Data-Driven Hybrid Beamforming for Uplink Multi-User MIMO in Mobile Millimeter-Wave Systems

To enable user diversity while balancing tradeoffs between cost and flexibility, we exploit hybrid analog-digital beamforming for multi-user mobile systems. By combining array and spatial signal-processing techniques, highly directional beams can be formed with high beamforming gain to achieve sufficient link budget. In this way, fine selection of codewords for analog beamforming is essential to ensure the uplink rate, which may increase the latency of establishing a reliable communication link, especially for mobile millimeter-wave communication systems. In order to guarantee the reliability of communication, adaptive data-driven beam tracking is proposed to find the suitable beamformer/combiner pair to achieve the given signal-to-interference-plus-noise ratios constraint. Unlike the model-based approach, the proposed approach is dependent only on the real-time measurement data based on a dynamic linearization representation of a time-varying pseudo-gradient parameter estimation procedure. Numerical analyses show that the proposed beam tracking algorithm can achieve good tracking performance with lower training overhead compared with traditional schemes.