Different Methods for Loop Tuning in Robotics
PID loop tuning given the natural frequency of the parameter gains can be manipulated using different methods. Here are some of the most popular loop tuning methods that are tested by times and trusted by people all over the world.
Ziegler Nicholas Method is arguable the most popular and reliable method for tuning a PID loop. This method includes setting the D and I gains to zero value. The parameters which are monitored and manipulated in this method are Proportional Gain (Kp ), Ultimate Period (Pu) and Oscillation Period (Tu). The P, I and D gains are now set against oscillation period and ultimate gain so that desired output is achieved.This method can be used for both the open and closed loop systems.
The basic formula used in the Ziegler Nicholas method for tuning is : Kc = 0.45Ku and Tu = (Pu/1.2)
Cohen-Coon Method is suitable only for open loop systems and it can only be used for time delay models that belong to t he first order class. It corrects the steady-state response given by the Ziegler Nicholas method. Cohen Coon method is an offline method, whereas Ziegler Nicholas method is an online method. For every tuning cycle, a steady state must be achieved because in offline methods steady state only leads towards step change in the input.The corresponding image shows parameters used in the Cohen Coon method.
Automatic Tuning based on relay feedback is another method that is an alternative to the conventional continuous cycling technique. This method is also known as Autotune Variation (AV) Method. It is performed for closed-loop systems and it is efficient for long time constant processes as compared to the conventional hit and trial or step methods.
Advancement in technology has made it possible to perform loop tuning and visualize the changes taking place on screen. LabView is one such tool that helps in monitoring and optimizing control systems using graphical flowcharts. Another tool is Robust Control Toolbox from Mathlabs. This tool minimizes overshoot and keeps a check on the steady state errors. It also offers approximation algorithms for model order reduction.