Soutenance de thèse : Shashwat MISHRA

Supporting Massive Machine Type Communication in Beyond 5G Networks: Fundamental Limits and Practical Heuristics

Doctorant : Shashwat MISHRA

Laboratoire INSA : CITI

École doctorale : ED160 : EEA (Electronique, Electrotechnique, Automatique)

This work focuses on supporting massive access in beyond 5G networks. First, we derive the fundamental information theoretic limits of connectivity under the spatial continuum model considering Rayleigh fading. We prove that achieving perfect reliability is impossible in such a setup and therefore we evaluate the fundamental EE-SE tradeoff considering outage. We extend the formulation to the case of discrete spatial continuum model as well as the case of multiple fading states and show that the discrete model approaches the continuum case with relatively few levels of superposition. Following this theoretical analysis, we present a graph-matching based heuristic for resource allocation in NOMA-aided massive access with multiple fading states. This pragmatic approach relies on grant-based mechanism such as the fast uplink grant to efficiently schedule resources for NOMA-based massive machine type deployments. We present a comparative analysis of the fundamental results and the heuristics and show that the heuristics closely approach the fundamental limits. This work comprehensively covers both the uplink and downlink massive access scenarios considering the respective power budgets as well as quality of service constraints. Moving to the multi-antenna setup, we present a graph neural network-based framework for real-time power allocation in cell- free massive MIMO networks to support massive access. This part of the thesis concentrates on formulating novel graph representations for the network that are used in a supervised learning framework to adaptively learn approximating the maximum ratio transmission precoding for partially and fully connected cell-free massive MIMO systems.