[Autopilot] Re: Fixed wing aircraft (Jake McGuire)
Status: Alpha
Brought to you by:
tramm
From: Carstens N <car...@su...> - 2004-09-25 23:47:57
|
Hi Jake. =20 Hope this does not come as old news... =20 Does this same problem also arise if you try to perform a coordinated turn in a helicopter with the current system, as opposed to "stop-yaw-start"? Yes, i would say. But you seem to understand the basic problem, so i = will not bore you with detail. But the long and short of it: = accelerometers measurements can only be used uncorrected to calculate = pitch and roll angles if there is no acceleration. The Rotomotion / = source forge autopilot EKF assumes the accelerations are small and of = short duration - well, in the old code. That is NOT the case during = coordinated turns. Old story: direction is changing all the time, = although you might be flying at a constant speed. What is worse is that = it is very difficult to build a good IMU without GPS or / and good rate = gyroscopes...=20 =20 So, the sensors being used to correct your estimates (accelerometers), = might give you bigger problems!! You are lying to your EKF by using = accelerometers to correct estimates during accelerated flight. If you = can find "better measurements" of your pitch and roll angles, use them. = Not the accelerometers...=20 =20 Sorry, talking too much. =20 Cheers, Nicol. |