Remember the cool Rockwell Collins video from last year, the one in which they blow most of the right wing off a scale F-18 model in flight, then land the unmanned aircraft safely? Well, the company has just posted a new version of that video, in which it shows what happened when its damage-tolerant flight controls were turned off.
Video: Rockwell Collins
Collins has just got a contract for the next phase of DARPA's damage tolerance program in which it will demonstrate automonous recovery and safe landing after "more realistic damage" - which, says the company's controls guru David Vos, means blowing off "a big fraction of the wing, horizontal tail and vertical tail". That should be interesting to watch...
I don't pretend to understand completely how it works - after all Vos is the man who brought us the autonomous unicycle, and control theory is way over my head - but basically there is outer-loop "automatic supervisory adaptive control" that returns the aircraft to controlled flight in about a second by using the whole vehicle as an actuator, controlling its orientation to generate the required forces and moments. Then there is inner-loop "model reference adaptive control," which recovers baseline performance over about a minute by comparing the damaged aircraft's responses with a stored model and increasing the control gains until the two match.
Especially interesting when some of the parallel sub-system survivability efforts are also considered, e.g. SECAD. Given current UAV loss rates and the cost of platforms such as X-47B (which may limit their use to dull & dirty only...), success with efforts such as this is a must.
That is so trippy cool that my mind just can't wrap itself around a major control/lift surface being ejected and then the craft regaining control. Camera tricks, I tell you ;)
Also, I want this technology adapted immediately to commercial aviation in the likely case I am not on a flight with Capt. Sullenberger at the controls.
franco12, I do not think that the implemented flight algorithm was specific with that damage. I assume more that they have multiple algorithms for damages like wing, aleiron, tail. Would be nice to know if those could be mixed... For sure implementing them is a lot of code and mostly testing. Kind of multiple airplanes...
franco12 and ghemago - my understanding is it doesn't really matter what the damage is, the system is driving the damaged aircraft's response as close as possible to the reference model stored on board, using the surviving control surfaces.
There must be a limit to the damage than can be overcome, or the performance that can be recovered, but whether it is clean or "dirty" damage appears less of an issue.
Also, where previous damage tolerance control systems used a look-up table - ie, aileron lost = use stabilator, etc - the system does this in real time by measuring the aircraft's actual responses to control inputs and learning what is still working and what is not...real fast!
Also, I want this technology adapted immediately to commercial aviation in the likely case I am not on a flight with Capt. Sullenberger at the controls.
http://www.youtube.com/watch?v=-_EXtBEaBbs
;-)
Would be nice to know if those could be mixed...
For sure implementing them is a lot of code and mostly testing. Kind of multiple airplanes...
There must be a limit to the damage than can be overcome, or the performance that can be recovered, but whether it is clean or "dirty" damage appears less of an issue.
Also, where previous damage tolerance control systems used a look-up table - ie, aileron lost = use stabilator, etc - the system does this in real time by measuring the aircraft's actual responses to control inputs and learning what is still working and what is not...real fast!