Hello all,
I have read some paper about quadrotor. almost of them state that it is nearly impossible to control the quadrotor using a PID controller because of the non-linearity, cross coupling and big disturbances (wind) (See equations). So if it is like this why we guys do not try to use some kind of more advanced and professional algorithm to control the ArduCopter. Controller which are more published are:
1- Optimal Control (LQR)
2- LQR with Gain Scheduling
3- Feedback Linearization
4- Sliding Mode Control
for all of the above algorithms one needs to identify the system parameters and to measure the speed of the four motors. this can be easily done by using a simple optical incremental encoders on each propeller.
What are exactly the reason which preventing DIYdrones community to do this?
Regards
Abusus
Replies
There is no need for an optical encoder. Sensing a motor phase is simpler and more reliable.
Nevertheless i'm not sure that a speed sensor is really needed for a better control. The thrust can be estimated quite easily from the motor / propeler thrust curve and mechanical inertia parameters.
The main problem is the added complexity, processor ressource, and reliability if the parameters are not set correctly. Measuring motors, propelers, and other needed parameters for a perfect control is out of access for most users.
PIPI or PID loops are simple enough to work in most situations without complicated settings and specific adaptation.
The project needs to stay as simple as possible so that "normal" users can jump in it.
Look up
get_stabilize_roll(long target_angle)
get_stabilize_pitch(long target_angle)
in attitude.pde.
You should be able to insert these equations there. Just replace the PIPI loops.
Jason