Multi-channel stochastic resource allocation and dynamic access scheduling

Modern communication system often has the ability to transmit signals on multiple communication mediums (e.g., RF, visible light) or interfaces (e.g., MAC layer protocols) at the same time. In this thesis, they are generalized. While each channel has different characteristics, a centralized controller with channel condition information will be able to schedule the resource allocated to each channel to achieve various optimization criteria. In this thesis, we focus on two usage scenarios: Indoor hybrid free-space optical (FSO)-WiFi femtocells and multi-channel satellite communication (SATCOM). For the Indoor hybrid free-space optical (FSO)-WiFi femtocells, a smart network controller is designed to determine which channel/interface to use for a specific user/time slot combination to balance some pre-specified objectives such as load balance. In particular, this problem is modeled as a dynamic scheduling problem, which is a Markov decision process problem t! hat is solved using a deep-Q reinforcement learning (RL) framework. For the SATCOM scenario, a smart network controller is proposed to transmit information securely on different channels to mitigate jamming and eavesdropping attacks. The proposed approaches combine elements from game theory and information theory to provide provably secure protocols from an information theoretic viewpoint.

Co Advisor: Bella Bose
Co Advisor: Thinh Nguyen
Minor Advisor: Liang Huang
Committee: Alan Wang
Committee: Arun Natarajan
GCR: John Parmigiani