Self-Interference Cancellation for Full-Duplex Underwater Acoustic Systems

This work develops and investigates self-interference (SI) cancellation (SIC) techniques for full-duplex (FD) underwater acoustic (UWA) systems. To enable the FD operation in UWA systems, a high level of SIC is required. The main approach used in this work is the digital cancellation based on adaptive filtering. A general structure of the digital canceller is proposed which addresses key factors affecting the SIC performance, including the power amplifier and pre-amplifier nonlinearities, up- and down-sampling effects. With the proposed structure, the SI can be effectively cancelled in time-invariant channels by classical recursive least-square (RLS) adaptive filters, e.g., the sliding-window RLS (SRLS), but the SIC performance degrades in time-varying channels. A new SRLS adaptive filter based on parabolic interpolation of the channel time variations is proposed, which improves the SIC performance at the expense of the high complexity. To reduce the complexity, while providing the high SIC, a new family of interpolating adaptive filters which combine the SRLS adaptive algorithm with Legendre polynomials (SRLS-L) is proposed. A sparse adaptive filter is further proposed to exploit the sparsity in the expansion coefficients of the Legendre polynomials. For interpolating adaptive filtering algorithms, the mean squared error is unsuitable for measuring the SIC performance due to the overfitting. Therefore, a new evaluation metric, SIC factor, is proposed. The SIC performance of the proposed adaptive filters is investigated and compared with that of the classical SRLS algorithm by simulation, water tank and lake experiments. Results indicate that the proposed adaptive filters significantly improve the SIC performance in time-varying scenarios, especially with high-order sparse SRLS-L adaptive filter. Furthermore, SIC schemes with multiple antennas are investigated to explore the possibility of achieving extra amount of SIC in acoustic domain and cancelling the fast-varying surface reflections by adaptive beamforming.