An adaptive PI controller for non-Gaussian stochastic systems

In this paper, a new algorithm for an adaptive Proportional-Integrator (PI)controller for nonlinear systems subjected to stochastic non-Gaussian disturbance is studied. The minimum entropy control is applied to decrease the closed-loop tracking error under an iterative learning control (ILC) basis. The key issue here is to divide the control horizon into a number of equally time-domain intervals called batches. Within each interval there are a fixed number of sample points. The design procedure is divided into two main algorithms, within each batch and between any two adjacent batches. D-type ILC laws are employed to tune the PI controller coefficients between two adjacent batches. However. within each batch, the PI coefficients are fixed. A sufficient condition has been established to guarantee the stability of the closed-loop system. An illustrated example of one-link manipulator with revolute joints actuated by a DC motor is included to demonstrate the use of control algorithm, and satisfactory results have been obtained.