On the selection of working channels in a channel-hopping cognitive PAN

In a channel-hopping Cognitive Personal Area Network (CPAN), the main objective of the working channel selection mechanism is to avoid collisions with unpredictable primary user activity. As the result, the manner in which the working channel for the next hop is selected, is among the most important determinants of CPAN piconet performance. In this paper we investigate the performance of two working channel selection algorithms and compare them with the simple random selection approach. We show that a simple algorithm where the working channel is selected amongst those that have most recently turned idle offer better performance in the homogeneous case (i.e., where the primary user activity on all channels follows a similar pattern), while the case with heterogeneous primary user activity on different channels is better handled by the algorithm in which predicted probability distribution of channel idle times is used to select the next working channel.