The basics of our imagery system is the (linux) PC on the plane, called the Flight PC, and the camera(s) it is attached to.
Currently the Flight PC is a microATX desktop PC with an Intel Atom Core2 Duo processor, running Debian Desktop Linux.
Issues with the current setup:
- Powering it efficiently is hard (12V is difficult to supply on flight batteries, and by the end of some flights is down to 10.8V)
- The microATX formfactor and computer was not made for this kind of deployment (embedded in a flying, vibrating, occasionally-rough-landing vehicle)
- Very overpowered for what we are using it for (capture images and store to disk/wirelessly transmit to the ground)
- Expensive to replace, in cost and construction time (building the custom casing for it to house in the plane
SOLUTION: The PandaBoard?
It's using TI's OMAP4xxx series, a dual-core ARM processor, that we believe is gonna be the solution to our current issues.
What it fixes:
- Powered off only 5V, this is very easily regulated down from the servo battery voltage, so will never be a problem (also uses less power)
- Smaller formfactor, can be mounted onto a panel for quick release/insertion in the vehicle.
- Fits our requirements for camera control AND has the ability to control an extra video camera. We're planning on using this to get a live video feed in addition to the stills we currently get.
- Has the right balance of computing power for our needs, anything more intensive than what we already do is offloaded to the ground station wirelessly. Runs Ubuntu Linux!
- Ethernet to connect to our embedded wireless router (Ubiquiti M5-High Power 802.11n bullet units running embedded linux)
- Hardware-accelerated JPEG compression/decompression. This means we can use the PostGIS? spatial queries to cut out overlapping sections of images and images outside of our bounds (and thus speed up wireless throughput of imagery).
Additionally this (new) platform we are going to open source, so that the DIY Drones community, the Ardupilot guys, other open-source autopilots, and even our competitors can use it! We believe it has a lot of commercial applications, including some that we've looked at in the past as an autonomous aerial imagery platform research group:
- Agriculture monitoring/management (image huge farms to save time driving around looking at crops) both visible-spectrum and infra-red
- Geological and Ecological surveying. These guys are using imagery platforms that are mostly developed in-house (with expensive engineering time)
- Crisis Management. Google has talked about the possibility of these platforms being deployed as aerial mobile phone towers/video surveys in the case of natural disasters where local infrastructure is knocked out.
- Emergency Response. We've talked to the coast guard about the possiblity of cheap aerial imagery platforms being used for search-and-rescue (open source platforms are about 1/10th the price of commercial ones, so they can afford to lose a few to the sea). Additionally wilderness response could use them for the same thing, we've discussed the possiblity of having an imagery platform that finds lost hikers and drops smoke signals near them for the helicopter rescue crew to find.
- And more, thats all i'm listing for now.
We're looking to have this system in place by the summer 2011 competition. Lets go for another first place guys!
- Flight Computers
- TweedleDee - onboard computer
- Leviathan - onboard computer
- AntennaRF - discussion of antennas relating directly to the flight computer wifi link
- WebsterTruthLocations? - how to correct imagery for hourly GPS drift