Obstacle avoidance reliability at speed?

[macfawlty]

This isn't something I'm inclined to run a series of tests on, but I'm trying to assess the reliability of obstacle avoidance when traveling at speed... not lightning speed, but say 15-20mph. I'd assume the M2P calculates speed/direction/distance to obstacle, but I'm hesitant to fly toward a brick wall at speed to find out.

 

[paulatkin73]

it works consistently enough for large non-reflective objects and if you did not modify P mode too much to boost the max horizontal speed/angle.

 

[Kilrah]

And it's something you should treat as something that "has a non-zero chance to save you from a mistake you did", but not as something to rely on. You still need to make sure you're not flying into something...

 

[brett8883]

This isn't something I'm inclined to run a series of tests on, but I'm trying to assess the reliability of obstacle avoidance when traveling at speed... not lightning speed, but say 15-20mph. I'd assume the M2P calculates speed/direction/distance to obstacle, but I'm hesitant to fly toward a brick wall at speed to find out.

I was sure they they were more trouble then they are worth for a good year or so. The only reason I even had them on was so I didn’t get the warning message that OA was turned off.

Then not too long ago I was flying full steam ahead and out of nowhere some birds crossed in front of my path. I actually thought they were off line with the drone but drone comes to screeching halt and doesn’t hit the birds. I checked the video later and sure enough the birds fly right in front of it. I’m not so sure I would have actually hit them but it would have been close.

Was totally shocked it saw a moving object like that and reacted. Only time I’ve ever had a true positive OA brake situation while flying manually(had a few on auto-RTH which saved me but this was the first manual one .)

 

[paulatkin73]

something that "has a non-zero chance to save you from a mistake you did", but not as something to rely on.

it is overall something that is not easy to swallow for folks who are not into this modeling hobby - one can and always should expect a failure - and be preemptively ready for it. there is no such thing as a 100% reliable model. there is always stuff that can brake, burn out, controllers can fail and crash, etc.
with properly made models like dji ones such probability is not high - but it is not at 0%, quite far from it.

people really do not seem to grasp it well until they go over their own first flyaway or crash or any other kind of a malfunction, user invoked or not.

 

[old man mavic]

I was sure they they were more trouble then they are worth for a good year or so. The only reason I even had them on was so I didn’t get the warning message that OA was turned off.

Then not too long ago I was flying full steam ahead and out of nowhere some birds crossed in front of my path. I actually thought they were off line with the drone but drone comes to screeching halt and doesn’t hit the birds. I checked the video later and sure enough the birds fly right in front of it. I’m not so sure I would have actually hit them but it would have been close.

Was totally shocked it saw a moving object like that and reacted. Only time I’ve ever had a true positive OA brake situation while flying manually(had a few on auto-RTH which saved me but this was the first manual one .)

had a similar experience with my MPP 250ft elevation 1300 ft out suddenly message breaking centre sticks, returned home and it was not until i watched the video that i saw several birds had flown very close to the front of the mav

 

[macfawlty]

I was in Kauai and Maui the last 2 weeks flying/following the cliff lines. In and out jaggy 100' cliff lines and beaches. I shoot video almost exclusively. Sometimes I start with gimbal down and tilt-up to reveal cliffs. Depending on the shot, the closer to the cliffs I fly, the better it looks. That foreground/background parallax... traveling forward w/ rotation/elevation change and gimbal tilt is a magical shot. Obviously, I can't see forward with the gimbal down, but I did tilt 2 or 3 times in a pass. It did stop a couple times, but I was traveling pretty slow.

The wind in Hawaii is fierce, but the M2P performed amazingly. Many would be reluctant to fly in those cliffside winds, but it's blowing all day every day, so... I saw it bouncing around a lot, but the footage is like butter.

I was sure they they were more trouble then they are worth for a good year or so. The only reason I even had them on was so I didn’t get the warning message that OA was turned off.

Then not too long ago I was flying full steam ahead and out of nowhere some birds crossed in front of my path. I actually thought they were off line with the drone but drone comes to screeching halt and doesn’t hit the birds. I checked the video later and sure enough the birds fly right in front of it. I’m not so sure I would have actually hit them but it would have been close.

Was totally shocked it saw a moving object like that and reacted. Only time I’ve ever had a true positive OA brake situation while flying manually(had a few on auto-RTH which saved me but this was the first manual one .)

That's pretty amazing. Good to know the birds are protected. I've been concerned with many locations, particularly near oceans, marshland, etc. Birds are everywhere and sometimes they'll do a fly-by to see what's up as happened in Kauai a few times last week.

These video stills are from Maui, Nakalele Blowhole, near Kapalua. There video is more dramatic when you fly close to objects. The M2P did stop on a few occasions. I'll upload some video clips later. Just finished exporting 89 edited clips.

 

[macfawlty]

it is overall something that is not easy to swallow for folks who are not into this modeling hobby - one can and always should expect a failure - and be preemptively ready for it. there is no such thing as a 100% reliable model. there is always stuff that can brake, burn out, controllers can fail and crash, etc.
with properly made models like dji ones such probability is not high - but it is not at 0%, quite far from it.

people really do not seem to grasp it well until they go over their own first flyaway or crash or any other kind of a malfunction, user invoked or not.

