Dynamic Control for IAB Systems with Mixture of Latency- and Throughput-Sensitive Traffic

Abstract

The Integrated Access and Backhaul (IAB) architecture proposed by 3GPP is expected to provide a cost-efficient deployment option for 5 G New Radio (NR) systems and should support all the services defined in 5 G. However, IAB relies on multi-hop wireless communications, which makes the packet latency a critical metric for such systems. Latency optimization in IAB is further complicated by the half-duplex constraint forbidding IAB nodes from transmitting and receiving simultaneously over multiple antennas. As a result, active links must be dynamically scheduled to avoid conflicts and satisfy requirements of all supported traffic types in both uplink and downlink directions. In this paper, by utilizing the tools of Markov decision processes (MDP), we develop a reduced-complexity control policy for 5 G self-backhauled systems simultaneously carrying latency- and throughput-sensitive traffic. The proposed policy minimizes the latency metric under low and moderate traffic conditions operating in ultra-low-complexity mode and switches to a more computationally demanding throughput-oriented regime when the offered load increases. Our results show that the core part of the policy – the proposed ultra-low-complexity tree level switching control (TLS) – minimizes the latency for high-priority latency-sensitive traffic over the whole capacity region of the system but also for dominant throughput-oriented traffic over 75–85% of the capacity region.