Dynamic time-domain duplexing for self-backhauled millimeter wave cellular networks

Millimeter wave (mmW) bands between 30 and 300 GHz have attracted considerable attention for nextgeneration cellular networks due to vast quantities of available spectrum and the possibility of very high-dimensional antenna arrays. However, a key issue in these systems is range: mmW signals are extremely vulnerable to shadowing and poor high-frequency propagation. Multi-hop relaying is therefore a natural technology for such systems to improve cell range and cell edge rates without the addition of wired access points. This paper studies the problem of scheduling for a simple infrastructure cellular relay system where communication between wired base stations and User Equipment follow a hierarchical tree structure through fixed relay nodes. Such a systems builds naturally on existing cellular mmW backhaul by adding mmW in the access links. A key feature of the proposed system is that TDD duplexing selections can be made on a link-by-link basis due to directional isolation from other links. We devise an efficient, greedy algorithm for centralized scheduling that maximizes network utility by jointly optimizing the duplexing schedule and resources allocation for dense, relay-enhanced OFDMA/TDD mmW networks. The proposed algorithm can dynamically adapt to loading, channel conditions and traffic demands. Significant throughput gains and improved resource utilization offered by our algorithm over the static, globally-synchronized TDD patterns are demonstrated through simulations based on empirically-derived channel models at 28 GHz.