Just wanted to fill you in on a small UAV altitude record that was just set (still unofficial at the moment) using an autopilot similar to the ones discussed here. Here's a story about it: http://news.stanford.edu/news/2009/september14/airplane-self-pilot-091809.html
The record attempt was made by a group of graduate students who took a course this spring in the Aero/Astro Department at Stanford University. I was the teaching assistant for the course. This year the goal for the course was to design a small UAV to beat the remote control altitude record (~11, 000 ft). There were four teams of 4-5 students that had 10 weeks to go from a blank sheet of paper to a successful vehicle. The autopilot and propulsion system were provided, but the teams had to write their own control software and design/build the vehicles. The only sensor allowed was GPS (+ an barometric pressure sensor for the record attempt).
At the end of the class in June we had a flyoff where the teams had to demonstrate their vehicles by repeatedly climbing between 50 ft and 400 ft (to stay within FAA regulations). All four team succeed in demonstrating greater than 10,000 ft of cumulative altitude gain. A few students continued working on the project over the summer, culminating in a world record altitude attempt a couple weeks ago at the NASA Dryden Flight Research Center.
The altitude record attempt was for an autonomously controlled electrically powered UAV weighing less than 5kg, FAI Category U.2a Group 2. The news article gives a good account of the record attempt flights. The record altitude was 2177 m (7142 ft), which ended up being limited by winds aloft, not battery energy. Only about 40% of the charge was used getting to the record altitude.
Here are some pictures of the airplanes and the autopilot/GPS boards.
I designed and prototyped the autopilot for the course early this year. It uses the dsPIC30F4011 microcontroller and the uBlox LEA-5H + sarantel helical GPS antenna (basically the same as is being sold now, but in a slightly larger footprint). The xBeePro or xBeeProXSC is used for commands/telemetry. RC controller commands are read by the custom ground station software using a transmitter-to-USB cable and sent over the xBee to the airplane. This software based failsafe was not ideal, but it was the lightest option. The autopilot has inputs available for other sensors (thermopile, IMU), they just weren't integrated for the class and there isn't any software written to handle them.
If anyone is interested I should be able to post a BOM, the board files and the autopilot/ground station software, but it likely won't be supported at all since we are working on a smaller/lighter design to replace this one.