Learning to fly - the teachin approach and the Kohonen call
Hi all,
well, from my old-fashioned opinion there is currently too much discussion about the best way to do AHRS in this forum. Kalman versus other approaches like Bill Premerlani's one. Remember that the whole DIYDRONE story started with the Paparazzi project in Europe (long before Arduino and the DIYDRONE page) and remember that as early as the Paparazzi project the problem was solved by taking advantage of absolute referencing by means of IR sensors. The US company who did introduce this technology is still alive because it takes care about the "ordinary" RC pilot (like me). From a very pragmatic point of view I do understand the discussions, because I also like to put things into the fuselage instead of sticking/patching sensors to the outer skin. But the only reason for this preference is that I don't want to kill performance of a streamlined plane optimized to win competitions in whatever F3x. SO: As soon as semiconductor industries find a market for micro IR sensors the problem might be GONE.
What I would like to have in my plane is a gadget that learns my flying style should I fail in a critical situation. Just imagine that you want to achieve two things: (1) That your carbon-kevlar plane will not crash into a group of people if YOU loose control and (2) that your plane rescues itself from crashing at all. The first steps towards these goals seems to me that one needs to start thinking about teaching your plane to act like yourself would act is such a situation. The plane is a robot that continues to learn from you while yourself is in control. In a critical situation it is waypoint programmed to return to a homepoint. But the way it does this maneuver has been learned from you. Like other maneuvers.
Anybody interested in a project on evaluating "teach in" algorithms?
Natalius
Replies to This Discussion
-
DeveloperPermalink Reply by Mark Colwell on -
I would be interested but, this will need a powerful processor or two.
-
Permalink Reply by Natalius Kiedro on -
Well, I'm not really convinced about your statement. Things can be shared between ground and air if transmission of data is no strong bottleneck anymore (e.g. by Kixlines). A collegue (maybefriend) of mine has published a collection of articles on a relating issue in a book: Flying Insects and Robots: Dario Floreano, Jean-Christophe Zufferey, Mandyam V. Srinivasam, Charlie Ellington (Eds.), Springer, Berlin, Heidelberg, 2009. I believe in the KISS-principle (Keep it simple and stupid). So what we as hobbyists/amateurs can do is to think about how to make it available to everybody. I'm not sure whether Teuvo's approach is the best what we can do in the moment. Maybe that more classical approaches (such as Simplex) are more appropriate for this kind of problem. In any case I would love to have something which allows me to teach my plane to learn my style of flying - so that switching from manual flight to autopilot is a smooth to almost natural transition.
Contests
Season Two of the Trust Time Trial (T3) Contest has now begun. The third round was a reliablilty/aerial photography round for both planes and copters, which is now closed. Stay tuned for the next round, beginning soon.A list of all T3 contests is here.
Top Content
Groups
-
South African DIY Drones…
19 members
-
DIYDrones - CNC Project
59 members
-
UK Ardupilot Group
28 members
-
ArduCopter (traditional …
114 members
-
ArduPilot (legacy) User …
94 members
© 2012 Created by Chris Anderson.
