A strategy to tune a PID is to start with 0 I, 0 D, and add P until it becomes
unstable (then do other steps, but this is not relevant now). So I initially
set my values to that. The procedure was: - change value in systems.cfg
directly within PackageSources. - rebuild package - Test No value in the
proportional control affected the result. According to the AP debugger, there
was still an I component. Are these reloaded by triggering a package rebuild?
Are they used at all?
Where did this non-traditional idea of tuning a pid, by 1st setting a value of
Integral come from ? The SDK ? I suggest you try: To tune your PID
controller manually, first the integral and derivative gains are set to zero.
Increase the proportional gain until you observe oscillation in the output.
Your proportional gain should then be set to roughly half this value. THEN,
you can add in I & D to modify the response for faster settling etc Ref: Do
your own research if you are unfamiliar with Tuning Pids. Google search -
How to tune a
PID
@N6722C , indeed, if you read again my post, that is what i am doing. I
however did not get to the part of increasing P until oscillation is observed,
as no matter the value, this would not happen. The AP debug window still
showed an I component being output, as if my entries were ignored. This is the
jist of my question. Are these PID parameters even being read?
As well as details on setting the AI-Pilot Pids (ie In case of AI Pilot
issues) Additional modes were added to the AP Pids, to account for initial
condition when the AP is turned on/off. Should there not also be similar mode
options for the AI-Pilot Pids ?
It will be great to know which parameters work, and which do not. So far, I
can tune via AI.cfg the main pitch and roll PIDs, but anything like heading,
VS, GS, does not seem to respond to even the old FSX legacy parameters.
Would be really good to get answers to this. The SDK details many entries in the systems.cfg file to tweak the autopilot, but apparently none are implemented? Will any work if added to the ai.cfg stick and rudder section? For four years I am sure many developers spent countless and completely futile hours trying to tweak the systems autopilot entries.
Isn’t it kind of weird the config for the ai version of an aircraft would control the piloted version of the aircraft?
Is that just to reduce the instances of the definition. If the ai.cfg is controlling what the ai co-pilot does, I understand, as, well, I imagine the ai-copilot is the same controller of the aircraft as the controller for ai aircraft. Either way it’s confusing if there’s one version of for all (ai and live, one version eliminates the possibility of having different versions for each case), or managing one set of data in multiple locations.
Sometimes having an option (in this case, an aircraft version and an ai version) can just make things more confusing and not add much utility.