Optimal input filters for iterative learning control systems with additive noises, random delays, and data dropouts in both channels

Abstract

The convergence performance of wireless networked iterative learning control systems is affected by additive noises, random delays, and data dropouts in both sensor-to-controller and controller-to-actuator channels. In order to guarantee the convergence performance of such systems, an input filter is designed in front of actuators to estimate the controller updated input under the effect of those uncertainties. Specifically, a P-type learning controller is considered firstly, and then a model is developed to describe the transmission of sensor measured output data and controller updated input data under the mixed effect of those uncertainties. On the basis of state augmentation, the two developed data transmission processes are further combined with the controller learning process to build a unified filtering model. According to this filtering model and the projection theory, an optimal filter in linear minimum variance sense is designed at the actuator side to estimate the controller updated input in iteration domain.The convergence performance of the norm of filtering error covariance matrix is proved theoretically, which means driven by this accurately estimated input, the convergence performance of networked systems adopting the P-type learning controller is guaranteed. Finally, some numerical results are given to illustrate the effectiveness of the proposed method.

Optimal state filters for networked iterative learning control systems with data losses and noises

“…It is evident that the convergence of networked ILC systems relies on the input not only at the actuator side but also at the controller side. In References 33, 37, and 38, we have successfully achieved the convergence of input at the actuator side. In the upcoming section, an optimal state filter will be developed at the controller side to indirectly obtain the actual output, thereby improving the convergence of the controller‐transmitted input.…”

“…In fact, we have proposed filtering methods to enhance the convergence of input received by actuators under the impact of various communication disturbances. 37,38 As a further work along the research direction, this article aims to enhance the convergence of input transmitted by ILC controllers under the impact of data losses and additive noises in both channels. First, we consider a class of linear discrete-time invariant objects controlled by proportional ILC controllers, the learning gain of which is pre-selected to ensure the object convergence without communication interference.…”

“…Once these two types of input converge, networked ILC systems can achieve satisfactory tracking performance. In fact, we have proposed filtering methods to enhance the convergence of input received by actuators under the impact of various communication disturbances 37,38 …”