PD-type Iterative Learning Control for Support-Vector-Machine-Based Digital Pre-distortion of Power Amplifier

Digital pre-distortion (DPD) is a commonly used approach for compensating the power amplifier(PA) nonlinearities. However, the accuracy of this scheme depends heavily on the establishment of the DPD model. To solve this problem, a new scheme is proposed, where the DPD model is built using a twin support vector regression (TSVR) machine, and the modeling accuracy is guaranteed by a new proposed proportion-differentiation iterative learning control (PDILC) algorithm for the training signals. The technique proposed here first uses PDILC to identify the optimal PA input signal that drives the PA to the desired linear output response. Once the optimal PA input signal is identified, the pre-distorter is estimated using TSVR. Finally, this pre-distorter model is cascaded with the PA to form a linear system, which achieves the linearization of the PA and obtains higher spectral utilization. Extensive experiments validate that the proposed PDILC-based TSVR-DPD outperforms many methods.