Controller Tuning is an Art

Controller tuning is an art. Obviously, there are ways to do this that can slash the time it takes significantly, but if you don’t understand the math behind it – really, all you’re doing is turning some knobs and seeing how a signal responds to it. You’re trying to match the signal you’re tuning to the input signal, which is constant, by iterating over different values of your variables in your controller.

Let’s go back to standard tuning methods and why they don’t (always) work. Each controller that you will deal with is unique, and will control different parameters for different systems. These parameters are changed by control engineers, who (by means I know not) have developed an intuition for how different tuning parameters affect a control loop’s behaviour. System not stable enough? Maybe increase the proportional gain. Not responsive enough? Try the proportional gain and the integral time. A good control engineer would be able to look at a system, see how it is vibrating or moving, and tell you if anything is wrong with it.

Why am I writing about this? Perhaps because I spent an entire day tuning and retuning a controller in SIMULINK and realising that I did not know what any variable in my controller did. I won’t go into the technical details, as somehow I’ve found variables that make my controller work (while, at the same time, having no idea how any of it works), but the math behind it is still a mystery to me. I just want to shed light onto what control engineers do and how they’ve managed to make a profession out of pure black magic shenanigans. Controller tuning is an art – and unfortunately, I’m just not a very good artist yet.