Notices

I think you hit the nail on the head here.
It's probably much cheaper to buy the components individually. Having my n900 in a crash would be bad news.

Definitely not! But OTOH, Nokia gave me my two phones just to build my application... (see What is PUSH N900 for more info)

Previously I built systems using RC and video TX/RX parts, but they were not as elegant, easy to use or powerful as the one using the N900. And the N900 is so much more fun to use, too...

Now I'm looking for a cheaper but similar platform (maybe a BeagleBoard or Gumstix) to use together with good compact cameras (wide angle, 12MP+ res.). It would replace the N900 on the kite, but would still stream video to and be controlled from my N900 on the ground.

Hmm... maybe I should use one N8...

A free kick in the groin still hurts. Err.. I mean, a free N900 still costs 600E. I can think of 5000 uses for a free N900 (with selling being #1). with the rest of 4998 being at most 2 meters off ground.

Frankly, I like your project. But I would only do it if the phone could survive crashes (tucked away).

It still sounds like a project that is warranted by "free N900s" not "free N900".

My suggestion is: don't get kicked in the groin!

I fly cameras almost as expensive as the N900 (which costs about US$ 500 in the US, according to a quick search). I managed hundreds of flights in 5 years, and until today I never had a crash. It almost happened once or twice (when the wind suddenly died), but even then, it was not free falling and the rig that holds the camera would have absorbed most of the impact, if not all (it is designed that way).

I can understand that... but as I said before, flying the phone on a kite is, for me, way more fun than selling it!

Nobody can guarantee that. There are risks involved (in anything we do, right?), but if you manage them well, you minimize the chances of an accident. And if there were no risks and no challenges, it would not be as fun and rewarding, too.

I know the N900 is expensive for a cellphone, but for an autopilot, its not that bad.

I built a UAV in 2001 using an off-the-shelf autopilot that cost $5,000. For that price, the autopilot did not include a battery, GPS, or modem.

I would love to see an N900-controlled UAV!

Actually, I would really love to see someone make a USB servo controller for the N900.

We need USB host first and while there are some really smart people working on it it's not really working yet and might never be (at least without doing some hardware hacking)

BT works for interfacing with something that can generate servo controls (you wouldn't want to power it from the USB in any case)

As for other considerations: N900 is actually not a good device for autopilot or telemetry but that is not the point, the point is that it would definitely be a cool hack (a concept some people seem to have trouble understanding). Even if you would want to use python or some other interpreted language on top of non RT kernel there are better options (like the gumstix modules, though they're not cheap either), for a bit lower level there's a bunch of microcontrollers, the "arduino" platform (ATMega uC plus a bit of supporting HW and a standard layout for connecting modules designed for various tasks) being very popular for the purpose.