New Skydio2 problem

[Johny5]

I had my first flight with my new drone yesterday using the beacon and an iPhone 12. The beacon synced to the drone and the phone connected to the beacon wifi. I did a hand launch and set it to fixed tracking at the furthest distance, then took off on my EUC.

Things started well and then about 30 seconds later I hear a beep. I look at my phone and it says phone disconnected from beacon. I’ve got my phone strapped to my left forearm and the beacon in my right hand. WTH.

The drone is now way behind me and apparently flying to my last GPS location. It arrives above me and I use the magic wand to put it back to the side and it responds correctly. I take off and within a few second it does that same thing again.

I’m now thinking I’ve just wasted a lot on money on a pice of crap. I decided to bring the drone in closer to see if that helps. I take off and go for a few minutes and it happens again. The drone is literally 15 feet to my right and it takes off flying back to a point behind me.

At this point I shut everything down. I then decided to try just the drone and beacon and not connect the phone. After that it worked fine with no more disconnects.

So why would the phone disconnect from the beacon? And why would the drone care? I thought it should track me and if it looses visual track then it will use the beacon to try and locate me again. But it was right next to me looking right at me and when the phone disconnected from the beacon the drone just took off flying away.

Thank you.

 

[Ridefreak]

Its switch between visual and gps should be seamless, I've never had it look for me after loosing visual tracking. From my experience; at some distance (or subject image size) it decides to switch over to GPS tracking, it's usually somewhere near the second standoff distance on the beacon screen. When it does this the only indication is the icon changes on the screen. From your description it almost sounds like the phone switched and connected to the drone. I know you said that the phone connected to the beacon and that's what you want. The tracking is very limited using the phone and it will have trouble staying connected when it's used to track like the beacon does, esp with a faster moving object and or distance.

You're likely going to test it out more, I'd watch both the tracking indicator (visual vs gps) as well as checking the wifi when both the phone and the beacon are connected looking to see if for some reason the phone is taking over the connection to the drone. Also I'd see if there's some correlation to it switching to GPS and the loss of tracking, if it can't switch to gps for some reason it's going to lose you pretty quickly from what I've seen. Just standing still with both connected you should be able to watch it switch over to gps as you increase the standoff distance. Also it's brand new, I swear the SD gets better at tracking the more it's flown and required to perform that function. Mine got to where it almost never lost tracking, maybe once in 8-10 flights. Did the update and it lost me 3 or 4 time the first flight flying in a super easy wide open terrain. Next flight it lost me once, following two flights it didn't lose me at all. All were in the same area under the same conditions.

 

[Johny5]

There is just something very wrong with mine when using the app and the beacon. I’ve put the phone in airplane mode, made sure the phone was connected to the beacon and that the beacon was the only network that the phone could connect to.

It frequently disconnects and causes the drone to stop tracking me. Even when connected the video is very choppy. I’ve done a hard reset on the phone. If I don’t use the phone and just use the beacon, it tracks me perfectly and never disconnects.

 

[VEGASROBBI]

I think the current firmware is flawed and the guys that are having problems, like me, did not experience this in the previous firmware.

I'm going to try my next flight with the beacon only and no phone.

I didn't open a ticket but I did send a comment to Skydio support. Since many of us are having the same issue hopefully they can get it sorted with new firmware. Unfortunately many of us experiencing these issues just have to wait until we get help from Skydio.

 

[baksideDisasterFlicks]

I had this first flight out with new firmware. Mad disconnects from phone with beacon errors and beacon and phone were sitting side by side. Now I've had 5-6 more flights and last flight out was flawless. Now I was riding in a really open desert environment so that probably helped but part of me is like how smart is this thing? Is the AI learning how I fly and my hardware? Dunno but I'm really stoked about the last flight it was amazing. Four batteries, no disconnects, flew full battery with no disconnect for 28 minutes...

 

[VEGASROBBI]

I flew today with beacon only, no disconnects, no GPS tracking it was perfect. I think the issue may lie in the phone app/beacon because I only had issues with the app open.

Maybe the AI is allowing the S2 to learn because it does seem to get better using the app the more you fly it, but for me I'll use the beacon only until the phone app is sorted.

