Aside from building this site, my two big UAV projects are taking the Lego Mindstorms UAV to the next level (it's mindblowing, although I can't talk about it yet!), and building a prototype blimp UAV as a platform for a potential indoor aerial robotics contest.
This particular blog series is going to be about the development of the second of those, the creation of a sub-$1,000 fully autonomous indoor blimp UAV.
For the curious, my white paper on why we picked blimps, rather that quadcopters, helis, microlight planes, etc, is here
Now that we've settled on blimps, we had to start by trying what was already out there. To get a feel for how blimps handle, I got the best (affordable) RC blimp: the Plantraco Microblimp
, which is a lovely bit of superlight engineering, with a 2.4ghz RC system, fully proportional controls, and LiPoly rechargeable batteries (it's the silver one on our dining room ceiling at right). First lesson: blimps have HUGE latency issues. Basically the inertial momentum means that you need to anticipate all of your turns by many seconds. That means that an autopilot will not only have to know where it is in six dimensions (x,y,z yaw, tilt, pitch), but also have an internal model of the blimp's response time (which ranges from slow to slower). Fun!
The second blimp we checked out is the only autonomous blimp you can buy (without a security clearance). It's the BlubberBot
featured in Make Magazine last month (it's the the big white one in the picture above). I was particularily intererested in it because it's about the size we want for our blimp (46"--a bit less than four feet--long) and I was curious about how other people handled blimp autonomy.
(BTW, those weird pink and blue things hanging off the bottom are children's forks and spoons for ballast!. I wanted to find out how much weight that envelope could carry [answer: about 125 grams] and so I filled it with more helium than recommended. That meant extra ballast, which meant raiding the kids' silverware drawer.)
It took all day to build the kit, which requires soldering all the components to the PCB and connecting four sensors (light left and right, touch at front, and, randomly, cellphone proximity) and four output components (sound, vibration and two LEDs). It uses a pre-programmed ATMega168v
CPU and two motor driver chips for the two props. No vertical control.
You can see the PCB at right, with the cellphone sensor hanging off the bottom. Overall, it's missing quite a bit of what we want in a proper blimp UAV, from programmability and location awareness to proportional control. It's designed to seek out lights and bounce off walls, and I suppose under the right lighting conditions it might actually do that (mostly it just wanders about). But we need more.
One thing that I did learn from the BlubberBot, however: There's no need to keep the motors going all the time, as you might with a plane. The BlubberBot pulses the motors, then coasts while it reads its sensors, then pulses them again in 10 or 15 seconds. I think we'll do the same with ours. A great way to save power!