Title:  Transmission Performance Optimization in Fiber-wireless Access Networks using Machine Learning Techniques

Committee: 

Dr. Bloch, Advisor

Dr. Ma, Chair

Dr. Anderson

Abstract: The objective of the proposed research is to experimentally investigate and validate machine learning techniques for transmission performance optimization in radio-over-fiber (RoF) based fiber-wireless networks (FiWiN). The underlying technologies of RoF-FiWiN consist of digital RoF (D-RoF) and analog RoF (A-RoF), according to the nature of the radio signal transmitted within the fiber connection between the central office (CO) and the base station (BS). The proposal aims to alleviate the capacity crunch of the D-RoF using advanced modulations and nonlinear signal recovery/pre-distortion schemes enabled by deep neural networks (DNN). Moreover, the proposed research focuses on mitigating the complex interference in the post-5G radio access networks (RAN) and evaluating the corresponding performance based on a millimeter-wave RoF (mmWave-RoF) experimental platform. More specifically, a proactive interference avoidance scheme using reinforcement learning and an effective method for simultaneous self-interference (SI) cancellation and signal-of-interest (SOI) recovery using a dual-inputs DNN (DI-DNN) are experimentally demonstrated.