Efficient Hardware Architectures for High-Throughput Streaming Data RF and AI-Driven Communication Systems
Loading...
Author(s)
Mao, Xiangyu
Advisor(s)
Editor(s)
Collections
Supplementary to:
Permanent Link
Abstract
This thesis investigates novel hardware architectures to address the growing computational demands of next-generation wireless communication and AI systems. With the advent of 5G, the emergence of 6G, and the rapid expansion of AI-driven applications, modern RF systems must process increasingly complex data streams in real time while maintaining low latency, high throughput, and energy efficiency. However, conventional computing architectures, particularly those based on the traditional Von Neumann model, face fundamental limitations in data movement, memory bandwidth, and scalability, motivating the need for alternative computing paradigms. To tackle these challenges, this dissertation proposes a series of hardware innovations across three major domains: RF system emulation, low-precision beamforming, and in-memory computing for automotive radar systems. First, an FPGA-based real-time RF radar emulator is introduced using a Direct Path Computing Model, which reduces computational complexity from O(M^3) to O(M^2) and enables efficient, high-throughput waveform simulation across diverse RF environments. Second, an ultra-low bit precision (2-bit) digital Linear Embedded Beamforming system is presented, incorporating a novel quantization compensation technique to significantly reduce power consumption and resource utilization without sacrificing accuracy. Finally, a Processing-in-Memory-based hardware architecture is developed for AI-enhanced automotive radar, integrating a configurable in-memory computing engine with an all-to-all BENES on-chip network to enable scalable, energy-efficient, and low-latency processing of radar data streams. Together, these contributions establish a unified hardware-accelerated framework for intelligent RF communication in the era of AI and big data. The proposed architectures developed in this work advance the state-of-the-art in real-time RF emulation, energy-efficient beamforming, and memory-centric AI computation, paving the way for intelligent, high-performance wireless systems capable of meeting the demands of 6G and beyond.
Sponsor
Date
2025-04-28
Extent
Resource Type
Text
Resource Subtype
Dissertation