 

[dmac1]

I flew today with beacon only, no disconnects, no GPS tracking it was perfect. I think the issue may lie in the phone app/beacon because I only had issues with the app open.

Maybe the AI is allowing the S2 to learn because it does seem to get better using the app the more you fly it, but for me I'll use the beacon only until the phone app is sorted.

What phone are you using? Mine flew fine today with app to beacon, beacon to drone. Samsung S21, Android 12.

When I was looking at the app, I saw Skydio mention there are problems with some devices....now I can't find where that info was....

EDIT - I found the info about device issues:
Go to app info, Getting Started, scroll down to App Compatibility

 

[VEGASROBBI]

Google Play says my device is compatible. It is a Samsung A12 running Android 11. I've been using this device for months and haven't had an issue till the current update. Like RF says the more I flew it the more "normal" it flew with the app. Just flies better with the beacon only right now.

 

[dmac1]

I'm no AI expert, but I dunno about this 'the more I fly it, the better it gets' thing. I don't think AI is there yet...especially in a device that only costs $1300, and which isn't sold as having the ability to learn.

My previous drone didn't get better the more I flew it...it stayed pretty much the same with some totally effed up behavior toward the end. Maybe it was defective out of the box and never had the 'brains' to get better and maybe that is coloring my thinking....but even if so, I'm still skeptical.

I haven't flown my new drone enough to be able to say what it will do, but maybe I'll take it out again to today to do a walk test via app only in my yard and see if it still circles while it tracks like it did the other day, or whether yesterday's session following me on my bike 'taught' it something.

 

[dmac1]

Ok, it hasn't learned anything....other than maybe it cemented its 'thinking' that it doesn't like tracking me as I walk! LOL

I did a walking/tracking test flight in my yard via app alone with drone to first one side, then the other. Same as before, drone does circles, seems worse on right side vs left. Possibly if I had it behind it would be better, didn't try that. I positioned it with sun behind it going both ways.

I mentioned elsewhere that this new drone seems to be quicker than my original one on launch/turn around. That quickness seems to be contributing to the circling issue...coupled with my walking being slower than it wants to go. My original too didn't like tracking my walking pace, but it would just speed up and slow down. No pulling back into a circle aspect.

To try to explain it, essentially, it can start off ok, but then it gets ahead, so it slows. It doesn't stay in a flat plane when it slows, it tilts back like a helicopter coming in for a landing (flares I think they call it), and since its kinda sideways to me, it also pulls away at the same time. So it slows and pulls away, then accelerates and moves closer. Basically does circles with brief kinda straight stretches...can see the motion on app as I walk/watch.

Tomorrow will be a dirt bike test. I suspect the circling thing will not occur in that environment (I didn't see it in yesterday's bike test).

 

[Johny5]

Tech support got back to me and now they want me to try reinstalling the app and test with another phone if possible. I’ve got an iPad I’ll try it with as soon as we get a calm day. Been lots of wind lately.

I’m very skeptical that it has anything to do with the phone itself, as you folks are having the same issue on androids. But I’ll try regardless.

I’m also pretty skeptical about the learning, although I did have one successful phone to beacon flight recently. About half a battery. But I also did a factory reset (again) on the beacon right before that flight. That was that last opportunity I’ve had to test. 1 out of 1 since the 3rd reset doesn’t mean much

 

[Johny5]

Oh I just thought of one other thing that I changed before my last successful flight. I changed the data format from 1080p to 4K. I’d previously been using the lower resolution to keep file sizes down.

Now it seems safe to assume that the visual data being used for tracking has nothing to do with the format being written to memory. But could it somehow be related to a bug in how data is streamed to the phone, and that is what’s causing the interruption in the phone connection?

I know that may be wild speculation, but maybe you folks could include your video format choices in your evaluations.

