Flying my Solo this evening when it flew off and crashed across the street after my son took a picture of it while it was hovering in front of him. No warning - just went. Solo was completely unresponsive to control attempts. See the event here: https://youtu.be/m368_A9hIwQ

Uploaded log to 3DR. Really lucky no one was hurt as kids had just been watching. Second flight - battery was around 20% - getting a picture just before going to land it.

Broke 3 blades but no damage to Solo or GoPro.

Doesn't build confidence when it just decides to do its own thing...

Views: 10199

Replies to This Discussion

Makes perfect sense. To we laymen that might be expressed as "it's not as smart as it could be" or "it doesn't have the real secret down of combining all the various onboard sensors to come up with the best solution".....or something like that. 

Here is a post from over 2 years ago indicating that type of filtering on an earlier model of quad.

http://www.rcgroups.com/forums/showpost.php?p=24451035&postcoun...

Would such a filter be automatically in the OEM GPS module? Or does it have to be tailored perfectly to the combo of sensors that the particular machine uses? 

I don't know what version of Arducopter is in the Solo but Arducopter does use Kalman filters. There are some very clever and qualified people writing the code and this is an area they have spent a fair amount of time on. 

Now the new phantom 3 and Inspire GPS uses the GLONASS which is a Russia satellites along with the American sats. It says they can lock up to 16 sats. Does this help the drone keeps good lock? Would this help the Solo if it were able to use GLONASS? I have always been told to fly the least amount as possible in GPS mode which I do. As a still photog it works good, flying in stable mode then GPS hover when the shot is established. But Im sure the video guys are wanting the GPS lock for cable cam and POI autonomous flying.

Please don't get me wrong, I have no doubt that clever and qualified people are writing the code. 

I'm just basing this observation on my personnel experience between my Phantom and Solo.  My Phantom, in GPS hold, seems to drift alot initially.  After it's been flying couple of minutes, it seems to continue to decrease it's deviation over time.

The Solo did not seem to decrease it's deviation over time.  It holds within 6 feet or so, but it did not seem to improve during the flight.

I have no empirical data to prove or disprove my  personnel observation.

Mine has been rock solid very little deviation and that is why we were so surprised when it just took off. Of course 1+ hours is probably not much time to make much of a statement but we were very comfortable standing right next to it. The little blades aren't nearly as fearful as my 80's Hirobo Shuttle's!

This was nice to discover!

Going to manual mode automatically should be a big help!

"I suspect that the Solo code does not utilize any filters such as this."

This is an odd assertion to make considering it is openly and frequently advertised that the pixhawk uses an EKF for sensor fusion and state estimation.

http://copter.ardupilot.com/wiki/common-apm-navigation-extended-kal...

v3.3 and after, the EKF is the only option.  Prior to that, EKF is optional as the "old" tried and true DCM complimentary filter is used by default.  Both work very well, but the EKF "knows" when a sensor is getting sketchy and does it's best to work without it.

Hello Phillip,

Thank you for the response and information.  

I'm taking stabs in the dark while I await 3dr to get back to me on the incident posted in http://diydrones.com/group/solo/forum/topics/pilot-erorr-solo-error

As an outside observer, here are my thoughts:

  1. If this was a GPS glitch, today's software release should help as it will force the aircraft into a manual mode if the GPS looses position.  Might not save you from a mild crash, but it should allow you to quickly drop the throttle and get it on the ground.
  2. Flying that close to a house is super sketchy.  You need 8+ GPS satellites to maintain a position lock.  If you're that close to any sort of vertical object, it's very easy for a satellite to move behind the object (they're not standing still up there) and suddenly you go from good GPS to an erratic signal that may "lie" to the copter about where it is.  In most cases, getting a good GPS lock requires both overhead satellites and ones on the horizons.  This isn't to be critical of you, Most people don't understand GPS and I don't think 3D Robotics (or other drone companies) put enough emphasis on how important it is to avoid flying where nearby objects can block or reflect the signal of the satellites.
  3. As has already been pointed out, the RasPi2 can be reset with a camera flash.  It would be good to know if the PixHawk2 suffers from the same malady.  I think all commercial multicopter controllers should be inside of a Faraday cage for this exact reason.

Good luck and I hope things work out well for you.

I think these are moot points when you have a decent gimbal - that is, no benefit to flying manual and most pilots would probably lose their machines if this were true.

I am testing out the P3 and it does usually have 13+ sats and has never lost the GPS connection - even next to buildings and in large forests, etc. 

Not to say it's not possible, but all anecdotal evidence points to a vastly improved GPS lock.

3DR claims they made a decision that this was not so. That is, that GLONASS does not add anything to the mix. They will probably defend their decision, but it goes against a lot of research and reports.

"Dual-system receivers and antennas deliver a clear advantage for high-precision positioning in such applications. In fact, Taoglas’ real-world driving tests reveal quicker time to first fixes, with accuracy improving from meters to sub one meter."

Conclusion

The tests were stark in their results: GPS/GLONASS dual-system antennas definitively offer a noticeable improvement in accuracy and performance. Urban environments are the true test of performance, and dual-system antennas are clearly superior."

http://electronicdesign.com/test-amp-measurement/real-world-drive-t...

One of the reasons to implement a Kalman filter, or EFK, is to integrate all of your position data (GPS, accelerator, barometer)  and filter out erratic data.  In your scenario #2, if you lose a GPS sat or one of the sats gives bogus data, it should be filtered out and not totally destroy your vehicle position.

That was why I  was originally questioning if 3dr had implemented something such as the EFK to compensate for erratic input.

To get FAA certification for the avionics software I used to work on, we had to prove we could drop X number of inputs (GPS, altitude, INS, etc) and still maintain accurate position, speed, heading.   Losing one or 2 GPS sat signals after initial position was established was no big deal

The latest from 3DR blog (just slightly edited here) includes more on the 1.05 update. They again reiterate to be mindful of objects in close proximity. 

"Solo will now fly better in environments where GPS signal may be less accessible or more sensitive, such as areas with peripheral objects like buildings and trees that can interfere with reception. But even with this improvement you’ll want to be mindful when you choose where to fly. Please stick to wide open places without nearby trees or structures (or food trucks!) that could occlude or interfere with the satellite signals Solo relies on for its GPS lock. This is especially important for new users who might not yet be comfortable flying in FLY:Manual mode."

RSS

© 2019   Created by Chris Anderson.   Powered by

Badges  |  Report an Issue  |  Terms of Service