A Model Free Automatic Tuning Method for a Restricted Structured Controller by Using Simultaneous Perturbation Stochastic Approximation (SPSA)

摘要

A model free auto tuning algorithm is developed by using Simultaneous Perturbation Stochastic Approximation (SPSA). For such a method, plant models are not required. A set of closed loop experiments are conducted to generate data for an online optimization procedure. The optimum of the parameters of the restricted structured controllers will be found via SPSA algorithm. Compared to the conventional gradient approximation methods, SPSA only needs the small number of measurement of the cost function. It will be beneficial to application with high dimensional parameters. In the paper, a cost function is formulated to directly reflect the control performances widely used in industry, like overshoot, settling time and integral of absolute error. Therefore, the proposed auto tuning method will naturally lead to the desired closed loop performance. A case study of auto tuning of spool position control in a twin spool two stage valve is conducted. Both simulation and experimental study in TI C2000 target demonstrate effectiveness of the algorithm.