Add PID object PD limit

[lthall]

This feature lets the user limit the maximum PD output of the PID object. This is useful in large aircraft or systems where the actuators are slew limited.

[IamPete1]

The constrain is "invisible" because the pid_info object is not updated, this will make log review very difficult.

Its a little hard to say the best way to over come this. We could work out the ratio and decrease both P and D as logged. EG:

if (is_positive(_kpdmax)) {
    const float PD_scale = _kpdmax / MAX(fabsf(P_out + D_out), _kpdmax);
    P_out *= PD_scale;
    D_out *= PD_scale;
}

However doing it that way may still not make it clear as P and D will not "flat line" individually.

I think we would need that and a new limit flag. We can do a limit bitmask and in the existing uint8_t limit field so the log message size will not actually get bigger, although the pid_info struct will.

[xianglunkai]

@lthall
I doubt the practical value of this change because based on my experience in autonomous driving, there are position, speed, and even acceleration limitations on the controller output, which is sufficient. Of course, the integration of individual saturation constraints is necessary. Perhaps my understanding is biased, thank you!

[lthall]

I think we would need that and a new limit flag. We can do a limit bitmask and in the existing uint8_t limit field so the log message size will not actually get bigger, although the pid_info struct will.

That sounds like the best option to me.......

@lthall I doubt the practical value of this change...

@xianglunkai this is not something that the majority of vehicles will use however it is useful for very particular problems. One of these problems is caused by slew limited actuators. Slew limited actuators have a small delay for small output changes and much larger delays for larger changes. Basically the frequency response changes with the magnitude of the signal. By adding a PD limit you can ensure the system doesn't move to the lower latency and unstable region.

[bnsgeyer]

LGTM

[tridge]

@lthall I've force pushed the change we discussed to remove the constructor changes

[lthall]

Thanks Pete!!! I like it.

[lthall]

Mark as advanced