Re: [Paparazzi-devel] Discussion point around toss and fly drone

[Felix Ruess]

Hi guys,

with most currently used sensors you might as well skip the initial AHRS alignment (lowpassing gyro to get initial bias).

The filter will rather quickly estimate the bias anyway... Also there is no assumption about the quad being level at startup...

The work Davide did on monocular vision is really nice, it will quickly recover scale and start the visual odometry/SLAM to keep position.

So for a "toss and fly" outdoor drone where you have GPS available this is not the critical part...

Cheers, Felix

On Wed, Aug 26, 2015 at 10:09 PM, Oswald Berthold <address@hidden> wrote:

hi,

wanted to throw that in, davide scaramuzza's group published some
related work earlier this year [1], see refs at the bottom of that
page. there seems to be additional ROS power at work too [2]. i don't
know how much lily is actually using vision for stabilzation but it does
use a dedicated relative positioning system (wearable dongle). as michal
says, its probably mostly about changing some assumptions about the
IMU/AHRS init procedure, levelness etc.

bst, opt

[1] http://rpg.ifi.uzh.ch/aggressive_flight.html
[2] http://www.ros.org/news/2015/05/the-lily-camera-developed-using-ros.html

David Conger writes:

> Thank you Michal for that thoughtful reply. I agree completely in all respects.
> My BeBop and ARDrone2 take considerable time before I fly them doing
> Mag calibrations, waiting for GPS lock etc. FWIW: My BeBop has Parrot
> firmware for the moment I am waiting for someone (TUDelft) to publish
> their workings with the SkyController (Wiki) engineering and latest
> with BeBop firmware work...I search, and find node.js and other simple
> things but nothing very detailed so I wonder if those who do the work
> are too busy to update the rest of the World on what they find?
>
>
> On 8/24/15, Michal Podhradsky <address@hidden> wrote:
>> Hello David,
>>
>> it is truly an impressive video. But it is a promo video, so I would
>> imagine that not every single time it takes off so flawlessly and they used
>> the "best" shots only. (here is the link btw https://www.lily.camera/)
>>
>> I don't think anybody tried it with Paparazzi, although it is possible.
>> IMU/AHRS can initialize very quickly, and you don't necessarily need to
>> initialize it levelled. One example would be external AHRS/INS devices,
>> such as GX3 or Vectornav, which give you orientation quickly after startup
>> (I assume they use gravity vector as a reference). Lily will have some
>> cheaper solution, probably even the very same sensors as Lisa board. So the
>> "only" missing part would be a better AHRS algorithm that doesn't need a
>> long initialization.
>>
>> I don't know about the GPS - maybe you can get some "fast" lock, although I
>> doubt it. Maybe Lily flies without GPS just in altitude hold (barometer)
>> until it gets a lock. The wrist control device is connected via bluetooth -
>> so maybe some kind of RSSI is used to guide the copter while it waits for
>> GPS lock.
>>
>> Anyway, if the company delivers what they promise, it will be a cool
>> product in a reasonable price range.
>>
>> That's my 2 cents:)
>> M
>>
>> On Wed, Aug 19, 2015 at 6:24 AM, David Conger <address@hidden>
>> wrote:
>>
>>> Hello, I hope everyone is having a good Summer. I start this
>>> discussion because I am often interacting with people who consider a
>>> MAV for a business or research etc. It seems the "Holy Grail" to have
>>> something that can power on and fly (even just hover in place) with
>>> zero setup. The follow a GPS dongle is also something I can see many
>>> uses for (following wildlife or toss the dongle into the sea and let
>>> the current carry it etc).
>>> I saw a surprising drone video for "Lily". A feature is no setup just
>>> take it out and toss it to fly. Video seems to support this.
>>> Interestingly in this discussion list I remember someone asking about
>>> just this for their Paparazzi quadrotor.
>>> How is this possible to take Lily out of a backpack, power on (not
>>> even level on the ground), then toss it and it recovers and flies? How
>>> will Lily know about magnetic variation? GPS fix? Level attitude so
>>> quickly?
>>> It seems an interesting discussion because in all these years I have
>>> never seen anything like this or thought it possible, then again
>>> things are only impossible until they are done no?
>>> Has anyone done such a thing with Paparazzi based multi-rotors?
>>> Cheers,
>>> David B Conger
>>>
>>> _______________________________________________
>>> Paparazzi-devel mailing list
>>> address@hidden
>>> https://lists.nongnu.org/mailman/listinfo/paparazzi-devel
>>>
>>


--

Dipl. Inf. Oswald Berthold

Institut für Informatik
Adaptive Systeme (RUD25, Raum 4.119)
Humboldt-Universität zu Berlin

 Tel.: +49 30 2093-3116 | Fax.: +49 30 2093-3045
Email: address@hidden

_______________________________________________
Paparazzi-devel mailing list
address@hidden
https://lists.nongnu.org/mailman/listinfo/paparazzi-devel