Parallella Flight Controller / Robot / UAV Daughter Board

[Veqtor]

A lot of people seem interested in using the parallella as a flight controller for drones/multirotor platforms for doing various forms of enviromental analysis. A daughterboard specifically made for this would thus be very interesting...

Personally I would like to see:

* Flexible I/O that can connect to receivers typically used in drones, now I'm uncertain if it's possible to interface w. 5v logic in both directions... can we?
Typically the connections are three pin headers w. 5v, gnd and a pwm signal. This goes for input AND output.

* some new ESCs (electronic speed controllers) that are being developed will have I2C and CANBUS so this is also important.

* Barometer, Accelerometer, Compass, Gyro... these should be of very high quality of course

* One high res ADC for battery sensor, ultrasonic etc

* Maybe additional High speed USB? 2.0 or 3.0? for connection to kinect or other depth sensors for 3D mapping?

* HDMI input for image recognition and processing?

* Composite video output for transferring video via typical analog video tx?

* 3G/4G module? Xbee connector? Something else?

* Onboard regulator and voltage sensor for connection to 1S up to 6S lipo?

What are your thoughts? What does your application need? I think parallella is the ideal robotics brain so I think it's in everyones best interest to create a base platform that has enough features and flexibility for all our robotics needs. To make sure everyone is happy with the result I think we should pool our resources and ideas to make the ultimate daughter board for this.

Personally I wish to help launch an open, crowd-sourced and free 3d-wiki-earth map, where drone users and next-gen smartphone (w. 3d cam) users can help map our earth in high resolution point cloud data that can be freely used in non-profit research, entertainment, games? I realized there was a need for this when I wanted to build procedural sci-fi game where the game world is based on a real model of the earth in a worst case global warming scenario 500 years from now. Finding free data proved to be nearly impossible, only very low resolution data was available. Imagine how an open system could change this, not only could we map height in high res, but also weather, spectral information for geological and plant life analysis etc etc.

These are exciting times!

Here are some current offerings:

Multiwii Mega


Ardupilot Mega 2.0


PX4FMU

[theover]

You mean like a model chopper mixer circuit and servo adjustment ? Usually that doesn't require so much computationpower, but maybe with cameras connected and all kinds of sensors, that might become interesting.

[Veqtor]

I think this uses quite a bit of computational power:
http://www.youtube.com/watch?v=MvRTALJp8DM

Also, kalman filtering, neural networks, etc... But most important, to perform autonomous 3D scanning!

[theover]

Cool video !

I've done remote controlled ~ 1meter helium balloons with about 30W motors that was fun, but local flight control is a new dimension.

I still think it depends on how you control the motors and what type of sensor you use to get flight feedback data. Of course those loopings are fast motions, but how many variables does the (I presume) onboard controller measure and control ? For sure a camera and optics based flight control cost more than any contemporary MC!

T.V.

[aolofsson]

Veqtor,
Yes, that group has done some incredible work (and videos)! Not sure how the Parallella could help though. Seems like they are doing pretty well without us.
Andreas

[Veqtor]

[LamsonNguyen]

And then we make a holodeck! I like this daydreaming of yours, Veqtor.

[Veqtor]

Too bad it seems to only work in very dark rooms with a lot of dust in the air...

[timpart]

Ground based 3D scanning that looks quite processor intensive

Article

Video of device mentioned

Tim

[marcoschwartz]

Hello,

I think the Parallela platform is a great opportunity to finally see some easy-to-use robots that are controlled by neural networks. Indeed, there are a few neural-networks controlled robots out there, but there are usually used by universities only and the code is not available to everybody. Also, the networks are usually running on a computer because the tiny microcontrollers can't simulate large neural-networks ... which causes communication problems between the robot and the computer, and the robots are never truly autonomous.

So with the several cores embedded in the Parallela board, I see great applications in running neural networks to control mobile robots. So I would love to see a daughter board where you can plug many sensors (like ultrasound sensors) to exploit the high degree of parallelism that neural networks can offer, and end up with a platform where everybody can experiment with their own neural networks running on the robots. Anyone else would be interested by this kind of idea ?