Sorry that happened KlooGee. Low(er) light + blue cables + blue bridge structure = perfect storm. It's really easy to get a false sense of security with the SD2 but at the end of the day it's like anything, subject to the limitations of it's technology.Looks like not enough light for the OA to operate well.
You seem more impassioned about holding Skydio's feet to the fire than you about a tenet that's been at the core of flying since day one ... i.e. The "pilot" has ultimate responsibility for the craft he is piloting. No slick advertising, no ground breaking advances in technology, no unbelievable low price should ever sway someone for thinking otherwise. To do so just gives our wonderful legislators that much more chance to make things even better ...insert a sarcastic emoticon.I disagree with everybody who states that that crash is Klogee's fault. The Skydio 2 is marketed as a drone that will not crash. Of course, it has its limits like any technology, and that's totally fine. What I would expect from such a beautiful piece of AI-powered technology like that, though, is that it knows its limits and acts and warns me accordingly.
I see 3 ways the drone could "think":
I get the impression that Skydio 2 implements something most similar to #1. That means that the drone itself relies on its own OA to work reliably by default and therefore expects too much from its own OA.
- By default, fly. If an obstacle is detected, either avoid obstacle or stop.
- By default, fly. If OA doesn't work reliably (e.g. based on lighting conditions or if the OA cameras are dirty/obstructed), stop, show a warning and allow the user to override (which voids warranty in case of a crash). If OA works reliably and if an obstacle is detected, either avoid obstacle or stop.
- By default, don't fly. If and only if OA is working reliably and if no obstacle is detected, fly. If OA is not working reliably, show a warning and allow user to override (which voids warranty in case of a crash).
This is a consumer drone and as a pilot, I don't receive any specific training around the abilities of the drone's OA. "Avoid low light and thin branches" is not a clear instruction as I don't know what is too low light or which branches are too thin.
I as a consumer drone pilot am by no means qualified to reliably determine whether the drone's OA is working at any given time based on all the parameters the drone assesses (and that are not even known to me) or not.
Is there a way to make sure that every consumer drone pilot is aware of their drone's OA's limits, based on countless factors that are not even known to them? Probably not.
Is there a way to make the drone's AI aware of its own limits and to warn its pilots when limits are reached/exceeded? Yeah. My Mavic Pro that came out years ago has been doing it ever since. If it's too dark for OA, it shows a warning and I as the pilot know that I cannot rely on the drone's OA anymore.
Skydio 2 also shows you a warning when it's too dark, so as pilot, I assume that it is not too dark and that obstacle avoidance is fully functional if I don't get that warning.
If I get the warning, I will not rely on the drone's OA. If I don't get a warning, I will rely on the drone's OA. If I cannot rely on the drone's OA when I don't get a warning, I don't need a drone with AI-powered OA for $1k and I'll fly a $50 drone that I know will crash into everything and anything I make it crash into and I'll gladly assume full responsibility.
Well said! The limits need to be clear in documentation. Their own videos hype its capabilities.I disagree with everybody who states that that crash is Klogee's fault. The Skydio 2 is marketed as a drone that will not crash. Of course, it has its limits like any technology, and that's totally fine. What I would expect from such a beautiful piece of AI-powered technology like that, though, is that it knows its limits and acts and warns me accordingly.
I see 3 ways the drone could "think":
I get the impression that Skydio 2 implements something most similar to #1. That means that the drone itself relies on its own OA to work reliably by default and therefore expects too much from its own OA.
- By default, fly. If an obstacle is detected, either avoid obstacle or stop.
- By default, fly. If OA doesn't work reliably (e.g. based on lighting conditions or if the OA cameras are dirty/obstructed), stop, show a warning and allow the user to override (which voids warranty in case of a crash). If OA works reliably and if an obstacle is detected, either avoid obstacle or stop.
- By default, don't fly. If and only if OA is working reliably and if no obstacle is detected, fly. If OA is not working reliably, show a warning and allow user to override (which voids warranty in case of a crash).
This is a consumer drone and as a pilot, I don't receive any specific training around the abilities of the drone's OA. "Avoid low light and thin branches" is not a clear instruction as I don't know what is too low light or which branches are too thin.
I as a consumer drone pilot am by no means qualified to reliably determine whether the drone's OA is working at any given time based on all the parameters the drone assesses (and that are not even known to me) or not.
Is there a way to make sure that every consumer drone pilot is aware of their drone's OA's limits, based on countless factors that are not even known to them? Probably not.
Is there a way to make the drone's AI aware of its own limits and to warn its pilots when limits are reached/exceeded? Yeah. My Mavic Pro that came out years ago has been doing it ever since. If it's too dark for OA, it shows a warning and I as the pilot know that I cannot rely on the drone's OA anymore.
Skydio 2 also shows you a warning when it's too dark, so as pilot, I assume that it is not too dark and that obstacle avoidance is fully functional if I don't get that warning.
If I get the warning, I will not rely on the drone's OA. If I don't get a warning, I will rely on the drone's OA. If I cannot rely on the drone's OA when I don't get a warning, I don't need a drone with AI-powered OA for $1k and I'll fly a $50 drone that I know will crash into everything and anything I make it crash into and I'll gladly assume full responsibility.
Well said! The limits need to be clear in documentation. Their own videos hype its capabilities.
