As described in the source, there are several ways to control a nonlinear system with a linear PID controller. If no fixed operating point is considered, advanced methods like gain-scheduling should be used.
But what if only one classical PID controller is available? Then a compromise between all possible operating points has to be found. But how to proceed? Do you always use the operating point with the largest static gain as a basis or is there a better way? Is there any literature that deals with this question?