just a correction. to stabilize a system you don't necessarily need to know how it work. Good engineer (in france we call that domain "Automatique". It is the guys who can stabilize a building heating, a rocket, a servo, an hybrid car engine... an old branch of cybernetic) know how to extract key data from the behavior of the system....
after observing the behavior of the system after some changes and perturbation, and if possible some modelization typically the first things is to guess the number of captors and actioners needed to control the system. you should also guess/measure the incompressible delay that you cannot absorb... then you can modelize (phenomenologically) the system, decide a target of control (should it be, stable, fast, economic, simple, robust or fine...). then you can compute the optimal controller... you can also make an adaptive version of that controller that guess the key parameter all along, and keep nearly optimal despite changes, and non linearity or slow changes. 2011/12/24 Terry Blanton <hohlr...@gmail.com> > You have to > understand the reaction to understand what makes it unstable.