“…However, when is non‐minimum phase, this inversion is not possible as the learning filter then becomes unstable; in this case, there are methods in the literature that implement stable inversion approaches to build the learning function (see [33–35]). A stable inversion is necessary due to the fact that unstable filters can produce control actions that grow exponentially over time where, even over a finite duration, can become undesirably large (which can excite certain non‐linearities and even damage system components [36]). In either case, a model is required to build a learning function that best approximates the inverse of the plant dynamics.…”