Using Gyros and Accelerometers for attitude detection

[andyg]

I know that gyros and accelerometers can measure tilt in a stationary application but how can you use them in planes where a cordinated turn would produce a false reading? I've searched everywhere but all I can find is a bunch of really complicated answers. Could someone please maybe define a system like this in Layman's terms? Thanks


Andy Goessling
andy@riddlehill.net

[JettPilot]

Yes, they can provide false readings if in a coordinated turn long enough. Depending on how sophisticated the airplane is there are different ways of solving this... The airplane I fly requires a full 10 minutes to get aligned on the ground before the plane is moved at all. . A lot of electronics and computers keep them true in flight, more is involved there than I there is room to write on this forum ;) If we shut these down in flight we never get them back :eek: On general aviation aircraft, the vacuum driven gyro has averaging valves on them.. If it senses gravity not being directly below the artifical horizon, it VERY SLOWLY starts to correct to gravity - With this system, a long side force will definately start to make the horizon drift if it present long enough..... Very simple but it works just fine for general aviation airplanes..

[andyg]

This works for fulscale aircraft a rc plane would definetly use a different kind of gyro. Thanks for the info. Anyone have some RC applications?

[JettPilot]

Heat sensing seems to be the way to go. The FMA ones are very cheap, work well, and are easy to set up and use, cant beat that :D . They work day and night over all types of terrain. If you must fly thorough clouds, I would incorprate the FMA system with a heading hold Gyro that would keep the plane level until it exited the the cloud. I never tried it, but it would be easy to do....

[sesat]

Andy, I presume you specifically want to know how this is done with inexpensive (<$1000) sensors. It is often done coupled with GPS too, which quickly provides rough estimates with which to do the inertial processing calculations.

If the outputs of these sensors are perfect, you could use gyro's alone to determine attitude (relative to some initial value corresponding to what we people consider level flight heading north) by continually integrating the output of the rate sensors. We could then use perfect accelerometers to determine body acceleration, because we know which way is "down" and north from the gyro's. gravity nonsense deleted We then integrate body acceleration twice to determine attitude CORRECTION inertial frame acceleration, not B-frame. use attitude from gyro to convert b-frame accel to vector relative to north/down . Again. If everything is perfect the difference between the two values will be the change in attitude from the first calculation step to the last calculation step. Have your computer crunch the data so fast that we can assume this change to be null.

We don't have perfect sensors, and the difference between the values depends on the errors of the outputs of each sensor. The differential of the GPS velocity data output gives you another estimate to work with. A statistical tool called a Kalman filter keeps track of the estimated errors of all these sensors, the estimated real values the sensors are ‘supposed’ to sense, and the estimated real values of the vehicle dynamics.

Initial estimates become increasingly critical with cheaper/dirtier sensors; hence, the initial estimates of the spectral densities of various noises need to be well-tuned in order for the filter to converge to the true value.

Ram.

edit important corrections in bold!