Yes, I always anticipate a crash. That's why I have insurance. My biggest concern is losing footage. I was certain to lose my P3 4k at some point as the range wasn't adequate (even with the Itelite DBS mod) and no obstacle avoidance. I did lose it in the mountains on forced critical battery landing, but it happened to descend next to a frequented trail and bridge. A Mennonite family saw it, retrieved it and took it to the ranger/discovery center where I was flying from.

 

[johnmeyer]

I only have one experience with the collision avoidance at speed.

I decided I wanted to fly really low over some bushes into a clearing, and then drop into the clearing. The bushes were only thirty feet away from me and I was standing perpendicular to the line of travel, so I could easily judge altitude and proximity to the bushes. I was going full-tilt in P-mode and about eight feet from the bushes, the drone came to a screeching halt. I'm 99% certain it wasn't about to hit anything, but it was, by my own design, certainly going to be close.

So, in my one experience, the sensors stopped the drone at full (P-mode) speed, and did so in time to prevent a collision, or even the potential for a collision.

I have no idea, when using the Waypoint and other automated modes, all of which use additional sensors, whether these "secondary" sensors work as well as the forward sensors.

 

[f13dfx]

Gentlemen:

Planning to do some creative video this weekend & did a low level test on shallow still water. Was wondering given a particular shot I’m planning & cognizant of obstacles around, should I fly in P or S mode when trying to do a video where I pass under a walking suspension bridge 20 feet above some rapids? I wouldn’t be doing any crazy speeds, just a slow pass. Thanks in advance!

 

[macfawlty]

Well, if you're flying slow, just use P.

I was flying over the ocean swell in Kauai realizing the ocean swells a lot more than I thought. The Mavic lifted to avoid touching the water.

 

[pls1911]

I've gently tested the object avoidance feature on several mediums, and found it works quite well 99% of the time.
Issues arise when the objects are less than solid/dense, like ponderosa pine branches loaded with pine needles swaying in a breeze. a few hundred branches of an open canopy with several hundred thousand needles all moving in different directions and different speeds drives the Mavic Pro sensors nuts trying to avoid multiple moving objects simultaneously.
Forget control, and RTH at normal altitude has it trying to navigate up though the moving canopy. That one experience resulted in a "get it on the ground NOW defensive command which semi soft side hill landing on rocks which cost me a lens filter replacement. Flying the same scenario without the auto object avoidance is a no brainer with proper planning and careful control.

So, I'd say most of the time auto object avoidance feature can protect you from yourself, and we all make mistakes.
Just don't make the mistake and misjudge the system limitations.

 

[CanadaDrone]

IMO the best way to think about obstacle avoidance is to pretend you don't have it. Never trust it, and fly as if you didn't have it. It's there as a fail safe, and might save your a$$ one day (most of us probably have stories), but never assume or expect it to do something. In my experience it works very well, but the one time you trust it implicitly or test it, it might cost you $2000.

It's kind of like the airbag in your car - you know it's there and you hope it works when it needs to, but the consequences of intentionally testing it and finding out it doesn't work as expected are not worth it. Of course that is an extreme example, and with a drone you can safely and gently fly the drone toward yourself or a wall and test it in a controlled manner, but those aren't the scenarios where we need obstacle avoidance - it's for that bird that comes out of nowhere or that tree branch or power-line we didn't see.

 

[johnmeyer]

... the consequences of intentionally testing it and finding out it doesn't work as expected are not worth it.

Without totally disagreeing with that, I think that some testing is quite useful, although not at full speed. For instance, I live in mountainous terrain and was very interested to know if, while trying to fly the drone up a hill, would the collision avoidance work to keep me from flying horizontally into the gently rising slope?

I did the test at slow speed, and the sensors stopped the drone. Of course I did not test this at 30 mph.

Along those lines, however, I have seen some pretty amazing footage of people who have indeed pushed the collision avoidance to its limits (they must be very rich, or else someone else is paying for their drones). The results are pretty impressive.

Just search YouTube and you'll find all sorts of videos.

[edit]Here's a link to a YouTube search that provides you dozens of videos to watch:

Mavic 2 Collision Avoidance

 

[Kilrah]

I decided I wanted to fly really low over some bushes into a clearing, and then drop into the clearing. The bushes were only thirty feet away from me and I was standing perpendicular to the line of travel, so I could easily judge altitude and proximity to the bushes. I was going full-tilt in P-mode and about eight feet from the bushes, the drone came to a screeching halt. I'm 99% certain it wasn't about to hit anything, but it was, by my own design, certainly going to be close.

That's one of the reasons I always have them disabled, I often do close flybys

 

[old man mavic]

while no one would not agree that collision avoidance is useful ,in certain situations such as flying slowly in restricted places. it can be fooled by bright sunlight water and foliage and also physics ie inertia, a drone needs time to stop, and will continue to move depending on speed and wind direction often with a fatal outcome for the drone as others have said do not rely on it to save the drone from every situation you are the PIC and you should always remain so