= Integral action
{title2=$\dot z=e$}
<Integral action> adds memory of the <tracking error> to a <controller>: $z(t)=z(0)+\int_0^t e(s)\,ds$. A persistent <tracking error> keeps changing $z$, so an <equilibrium point> with constant $z$ must have $e=0$. The accumulated value can remain nonzero after the <tracking error> has vanished, maintaining the <control input> needed to cancel a constant <control disturbance>. Convergence still requires a stable <feedback control> design. https://www.fbswiki.org/wiki/index.php/Introduction[Åström and Murray, Chapter 1: integral action]{ref}
Back to article page
