What is your background?
I obtained an engineer diploma with major in microelectronics in 1998 and then worked 10 years as software developer for Air Traffic Control while teaching computer science classes at the French National University of Civil Aviation ( ENAC ).
In 1985, I designed my first expansion board - for a Gumstix Verdex motherboard in 2005. My board, named Classix ( http://paparazzi.enac.fr/wiki/Classix), featured two LPC2148 and a Verdex socket. At the time, there wasn't an extensive offer for affordable COM module featuring Linux support andthe Verdex appeared like a natural choice.
The goal was to offer ample processing power ( the previous board was based on a pairof atmega http://www.nongnu.org/paparazzi/gallery_v1_2.html) for playing with Kalman filtering and predictive control, which are both CPU hungry fields. The road Paparazzi followed at the time went more along using simple and optimized algorithms and thus the Classix never got very popular. The Tiny (http://paparazzi.enac.fr/wiki/Tiny), featuringa single LPC2148 and an integrated GPS receiver became the most usedPaparazzi autopilot.
In the recent years, the maturity of the Paparazzi project and of micro UAV flight control in general, brought back the need for higher processing power and the ease of development brought by an operating system.
Here's a video of one prototype http://www.youtube.com/watch?v=LWTi2Dgh0Fw
In the same time, I've always found that coding "hard" real time algorithms like control loops or peripheral drivers was easier on an OS-less and "simple" processor. In the light of this, I designed Lisa/L as a two processors board, featuring a 72Mhz CortexM3 for the real time/low level part and a Gumstix Overo for the high level/number crunching stuff. Of course the board can be operated without the Overo and runs the current Paparazzi code base in that configuration. I chosed SPI for the inter MCU link in order to have a high bandwith with alow CPU load ( there are DMA on both ends ).
The dual MCU configuration is also interesting in term of safety, as thecritical ( and often simple) code needed to maintain flight can be segregated to the "real time" MCU rendering the vehicle immune to a fault in the "complicated" code running on the Overo.
Which Gumstix products are you using in the project?
As mentioned above, a Gumstix Overo COM.
Why is Gumstix technology the best solution for the project?
Mostly becauseof the open nature and wide community support of Gumstix. I love the fact that they are using Open Embedded. Between the Classix and Lisa/L, Iplayed a little with a LPC3180 board from phytec but quickly gave up after realizing I was on my own. Phillips and Phytec boast Linux support, but what they really did was a one
shot port, not pushing any patch upstream, hence rendering their effort useless. The silicon may be good but it's doomed by the lack of software.
What projects do you have planned for the future?
I think.. I hope... that the current board will provide our project with a couple of years of serenity, hardware wise, and let us focus on software. After that... we'll see
Will gumstix technology be part of your future plans?
Probably so because of their "open" strategy.
