Cognitive Radio for Asymmetric Cellular Downlink with Multi-User MIMO

TL;DR

The paper addresses spectral efficiency for infrastructure-based cognitive radio in 5G downlink where active PU locations are unknown and both PBS and SBS employ multi-beam transmissions. It develops a two-phase offline-online algorithm: offline learns PBS-SBS and SBS-UE channels across UE locations and builds SINR tables, while online detects active PBS beams and selects SBS beams by solving a probabilistic interference constraint $Pr(SINR < \theta) < V$ while maximizing SBS beam usage. Validated with ray-tracing channel models, the approach demonstrates that careful directional beam decisions and power tuning can increase SBS throughput while keeping PU interference within acceptable limits. This provides a practical framework for realistic spectral efficiency gains in infrastructure-based CR deployments for 5G networks.

Abstract

Cognitive radio (CR) is an important technique for improving spectral efficiency, letting a secondary system operate in a wireless spectrum when the primary system does not make use of it. While it has been widely explored over the past 25 years, many common assumptions are not aligned with the realities of 5G networks. In this paper, we consider the CR problem for the following setup: (i) infrastructure-based systems, where downlink transmissions might occur to receivers whose positions are not, or not exactly, known; (ii) multi-beam antennas at both primary and secondary base stations. We formulate a detailed protocol to determine when secondary transmissions into different beam directions can interfere with primary users at potential locations and create probability-based interference rules. We then analyze the "catastrophic interference" probability and the "missed transmission opportunity" probability, as well as the achievable throughput, as a function of the transmit powers of the primary and secondary base stations and the sensing window of the secondary base station. Results can serve to more realistically assess the spectral efficiency gains in 5G infrastructure-based cognitive systems.

Figures (4)

• Figure 1: Illustration of primary and secondary beams with scattered PU and SU; inactive beams are shown with light shading.
• Figure 2: PMO Comparison for different $\sigma^2_{\rm SBS} \backslash \sigma^2_{\rm PBS}$ as $\circ$: -20 dB, $\square$: -5 dB, $\triangle$: 5 dB. MDBA and BDBA for 5 dB overlap.
• Figure 3: PCI Comparison, for different $\sigma^2_{\rm SBS} \backslash \sigma^2_{\rm PBS}$ as $\circ$: -20 dB, $\square$: - 5 dB, $\triangle$: 5 dB. Our method and MDBA for -20 dB overlap. MDBA (5 dB) and BDBA (-5 dB) overlap. Our method (5 dB) and MDBA (-5 dB) overlap.
• Figure 4: Throughput for the Secondary Users