UCLA

Future wireless applications continue the trend of increasingly demanding bandwidth, data rates, and connectivity, forcing systems to operate at higher frequencies to exploit the abundant bandwidth and with more antennas to exploit array gain and beamforming capabilities. However, the increasing array aperture and bandwidth start to challenge the established channel and array response assumptions, and these changing assumptions require both new array architectures and new algorithms suited to the changing channel characteristics. First, the frequency dependency of the array can no longer be ignored, requiring true-time-delay (TTD)-based array architectures to enable low-cost frequency-dependent control. Furthermore, as the array aperture grows, more users fall into the near-field region, complicating the beamforming design, since the user channel now depends on both angle and distance. This also complicates localization and beam training, as the search must now be performed over both distance and angle. TTD arrays, originally proposed to overcome the frequency dependency caused by the fixed antenna spacing, have also been utilized to realize split beampatterns which map subbands to different spatial regions, allowing the serving of multiple users at different angles with only a single RF chain. However, current algorithms for split beampattern generation either require high computational complexity or memory, or cannot operate under frequency dependency, while heuristic models are restricted and reduce usability. In Chapter 2, we propose a fast and efficient split beampattern synthesis algorithm based on the mathematical structure of the split-beam synthesis problem, which requires only a fixed generator beampattern dictionary, resulting in a low-cost alternative that also provides fairness. However, the parameters of the proposed algorithm are not optimized for the average received power. In Chapter 3, we propose a data-driven parameter optimization approach for the proposed algorithm to maximize the average received power, and show that the optimized algorithm matches a low-complexity state-of-the-art benchmark algorithm while providing a fairer power distribution among subbands. Likewise, the near-field beam training problem is addressed in Chapter 4, where we show that, by utilizing TTD arrays and a special configuration called rainbow beams, we can virtually partition the array into far-field-operating sub-arrays, recover each sub-array’s signal in post-processing, determine each sub-array’s angle, and localize the user with triangulation. Therefore, this thesis provides strong solutions to the rising challenges of next-generation wireless systems by addressing both of the changing assumptions.