Robust and Efficient Kernel-Based Digital Self-Interference Cancellation Using a Priori Knowledge in Full-Duplex Transceivers

Self-interference poses a significant challenge to in-band full-duplex (FD) wireless communications, particularly due to nonlinear distortions introduced by hardware impairments. This paper presents an enhanced digital self-interference cancellation technique based on a kernelized version of the adaptive projected subgradient method (APSM). A key contribution is the incorporation of prior knowledge of the system represented as convex sets. This integration improves the robustness and accuracy of interference cancellation and reduces the computational complexity at the same time. By leveraging parallel projections in the APSM, the proposed algorithm achieves fast adaptation to dynamic channel conditions in FD wireless communications.