On the Provisioning of Ultra-Reliable Low-Latency Services in IoT Networks with Multipath Diversity

Abstract

The unprecedented increase in the number of smart connected devices invoked a plethora of diverse applications with different performance requirements stipulating various network management strategies. Ultra-reliable low-latency communication (URLLC), one of the promised 5G dimensions, is expected to enable mission-critical applications while adhering to different levels of reliability and latency requirements. At its core, URLLC rests on the notion of providing stringent reliability and latency requirements, in which guaranteed network availability becomes a necessity. In this thesis, we propose different approaches for providing URLLC services in two different contexts: wired and wireless network scenarios. Motivated by the fact that extreme reliability and lowest latency guarantees incur extremely high costs, we aim to provide these requirements using multipath diversity. However, providing such guarantees requires careful resource provisioning and management. In the fi rst part of this work, we utilize the notion of Network Slicing (NS), one of the key paradigms that can offer performance guarantees through customized network management of software defi ned networking (SDN) as well as path diversity. Multiple disjoint paths may be selected to ensure the reliability and latency requirements of supported URLLC-based applications. We formulate the problem as a mixed integer program and then propose a decomposition approach that can efficiently associate each application to its corresponding paths while meeting its strict quality demands. In the second part of this work, we leverage unmanned aerial vehicles (UAVs) as redundant flying edge servers to fulfi l the reliability and latency requirements of devices in a given IoT network. UAVs fly closer to devices which target reliability cannot be met by the serving base station (or access point) equipped with mobile edge computing (MEC) capabilities. We de ne the constraints of the problem, model it as a Markov Decision Process, and propose a reinforcement learning-based solution to optimize the UAV trajectory. Simulation results are presented for both parts of the thesis to illustrate the effectiveness of the proposed solutions and algorithms in comparison with optimal solutions and baseline algorithms.