AdaPilot is a new project to create a highly reliable open source autopilot using the Ada language. From the Project Overview:
AdaPilot, started in 2016, is a next-generation Open-Source software project with a reference hardware development kit, created by the AdaPilot Community. It has been designed from the ground up by a community of passionate Ada and ARM developers from around the globe, with its core design principles being quality, safety, and ease of use. It is aimed at providing a DO-178C certifiable safety critical flight control system composed of a single or redundant flight control computer with basic autopilot features, an attitude & heading reference system, a modular & customizable mission controller, a secure data-link, a electronic speed controller, a handheld ground control station and a mission management & planning software. The community can help to contribute to the AdaPilot code written in Ada-SPARK following the Core-Team advices and approvals. The reference hardware development kit forms the base for the AdaPilot environment, and through the use of the CANaerospace aviation bus protocol, the flexibility and integrability with third-party systems are only limited by the imagination and skills of the developers.
The current code base is targeting multirotors up to 8 rotors. However, the platform is flexible and can be adapted to fixed wing or single rotor, as well as for land-based and maritime applications.
Comments
new generation? who will learn ada and compile it on a stm32.. . i don't need to learn by mistake to crash more than current solution. why not make VHDL and work on a old fashion lattice fpga.
Hopefully one Monday morning we can link the Ada modules to ArduPilot and fly certified and without any issues in the afternoon.
This adapilot is an European FP7 subsidized project. Like most of these EU subsidized projects, this will fall into oblivion once the international consortium will have consumed all of the subsidies. Especially when the underlaying topic is nothing new and late versus what already exists, like Ardupilot.
As all of these EU FP7 projects, big companies won the bid and are in the driving seat. They have the nerve on the web page project's description to explain they'll produce expensive products and services, not targetting the DIY/RC market, because they will be "professionnal" level solutions. Come on, what a bunch of rubbish.
The first flight controller is called 'AdaRacer'. Is that targeted at the 250 size FPV racer. I'm sure that class of drone is statically more likely to crash before any MTBF of flight stack, parts etc... ;-)
http://adapilot.jappster.de/forum/index.php/adaracer?blog=9
(it's apparently based on PixRacer OSHW design)
The other aspect of creating a high reliability is the hardware that it runs on. The Hercules line of processors is designed for safety critical applications. I think NutX has been ported to one of the processors this line.
https://en.wikipedia.org/wiki/Hercules_%28processors%29
These "Safety" processors have 1 bit ECC and 2 bit error detection on all memory and on chip I/O registers. Processor results are checked after every clock cycle with errors being flagged at a hardware interrupt level.
Yeesh. I studied that in college. *shudder*
wait first flight before to evaluate the quality of solution ... there are some video about it ?
I would say one of the biggest factors in making a robust system is to limit the autopilot to a core of basic functionality that just work. Something open-source by nature has a tendency to not be good at (not that proprietary is any better for that matter). Once you have a fully functional core that 'just works', you can then start looking into doing fancy statistical stuff to mathematically prove it's robustness.
But this is very much a chicken and egg problem, since autopilots are still far from being a solved problem. Real world physics, radio signal conditions, electronic and mechanical problems will still continue to behave in unpredictable ways, regardless of how mathematically robust the source code is.
Also the very nature of this would mean that feature updates would be few and far between. Something that while attractive to the professional marked, tend to limit the larger community appeal.
There is also the http://smaccmpilot.org high assurance autopilot for those interested