Accession Number : AD1030771


Title :   Intelligent Spectrum Handoff via Docitive Learning in Cognitive Radio Networks (CRNs)


Descriptive Note : Technical Report,01 Dec 2012,01 Sep 2016


Corporate Author : University of Alabama at Tuscaloosa Tuscaloosa United States


Personal Author(s) : Hu,Fei ; Kumar,Sunil


Full Text : https://apps.dtic.mil/dtic/tr/fulltext/u2/1030771.pdf


Report Date : 01 Mar 2017


Pagination or Media Count : 87


Abstract : In this project, we target the design of a Docitive Learning (also called transfer learning) based spectrum handoff design in Cognitive Radio Networks (CRNs). If the current channel quality is below a threshold, the secondary user (SU) should make one of the following decisions: (1) stay in the same channel and wait for it to become idle again (this strategy is called wait-and-stay), (2) stay in the same channel and adjust the channel parameters according to the varying channel conditions (this strategy is called stay-and-adjust), or (3) switch to another idle channel that meets its QoS requirement (this is the conventional spectrum handoff). In this project, we have applied a critic-based transfer learning model to implement an intelligent spectrum handoff with both node-to-node learning and self-learning. Additionally, a multi-teacher-based transfer learning scheme is used to learn handoff parameters from multiple neighbors. We have also built a comprehensive CRN testbed for spectrum handoff test. Such a testbed has other essential CRN components, such as spectrum sensing, mining and handoff.


Descriptors :   cognitive radio , wireless communications , hidden markov models , signal processing , supervised machine learning , false alarms , multiple access , packet loss , algorithms


Subject Categories : Radio Communications
      Computer Systems


Distribution Statement : APPROVED FOR PUBLIC RELEASE