Skydio has pushed the bar up with this technology. Given the high volume of footage gathered where SD did not fail, as seen on YouTube and elsewhere, where SD is pushed through forests and bridges (high risk) I'd say the failure ratio is low.
The Pilot is always IN COMMAND. Don't fool yourself that any technology is perfect. Commercial airliners have autopilots that can land in a storm but when was the last time you saw one without a pilot?
Skydio is not perfect and never will be, there will always be that perfect storm. This guy was pushing into a complex environment with changing light conditions, even a seasoned fpv Pilot would have had a hard time in that scenario.
Marketing aside SD allows us to get better perspectives, more usable footage when flown properly and with COMMON SENSE. Don't believe the hype, marketing is marketing all the manufacturers push the envelope.
Here's my M2 attacking my vehicle in daylight with obstacle avoidence on, and there are plenty more examples of DJI drones attacking subjects with OA active. Does DJI tell you to wear a helmet when using AT? This is not a scenario of the drone avoiding or stopping at an object; rather it actively attacking the object.
No technology is perfect and maybe we expect too much sometimes. I always fly responsibly and never give up LOS or pilot control. I see autonomous flight as an aid to my piloting.
GO4 constantly warns of wind, nfz, interference, updates, imu, battery warning, temperature - it's rather annoying and akin to a lawyer covering his *** so when something happens they can say we warned you. My DJI drones have crashed as much as any drone without the warnings.
Skydio will get better with time but it will never be what some expect, especially at this price point.
BTW in this video I could not override Active Attack with the sticks.
When I saw you joined this forum I thought here it goes. Every place I see you on the web with regards to a drone you're arguing for a DJI product or trying to convince someone that they're in the wrong according to you. If there was an ignore option here I'd have you on it.Dude...why does DJI have to come up all the time with you? DJI has never claimed the Obstacle Avoidance was perfect. In fact, to my knowledge DJI has never released videos of their employees or proxies demonstrating Active Track in any way even coming close to Skydio. Why not mention Yuneec who DID demonstrate and market their Intel Realsense Obstacle Avoidance as being capable of navigating "complex environments"? Besides, you know for a fact that there have been several updates to DJI's Active Track since this incident... Maybe "changing light conditions" or "complex environments" were at play for the M2P as well? Geez... I mean, you take a drone that's sold millions of units and point out the hundreds of failures but the Skydio 2 has already had several documented crashes with very few sales... With a drone that was marketed as "uncrashable"...
I don't know what happened with your drone but have you ever posted the flight logs? Did DJI take care of you? You've brought this up several times but I'm not sure what the resolution was? Since it's ALWAYS the pilot's responsibility, I guess this incident was your fault, right?
When I saw you joined this forum I thought here it goes. Every place I see you on the web with regards to a drone you're arguing for a DJI product or trying to convince someone that they're in the wrong according to you. If there was an ignore option here I'd have you on it.
Ha, I found it. Ignored!
Look back DJI and thier reps DID promote Active Track as having equal capabilities to a Anafi or Typhoon. They not only represented Active Track as having GPS to GPS but using a device barometer for dynamic altitude, a video was produced to prove such. Talk about misleading marketing. Showing an Inspire flying sideways at 45 mph following a vehicle when they didn't even have profile mode.
This is not what uspets me the most....logs lol. DJI knows active track misidentifies the subject and goes into attack mode. The reps know it and are ordered to keep it on the hush. Profit is more important than safety and that is not okay with me. Profit to pay lawyers for "acceptable" losses.
I haven't got my Skydio so I reserve judgment but the jury is in on DJI Active Track.
Competition is good maybe the Mavic 3 Active Track will work as intended. Personally I don't trust any put optical tracking. In the meantime DJI products are excellent, is that what you want to hear? Just use it for what it is good at, aerial photography and don't believe the marketing.
Bro, I have a couple dozen Active Track tests without any incidents. That's all I'm saying. I'm not trying to say you didn't have a problem, I'm saying that I haven't. Should I trust my 500 flights with the M2P or your incident? BTW the M2P has profile tracking... The Inspire has a 360 degree rotating camera, it doesn't need profile tracking... I might be missing something. I was watching the M2P launch live and I don't remember the Anafi or the Typhoon H being mentioned at all. I've also never seen any DJI representative claim GPS to GPS tracking... Again, I'm not sure I understand what you mean. Quite frankly, I'm pretty confident that I've never seen DJI do the type of trash talking Skydio has done in their ads and marketing materials. I may be wrong though. DJI has a YouTube channel, I've watched pretty much every video and haven't seen any mention of other companies.
For the record, the Parrot does an excellent job of GPS tracking to the controller/phone. I don't know how well it does over varying elevations but I seem to remember that it had problems. I've taken the M2P over varying elevations and it did "okay" but it's not perfect by any means. As you may know, I had the THP and had "issues" with it so I returned it.
I agree that GPS tracking combined with optical is vastly superior to either type alone. That's why I've dropped about $2000 on the Skydio 2...so far. You (and dirt bike boy) have very specific use cases that the Skydio 2 was seemly designed for. Unfortunately, my use case and possibly 90% of drone pilot's use case were an afterthought. We'll see what the future will bring but I doubt DJI is going to give up on the compass anytime soon so the Skydio 2 will continue to be my construction site, bridge, crane, etc filming drone for the foreseeable future.Even after I buy the Mavic 3 Pro, Mavic 2 Pro+, Mavic 2 Pro Platinum or whatever they decide to call it.