 

[Ridefreak]

I don't think I ever attributed the over time improvement to AI, but often sophisticated electronic devices build data tables that are used as part of it's operation. Unless you have a Tesla in your driveway the SD is likely the most sophisticated electronic device you own, it's not unreasonable to think the SD might do something similar. I've seen improved behavior over time originally when the drone was new and after the update despite it crashing for an apparently different reason. Despite the crash, tracking improved after that incident going from multiple disconnects to none and the gimbal shake improved almost to where it was before the update. SD requested I send the drone back for replacement or else it's likely I would still be flying it.
During what I'm describing as the flights where it improved the conditions were constant including obstacles as well as the RF environment. How does it improve that significantly in constant conditions and maintain that improvement? I don't know but to me its obvious the update reset something that affected it's connection and gimbal stability and it was very similar to the performance when the drone was new.

No exaggeration, I'm up around 100 flights that are conducted in similar conditions so I have had allot of opportunity to witness what I'm describing. You can shrug it off but I know what I'm saw and before the update it performed excellent, it wasn't that way when it was new, nor after the update but both times it improved significantly.

When I get the replacement I have a plan to start the flights off easy and simple in open areas with good reception, just like I did when new and move on to more difficult operations after some flights. You can dismiss this approach but it's worked for me and I had arguably better success with the SD then many of the others posting on here. So have some others who witnessed similar behavior.

 

[VEGASROBBI]

In the automotive world we call them adaptives. Most modern engines and transmissions and other components learn over time. You can call it AI or adaptive learning, whatever you want - I'm thinking the S2 has some of this.

GM had a class action lawsuit over their 8-speed transmission because they shifted terrible. It could take weeks and sometimes months for that transmission to learn but once it did it was beautiful. GM came out with a quick learn procedures but they never worked, so the drive cycle is like three pages long to get the solenoid characterization complete. You can call it mechanical learning or AI but there's definitely something to it. The transmissions are torque management based so they learn about the weight of the vehicle, how you drive... then adjust to the conditions. I'm assuming the S2 is doing some of this with the gimbal and flight characteristics.

 

[dmac1]

I don't think I ever attributed the over time improvement to AI, but often sophisticated electronic devices build data tables that are used as part of it's operation. Unless you have a Tesla in your driveway the SD is likely the most sophisticated electronic device you own, it's not unreasonable to think the SD might do something similar. I've seen improved behavior over time originally when the drone was new and after the update despite it crashing for an apparently different reason. Despite the crash, tracking improved after that incident going from multiple disconnects to none and the gimbal shake improved almost to where it was before the update. SD requested I send the drone back for replacement or else it's likely I would still be flying it.
During what I'm describing as the flights where it improved the conditions were constant including obstacles as well as the RF environment. How does it improve that significantly in constant conditions and maintain that improvement? I don't know but to me its obvious the update reset something that affected it's connection and gimbal stability and it was very similar to the performance when the drone was new.

No exaggeration, I'm up around 100 flights that are conducted in similar conditions so I have had allot of opportunity to witness what I'm describing. You can shrug it off but I know what I'm saw and before the update it performed excellent, it wasn't that way when it was new, nor after the update but both times it improved significantly.

When I get the replacement I have a plan to start the flights off easy and simple in open areas with good reception, just like I did when new and move on to more difficult operations after some flights. You can dismiss this approach but it's worked for me and I had arguably better success with the SD then many of the others posting on here. So have some others who witnessed similar behavior.

Assuming you were talkin to me cuz I was questioning its learning ability.

You have tons more time with S2 than I do. I didn't know you had troubles when it was new, that markedly improved with more flights. Mine didn't improve, hence my skepticism. Mine was never very consistent. Sometimes it would work seemingly fine, sometimes not so much. What frustrated me was, despite 40ish test launches, maybe even more...I lost track, I never could nail down a startup sequence that it would do over and over. Seemed like it should be doing something repeatable....hence part of my complaints to Skydio.

One thing seems apparent, the update screwed stuff up, possibly those data tables you mentioned...I'd not heard of those before, but that does make some sense actually that it builds them and draws upon them!!

 

[Johny5]

Well wouldn’t you know it, I managed to test 2 more batteries using app to beacon and had a total of…zero problems. Wow.

I tested half a battery with an iPad, and 1 and a half with the original phone which had all of the disconnects. All worked perfectly. Recording was at 1080p on a EUC at around 20mph in an open road area.

