Public defence in Communications Engineering, M.Sc. Mohsen Amidzade

The title of the thesis: Probabilistic cache policy design for cellular networks with stochastic geometry analysis

Doctoral student: Mohsen Amidzade
Opponent: Prof. Di Yuan, Uppsala University, Sweden, and Prof. Italo Atzeni, University of Oulu
Custos: Prof. Olav Tirkkonen, Aalto University School of Electrical Engineering, Department of Information and Communications Engineering 

To tackle the challenges arising from the growing demand of mobile data in contemporary cellular networks while simultaneously optimizing resource consumption in the network, novel methodologies need to be developed for adapting to the dynamics of the wireless networks. Edge caching is a promising solution that provides a resource-efficient mechanism to alleviate the unprecedented traffic escalation. The idea is to proactively store the most popular contents at the network edge and then to transmit the cached contents to requesting users using a cache delivery strategy. Analyzing edge caching for large-scale cellular networks poses an extraordinary challenge due to the high dimensionality of the parameter space. Further, cache policy design in this framework gives rise to intricate optimization problems. 

This dissertation thus focuses on devising efficient cache delivery schemes to optimize network performance metrics. To cope with the challenge of analyzing the developed schemes in large-scale networks, this thesis employs stochastic geometry—a probabilistic analytical tool modeling the deployment of network edges by Poisson point processes. This is complemented by a distributed probabilistic content placement strategy at the edges. The thesis leverages time-varying optimization and reinforcement learning techniques to tackle the intricacy of cache policy optimization problems. Hence, it efficiently optimizes the developed cache policy at the network level by utilizing network-wide parameters rather than focusing on an edge-specific approach. 

This dissertation develops and analyses a novel delivery scheme, termed orthogonal multi-point multicast, which exhibits enhanced quality-of-service in cache-enabled networks, when compared to conventional on-demand unicast schemes. Moreover, by combining it with conventional delivery schemes, a hybrid delivery approach is formed, leveraging the benefits of both multi-point multicast and on-demand unicast, leading to further improvements in quality of service. The findings of this thesis present energy -efficient delivery schemes with the potential to enhance conventional delivery schemes for cache-enabled cellular networks. Moreover, it sheds light on the applications of stochastic geometry, time-varying optimization, and reinforcement learning in the field of wireless communications.

Keywords: Probabilistic cache policy, Distributed cache placement, Multipoint multicast transmission, Stochastic geometry, Reinforcement Learning, Time-Varying Optimization

Thesis available for public display 10 days prior to the defence at: https://aaltodoc.aalto.fi/doc_public/eonly/riiputus/

Contact:

Doctoral theses in the School of Electrical Engineering: https://aaltodoc.aalto.fi/handle/123456789/53