Machine Learning Approaches for Wireless Spectrum and Energy Intelligence

• Author / Creator

  Wu, Keyu

• Cognitive radio and energy-harvesting technologies improve the efficiency of spectrum use and energy use in communication networks. However, due to the randomness and dynamics of spectral and energy resources, wireless nodes must intelligently adjust their operating configurations (radio frequency and transmission power). With machine learning as primary tools, this thesis addresses three spectrum and energy management problems.

  First, we consider a single-channel energy-harvesting cognitive transmitter, which attempts to maximize data throughput with harvested energy and dynamically available channel. The transmitter needs to determine whether or not to perform spectrum sensing and channel probing, and how much power for transmission, subject to energy status and wireless channel state. The resulting control problem is addressed by a two-stage reinforcement learning algorithm, which finds the optimal policy from data samples when the statistical distributions of harvested energy and channel fading are unknown.

  Second, we consider energy-harvesting sensor, which strives to deliver packets with finite battery capacity and random energy replenishment. A selective transmission strategy is investigated, where low priority packets can be dropped to save energy for high priority data packets. The optimal transmission policy, which determines whether or not a packet should be transmitted, is derived via training an artificial neural network with data samples of packet priorities, wireless channel gains, and harvested energy levels.

  Third, we investigate cooperation among cognitive nodes for reliable spectrum sensing given spectrum heterogeneity (i.e., spatial dependence of spectrum availability). Sensing cooperation can mitigate it. However, the challenge is how to model and exploit spatial correlations to fuse sensing data. To overcome this, spatial correlations among cognitive nodes are modeled as a Markov random field; and given data observations, sensing cooperation is achieved by solving a maximum posterior probability estimator over the Markov random field. Under this methodology, three cooperative sensing algorithms are designed for centralized, cluster-based, and distributed cognitive radio networks. These algorithms offer improved computational efficiency and reduced communication overhead.

• Subjects / Keywords

  • Spectrum management
  • energy management
  • machine learning
  • cognitive radio
  • energy harvesting
  • reinforcement learning
  • cooperative spectrum sensing
  • Markov decision process
  • Markov random field

• Graduation date

  Fall 2018

• Type of Item

  Thesis

• Degree

  Doctor of Philosophy

• DOI

  https://doi.org/10.7939/R3901ZX92

• License

  Permission is hereby granted to the University of Alberta Libraries to reproduce single copies of this thesis and to lend or sell such copies for private, scholarly or scientific research purposes only. Where the thesis is converted to, or otherwise made available in digital form, the University of Alberta will advise potential users of the thesis of these terms. The author reserves all other publication and other rights in association with the copyright in the thesis and, except as herein before provided, neither the thesis nor any substantial portion thereof may be printed or otherwise reproduced in any material form whatsoever without the author's prior written permission.