Neural network-defined physical layer: a new paradigm for software radio in the IoT era

The increase of the Internet of Things (IoT) has created a complex and heterogeneous wireless ecosystem, demanding IoT gateways that are both flexible and efficient. While Software Defined Radio (SDR) provides the necessary hardware adaptability, its potential is frequently undermined by significant software implementation challenges, including a lack of portability across platforms, prohibitive design complexity for advanced algorithms, and poor computational efficiency. This thesis posits that these persistent bottlenecks can be overcome by a paradigm shift in physical layer (PHY) design: reframing core communication functionalities as learnable, interpretable neural network (NN) models.

Our proposed NN-based framework addresses the dual problems of design and implementation. It automates the complex, expert-driven process of parameter optimization through data-driven learning, and it achieves inherent portability and efficiency by leveraging the mature ecosystem of deep learning libraries and hardware accelerators. The power and versatility of this paradigm are demonstrated through the design, implementation, and extensive experimental validation of three novel systems, each targeting a fundamental PHY-layer challenge: an NN-defined modulator that achieves high portability for signal generation across heterogeneous hardware; an NN-based Polyphase Filter Bank (NNPFB) that automates the design of complicated filters for efficient signal multiplexing; and an NN-based Cross-Technology Communication (NNCTC) framework that enables adaptive and general-purpose interoperability between disparate wireless protocols.

The results consistently show that our NN-based systems meet or exceed the performance of traditional, hand-crafted methods while offering superior flexibility, portability, and adaptability. This work establishes the neural network-defined physical layer as a practicable and potent new direction for designing the next generation of intelligent wireless communication systems.

Jiazhao is currently a PhD student in ISTD of SUTD, under the supervision of Prof Chen Binbin and Dr Jiang Wenchao. His research interests include wireless communication and IoT networks, focusing on the physical layer design. Prior to joining SUTD, he received a BEng degree in Electronic Information Engineering from the University of Electronic Science and Technology of China (UESTC) in 2019, and an MSc degree in Electrical and Computer Engineering from the National University of Singapore (NUS) in 2020. During his PhD studies, he also joined the Shanghai Wireless Group of Microsoft Research Asia as a research intern.