UCLA

Millimeter-wave frequencies offer abundant spectrum and high data rates, but their use comes at the cost of a severe propagation loss. The loss can be alleviated by establishing directional communication links with high beamforming gains between the base stations and users and by using dense network deployments.

Beam training is a procedure that reveals the best beam steering directions in a wireless channel. When performed with conventional phased arrays that have many antenna elements, beam training imposes a large overhead. In this work, we introduce true-time-delay array architectures as promising alternatives for phased arrays to solve the overhead problem. Compared to phased arrays, true-time-delay arrays can synthesize frequency-dependent beams and thus probe all angular directions simultaneously using different signal frequencies. We leverage this property to develop and analyze low-complexity digital signal processing algorithms for fast and accurate millimeter-wave beam training.

Unlike beam training, channel estimation has the goal to estimate all parameters of a sparse millimeter-wave channel. We exploit the channel sparsity and frequency-dependent beams of true-time-delay arrays to develop a frequency-domain compressive sensing based algorithm for channel estimation. We also analyze the performance of the developed algorithm in the presence of practical hardware impairments and we derive the lower bounds on the variances of channel parameter estimators.

In dense millimeter-wave networks with a small inter-site distance and a large number of users, directional beams can cause significant interference and prevent data-hungry users from satisfying their rate requirements. The user experience and the overall network performance can be optimized through coordinated user association and beam scheduling on a network level. Given the channel estimates between different pairs of base stations and users, we develop and analyze a new multi-step optimization framework for joint user association and beam scheduling. The main goal of the framework is to maximize the number of users with satisfied rate requirements while simultaneously suppressing the inter- and intra-cell interference. Since the framework includes NP-hard optimization problems, we propose an algorithm that attains a sub-optimal solution in polynomial-time.