll take for it to get back to me?

Visually, there seems to be nothing to repair. If I were Autel, I would send you a set of new propellers to thank you for the tests

Q/A are near the bottom of the page

How long will service take?​

Service times often depend on the nature of the issue and the availability of service parts at the service location. Be sure to ask how long the repair will take when you discuss service with Autel Robotics, or when you take your aircraft to an Autel Robotics Authorized Service Center.

 

[AstroNoot]

Here's their response.

 

[rvl]

A bit of a strange answer. In my value system, "we can ..." must be "we want ..." if Autel cares about improving the quality of software/firmware.
I would sync the flight logs with the Autel cloud and ask them to analyze this flight for errors. But it's up to you.
By the way Bizmodel of Italy sells the low noise propellers presenting the photo of the standard propellers The real low nose propellers looks so

 

[AstroNoot]

I sent them the details of the crash (date and everything) and sent them the log too. So they can use any file they want.

Their audacity to tell me to buy my own props motivates me more to just send them the drone and make them repair it. In the end, i can't be 100% sure it's not a hardware fault. Imagine if there was a person instead of that tree