Cross-Layer Optimization for Off-Network Public Safety Communications in 4G LTE and 5G NR

Thesis (Ph.D.)--University of Washington, 2023As public safety migrates from existing digital Land Mobile Radio (LMR) networks to Third Generation Partnership Project (3GPP) based FirstNet, an issue of key concern is coverage. To handle off-network scenarios, 3GPP has introduced Sidelink, a communication link that facilitates direct communication between user equipment (UE). In this thesis, we analyze and propose improvements to the two standards which constitute the Sidelink, Fourth Generation (4G) Long Term Evolution (LTE) Proximity Services (ProSe) and Fifth Generation (5G) New Radio (NR) Cellular Vehicle-to-Anything (C-V2X). As both standards are relatively new and poorly understood, we attempt to model them in terms of their Key Performance Indicators (KPIs). These models allow us to validate the performance of the network simulator 3 (ns-3) implementations of these standards and to determine how best to set the Physical (PHY) and Medium Access Control (MAC) layer parameters to optimize the KPIs. The challenge in any real-world scenario is that some number of PHY parameters that contribute to the performance of a KPI (e.g., the number of UE participating in the ad hoc network) are hidden from UEs, hindering the optimal setting of PHY/MAC parameters. To deal with this challenge, we use the previously developed models to aid in learning the hidden parameters during operation and set PHY/MAC parameters according to those approximations, thereby improving performance. In ProSe, we examine the performance of the ProSe Device-to-Device (D2D) direct discovery process in out-of-coverage scenarios. We model individual discovery periods as a slotted random access protocol with half-duplex (HD) UE and the entire discovery process as a modified coupon collectors problem. We use the open-source network simulator 3 (ns-3) to validate our model and evaluate the discovery process's performance as a function of the size of the resource pool, UE density, and transmission probability. We establish there exists an optimal transmission probability that minimizes discovery time for a given network configuration and develop a method to allow UEs to learn the optimal transmission probability during discovery by estimating the hidden parameters. For C-V2X, we present a novel distributed blind retransmission algorithm in out-of-coverage (mode 2) scenarios. Our contribution is intended as an enhancement to standard (3GPP) specified sensing-based Semi-Persistent Scheduling (SPS) resource re-allocation by opportunistically reusing blind retransmissions to improve average per-user throughput for UE-to-UE transmissions. We initially develop a detailed cross-layer model for (per user) throughput of 5G NR mode 2, supported by simulations using the reputable open-source ns-3 (www.nsnam.org) network simulator. This model and simulation analysis provided important insights into the underlying causative trends and led to two novel distributed adaptive algorithms for resource selection: an initial standards-compliant Dynamic Retransmission (D-Re) and a final standards non-compliant Opportunistic Retransmission (O-Re) algorithm. We show that both D-Re and O-Re render average per-UE throughput robust even in the absence of direct knowledge of 5G network parameters, while O-Re significantly improves the averaged per-UE throughput at high UE densities