Opportunistic spectrum usage scheduling: Time series approach

Prediction of idle times of the licensed users, based on the previous history of their utilization pattern is expected to help the opportunistic users choose the best available channel improving the spectrum utilization. In this paper, we propose a novel spectrum assignment technique that works as a service time planner as well as activity analyzer. The activity analyzer learns and identifies the traffic pattern over the channel of interest and applies suitable prediction techniques to predict the duration of next available idle time. The main advantage of the proposed activity analyzer is that it also takes into considerations of the time series variations of the OFF time traffic that may have some patterns like correlated or short-term correlated, cyclic or alternative, increasing or decreasing trends, seasonal effects or outliers in estimating the idle time of the licensed user spectrum. Our proposed prediction method has shown reduced prediction error as compared to the predictive methods without the time series variation considerations. Also the optimal spectrum assigner is introduced that adaptively utilizes the predicted OFF time information from the activity analyzer, and matches it with the job duration requirements of the opportunistic users that are provided by the service time planner. It is also shown that, our method outperforms the random channel selection methods with respect to improved spectrum utilization, job completion rate of opportunistic user and reduced collision rate with licensed user. Wireless traffic generated by OPNET for different applications such as email, FTP, HTTP and video browsing are used in simulations to prove the practicability of the proposed method.