Compressive Initial Access and Beamforming Training for Millimeter-Wave Cellular Systems

Initial access (IA) is a fundamental procedure in cellular systems where user equipment (UE) detects base station (BS) and acquires synchronization. Due to the necessity of using antenna arrays for IA in millimeter-wave (mmW) systems, BS simultaneously performs beam training to acquire angular channel state information. The state-of-the-art directional IA (DIA) uses a set of narrow sounding beams in IA, where different beam pairs are sequentially measured, and the best candidate is determined. However, the directional beam training accuracy depends on scanning beam angular resolution, and consequently its improvement requires additional dedicated radio resources, access latency, and overhead. To remedy the problem of access latency and overhead in DIA, this paper proposes to use quasi-omni pseudorandom sounding beams for IA, and develops an algorithm for joint initial access and fine resolution initial beam training without requiring additional radio resources. It comprehensively models realistic timing and frequency synchronization errors encountered in IA. We provide the analysis of the proposed algorithm's miss detection rate under timing synchronization errors, and we further derive Cramér-Rao lower bound of angular estimation under frequency offset, considering the 5G-NR compliant IA procedure. To accommodate the ever increasing bandwidth for beam training in standard evolution beyond 5G, we design the beam squint robust algorithm. For realistic performance evaluation under mmW channels, we use QuaDRiGa simulator with mmMAGIC model at 28 GHz to show that the proposed approach is advantageous to DIA. The proposed algorithm offers orders of magnitude access latency saving compared to DIA, when the same discovery, post training SNR, and overhead performance are targeted. This conclusion holds true in various propagation environments and three-dimensional locations of a mmW pico-cell with up to 140 m radius. Furthermore, our results demonstrate that the proposed beam squint robust algorithm is able to retain unaffected performance with increased beam training bandwidth.