Advantage of anti-windup

Question

What is the definition of anti-windup? How does it impose the constraints?

What are the advantages of MPC and anti-windup over each other?

Does anti-windup guarantee the constraints or does it just try respecting them?

Answer 1

Anti-windup is a concept for feedback controllers with integral terms, e.g. PID, to keep the integral term from „overcharging“ when regulating a large set point error. It basically saturates the integral term to keep the system from overshooting the set point.

The classic form of anti-windup, as described above, does not actually ensure satisfaction of input or state constraints of the system, it just enforces a somewhat „auxiliary constraint“ on the integral term.

EDIT: Anti-windup can also be used to ensure satisfaction of state and input constraints by exploiting the unwanted mechanic:

The classic saturation circuit gets extended by a predictor, that predicts the state that corresponds to the control input. A possible violation of constraints can then be detected with saturation. In that case a feasible control input, i.e. a control input such that the resulting state will not violate constraints, is calculated. The details of the approach are explained in this paper.

I am not familiar enough with the above approach to directly compare Anti-windup constraint satisfaction and MPC constraint satisfaction. I read once in some lecture slides about MPC (I'll link to them when I find them again)

If PID does the job, take PID, otherwise take MPC

so now that you know how to ensure constraint satisfaction with the former, just take the one that does the job.