Analysis of a modified Switchable Bayesian Learning Automaton for Cognitive Radio

One of the most important topics in Cognitive Radio communications is that all communication partners change to the same frequency at the same time. A critical aspect of this process is a powerful channel selection algorithm, because each channel switching process requires resources and bears the risk of connection loss. Therefore, it is important to choose the channel that can be expected to be available for the longest time. This requires collecting information about all usable channels and developing a selection strategy. In [1] a machine learning approach, the Switchable Bayesian Learning Automaton (SBLA), is proposed for this task. Recently, we have implemented that algorithm to our cognitive radio simulator, which was presented in [2]. This paper describes the implementation, points out the advantages and drawbacks of the algorithm, and introduces improvements for its use in real-time systems.