Optimal Channel-Switching Strategies in Multi-channel Wireless Networks.

Channel Allocation;Multi-channel Communication System;Online Learning;Throughput Optimality;Electrical Engineering;Engineering;Electrical Engineering: Systems

Wang, QingsiAnastasopoulos, Achilleas ;

University of Michigan

关键词: Channel Allocation;    Multi-channel Communication System;    Online Learning;    Throughput Optimality;    Electrical Engineering;    Engineering;    Electrical Engineering: Systems;   

Others  :  https://deepblue.lib.umich.edu/bitstream/handle/2027.42/108949/qingsi_1.pdf?sequence=1&isAllowed=y

瑞士|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

The dual nature of scarcity and under-utilization of spectrum resources, as well as recent advances in software-defined radio, led to extensive study on the design of transceivers that are capable of opportunistic channel access. By allowing users to dynamically select which channel(s) to use for transmission, the overall throughput performance and the spectrum utilization of the system can in general be improved, compared to one with a single channel or more static channel allocations. The reason for such improvement lies in the exploitation of the underlying temporal, spatial, spectral and congestion diversity. In this dissertation, we focus on the channel-switching/hopping decision of a (group of) legitimate user(s) in a multi-channel wireless communication system, and study three closely related problems: 1) a jamming defense problem against a no-regret learning attacker, 2) a jamming defense problem with minimax (worst-case) optimal channel-switching strategies, and 3) the throughput optimal strategies for a group of competing users in IEEE 802.11-like medium access schemes.For the first problem we study the interaction between a user and an attacker from a learning perspective, where an online learner naturally adapts to the available information on the adversarial environment over time, and evolves its strategy with certain payoff guarantee. We show how the user can counter a strong learning attacker with knowledge on its learning rationale, and how the learning technique can itself be considered as a countermeasure with no such prior information. We further consider in the second problem the worst-case optimal strategy for the user without prior information on the attacking pattern, except that the attacker is subject to a resource constraint, which models its energy consumption and replenishment process. We provide explicit characterization for the optimal strategies and show the most damaging attacker, interestingly, behaves randomly in an i.i.d. fashion. In the last problem, we consider a group of competing users in a non-adversarial setting. We place the interaction among users in the context of IEEE 802.11-like medium access schemes, and derive decentralized channel allocation for overall throughput improvement. We show the typically rule-of-thumb load balancing principle in spectrum resource sharing can be indeed throughput optimal.