Intelligent Self-Interference Mitigation for Integrated Radios

With the evolution of increasingly higher levels of integration in modern SoCs, combined with the insatiable demand for more data/information across a channel, the challenges of self-interference are appearing in a plethora of diverse applications from wireless and wireline transceivers, to biomedical systems (neural interfaces, imaging front-ends, etc) and quantum electronics. Increased spectral efficiency motivates the use of radios that simultaneously transmit and receive using the same frequency band — this is commonly referred to as In-Band Full Duplex (FD) communication. The strong transmitter self-interference in FD radios places extreme performance demands on the corresponding receiver with respect to linearity, noise figure degradation, reciprocal mixing, and as high a cancellation depth and bandwidth as possible. Moreover, while there have been several successful research efforts that demonstrate the feasibility of integrated self-interference cancellation techniques, these solutions require a high degree of tunability using complex and lengthy calibration algorithms, which further challenge the practical use of FD radios for commercial wireless systems. This presentation will review the challenges surrounding self-interference in a diverse set of applications from biomedical to wireless communication and radar sensing. This is followed by a survey of the last eight years of our laboratory research on integrated radio front-ends for FD wireless systems. Our more recent work on the use of convolutional neural networks (CNNs) for rapid adaptation of a full-duplex canceller will be highlighted toward the end of the presentation along with our thoughts for future research on highly adaptable integrated radio front-ends.