Mr. Hassan Nooh

Title: Hierarchically Layered Intelligent Non-Terrestrial Networks

For key application scenarios in future wireless systems (e.g., global coverage extension to unserved & underserved areas, and to realise the technical requirements for eMBB, mMTC and URLLC) a multi-layered non-terrestrial network architecture with intelligent decision-making capabilities is essential. While several configurations and application specific scenarios can be conceived within this general scope, many of the challenges that are yet to be addressed in literature tend to have a common theme with combinatorially large search spaces. Most such problems are NP-hard, and we may need to settle for approximate solution methods. These ailments are further exacerbated as requirements become increasingly stringent in time, capacity, and resource usage – among others.

Recent advancements in deep reinforcement learning (DRL) provide a general framework to tackle problems in the combinatorial optimisation landscape with increasingly encouraging results. The inherent data driven nature of these algorithms enables the structure of the problem to be exploited. While we do not need to restrict our attention to DRL alone, this provides us an opportunity to explore and characterise the performance bounds and feasibility of such systems utilising DRL-based architectures against other alternatives such as swarm techniques and evolutionary strategies.