On the closest 2 distances it maintained visual track. At the furthest 2 distances the app would flicker the gps marker over my body constantly but the tracking remained smooth. It stayed right on me the whole time. The beacon showed the eye symbol and never showed gps during the app flickering. It seemed to have no trouble rapidly transitioning its tracking method.

But the most important thing was that it never disconnected and flew away like it was doing originally. And I haven’t reinstalled the app yet. I wanted to test 1 change at a time, which was using the iPad.

So was the fix due to learning? 3 factory resets? A stealth update? Sunspots? Who knows. And is it actually fixed or was this just a fluke? Time will tell I guess.

I say that I’m skeptical that it can learn not to disconnect or to fly better. Being skeptical means that while I have doubt I still recognize that you could be absolutely right

I also did a little bit of walking. It didn’t do any of the small circles but I did notice very pronounced response to my body movements. If I swung my arm while walking, the drone swung with it. If I stopped moving my arm but maintained the same walking gate, the drone stopped swinging. It was kind of funny to watch. So could the small circles be responding to some body movement or gesture?

 

[Ridefreak]

Assuming you were talkin to me cuz I was questioning its learning ability.

You have tons more time with S2 than I do. I didn't know you had troubles when it was new, that markedly improved with more flights. Mine didn't improve, hence my skepticism. Mine was never very consistent. Sometimes it would work seemingly fine, sometimes not so much. What frustrated me was, despite 40ish test launches, maybe even more...I lost track, I never could nail down a startup sequence that it would do over and over. Seemed like it should be doing something repeatable....hence part of my complaints to Skydio.

One thing seems apparent, the update screwed stuff up, possibly those data tables you mentioned...I'd not heard of those before, but that does make some sense actually that it builds them and draws upon them!!

Back when it was new there wasn't alot to compare it to so much of that was attributed to not being familiar with the drone and some of it was. But after seeing it again after the update I recognized the behavior.

 

[dmac1]

Well wouldn’t you know it, I managed to test 2 more batteries using app to beacon and had a total of…zero problems. Wow.

I tested half a battery with an iPad, and 1 and a half with the original phone which had all of the disconnects. All worked perfectly. Recording was at 1080p on a EUC at around 20mph in an open road area.

On the closest 2 distances it maintained visual track. At the furthest 2 distances the app would flicker the gps marker over my body constantly but the tracking remained smooth. It stayed right on me the whole time. The beacon showed the eye symbol and never showed gps during the app flickering. It seemed to have no trouble rapidly transitioning its tracking method.


I also did a little bit of walking. It didn’t do any of the small circles but I did notice very pronounced response to my body movements. If I swung my arm while walking, the drone swung with it. So could the small circles be responding to some body movement or gesture?

Zero problems is good! Mine tracked me real well yesterday and avoidance of trees and branches was amazing! I had app to beacon, beacon to S2.

I put my phone in a pocket and I have a little sleeve on my Camelbak's shoulder strap for the beacon when I'm on my dirt bike, so I can't keep an eye on the displays while riding. I have watched the displays while I walk...I don't recall ever seeing mine flicker GPS over my body, but maybe I just wasn't looking at the right moment. Mine shows a circle going around me and actually I've not seen its following mode displayed via app (ie eyeball, eyeball with check, gps like on beacon).

I am pretty sure my drones circling wasn't due to arm swinging, but thats interesting! Did you have the beacon in the hand with the arm swinging? Pretty precise if so! In my case, it was pretty clear that the drone doesn't wanna go my walking pace...sometimes it paces me just fine (like in my initial tests in exactly the same place) but sometimes it speeds up/slows down and mostly when tracking on my right side it does the circles.

As I expected, on my bike yesterday, per the footage, it did no circling.

 

[Johny5]

Actually, yes, the beacon was in the hand that was swinging. But the drone was very close and on visual track at the time so I’m sure it was not related. I believe it visually interpreted the arm swinging as me walking faster and slower.

 

[Trailryder42]

It might be interesting to check to see if Sun activity and CME/solar weather that has been hitting the Earth, correspond to the glitches in Drone/GPS/Phone. We've been having a hell of a lot of Sun activity hitting us.