Flight - Offload wind compensation from navigation PIDs

[digitalentity]

Given:

1. Main goal of PH PIDs is to counteract position drift.
2. Main source of drift is wind.
3. Wind is mostly constant in speed and direction with occasionally gusts

Copter can use its velocity estimator to estimate tilt angles required to compensate wind the same way automatic throttle tilt compensation works. When wind compensation is uninitialised or not yet locked on required tilt angles common PosR PID will do it's job, but after wind compensation initialises, PosR PID will be required to only compensate for gusts, GPS drift, miscalibrated accelerometer and other minor causes of position drift. This will speed up the response or PosR PID and improve poshold stability.

[digitalentity]

As of 6ec7e15ee6676b312576a4048716d1cab98d0b98 wind compensation works and does it's job rather well. Now copter is able to quickly break from position adjustment to actual position hold. Althoug wind compensation should probably be reset when not used for a while. Need to figure the condition to do the reset.

[digitalentity]

Windcomp behaves better than common PIDs when changing attitude. Common NAV PIDs are being reset if pilot is touching the sticks, so when poshold takes over it has to figure the wind from the start - leads to drift. Windcomp is not reset on stick input - wind is calculated in global frame, it does not change rapidly, so PIDs have to compensate only for change in wind strength, not for the wind itself. It behaves much like PID I-term, but windcomp accumulator stores only the controller output that leads to zero velocity.

Today's tests proved windcomp to be working fine - copter was able to handle 20mph wind AND do maneuvers without much position drift.

[digitalentity]

Sometimes this causes copter to behave strange, especially when wind is gone while adjusting position. As a temporary fix I'm going to lower limit of wind compensation and add a slow decay to zero,

[digitalentity]

Well, latest tests indicates that wind compensation does not work as expected. It actuallt tends to amplify compass mis-configuration. With poorly calibrated compass and wind-comensation disabled copter is able to maintain position, but with wind-compensation enabled it is going in circles with increasing speed.
Conclusion: Wind compensation should go.

[digitalentity]

Removed in 7d08c4311c920051c95cba544a03398410ec8950

[digitalentity]

I'll keep it opened but blocked for the moment. I have a hunch that windcomp might have failed due to some other reason than algorythmic flaw.

[digitalentity]

New PID controller with anti-windup works really great by itself.