Thanks all!
When I was working on Tacocopter, we used 8-CineStars from Quadrocopter[1]. They were originally designed to fly pro cameras, so they have an incredible payload, great flight characteristics, all carbon fibre, etc. Downside is they're very expensive.
Most of our development was in the software, so we wanted a hardware solution that "just worked" Dealing with DIY motor controllers that can burn out really slows down the process. Aside, if you end up using the MicroKopter platform, I reverse engineered a lot of their serial protocol for our telemetry stack-- check it out on Github[2].
The cool thing to me about MAVs is they essentially become a flying computer. For your friends, I would pick the one that has enough power for decent acrobatics+payloads, but more importantly the easiest software development stack. Stuff from 3DRobotics is a good starting place.
[1] http://www.quadrocopter.com/
Would giving the drone a 4G radio with a data channel open to an EC2 instance be considered "autonomous", or "remote-controlled?" It always seemed like the best option to me, since the server issuing guidance could read from whatever internet sources you like (like grabbing a route from Google Maps, or somesuch.)
Anyway, I would add another big vote for 3DRobotics. They're the organization behind the enormous and successful Ardupilot project, which is an open source community that develops hardware and software for autonomous model vehicles (helicopters, multirotors, airplanes, and cars).
Their ready-to-fly multicopters and planes are packed with functionality and reasonably priced, although they are probably out of the price (and ability) range of someone completely new to the hobby. As other people have said, it's essentially a necessity to grab a toy quadcopter like the Syma X1 for around 40 USD to get a hang of flying them (and realizing how easy it is to crash them) without the chance of a major financial loss.
We definitely want some degree of autonomy with the drone, and most likely a quad/octoroter. A big things for my friends is payload and possibly custom actuators (grabber or claw).
The biggest benefit is that it lets you build your piloting chops without a huge financial risk. An acquaintance of mine is working to get drones into the hands of professionals who aren't interested in becoming drone experts (Realtors, etc), and he uses the Syma X1 to train folks before giving them the keys to larger, more dangerous and expensive drones.
Both are android users and prefer open-source hardware and software. The Parrot AR 2.0 looked good at first, but I'm unsure as to the level of community support for the android software.
They are also very skilled at low level programming, so some of the newer Linux based drones are appealing.
Runs the PX4 autopilot (which I believe is hackable), can be controlled by RC, has good support for camera, gimballs, attachments, etc.
The Parrot firmware is closed and I don't know about other firmware implementations, but the protocol is open so you can write clients to control is - https://github.com/felixge/node-ar-drone being one of the most well known.
If you have lots of time on your hands, though, you could build your own with ardupilot: http://ardupilot.com
Unfortunately it will be limited in its power, unless you do an extensive engine and frame mod as shown here http://www.youtube.com/watch?v=KESy--N6ags
The 3DRobotics Iris is an exciting new product that represents a more flexible, pro-amateur level tool. They're made to be put to work, not flown like a remote controlled helicopter, so you'll find good support for programmability here.
On the other hand, it's almost $1k fully built out, compared to half that for the Parrot.
The AR Drone is pretty deluxe for a starting point, plus it actually works out the box and is affordable. Whilst having hours of fun with it one could work out what the 'ideal' is actually going to be.
Or build an AeroQuad: http://aeroquad.com/content.php?s=9af9c997f8aed9a3ad958e32bf...
There's a lot of specific distinctions between piloted-flight versus autonomous ('drone') as well as size restrictions and such. First-person view (FPV) requires a second pilot using grounded line-of-sight, as well (http://www.modelaircraft.org/files/550.pdf).
I'm by no means trying to discourage you, just know if something might be a federal offense!
If you could include prices that would be amazing, but I can always look it up if you can just post the parts!
A proper quadcopter is different than those styrofoam quadrotor things you see. A copter will take longer to learn how to not crash, but you have more precise control.
Frame: http://shop.flitetest.com/multirotors/bat-bone-tri-370-kit/ but you can easily build a frame yourself from either wood or aluminum towel holder bars from home depot.
Flight controller: http://www.hobbyking.com/hobbyking/store/uh_viewitem.asp?idp... I have this one, the 2.0, but there's a 2.1 out now too.
Motors: http://www.hobbyking.com/hobbyking/store/uh_viewitem.asp?idp...
ESCs: http://www.hobbyking.com/hobbyking/store/__25363__Turnigy_Mu...
Tail servo (don't needs this for a quad): http://www.hobbyking.com/hobbyking/store/uh_viewitem.asp?idp...
Batteries: http://www.hobbyking.com/hobbyking/store/__9394__Turnigy_220...
Radio: http://www.hobbyking.com/hobbyking/store/uh_viewitem.asp?idp...
That's all the major parts. You'll need a few odds and ends like propellers, servo extensions, connectors for the battery and ESCs, etc. Hobbyking sells a power distribution board you can use if you don't want to make your own wiring harness.
So please, stay away from the factory drones, folks.
Forget amazon, that's not a place for hackers...
But beware of the hidden costs in Radio-Controlled flight - it's never a cheap date...