Predicting spectrum status duration using non-linear homotopy estimation based HMM for UAV communications

Abstract

As the spectrum resources are limited, unmanned aerial vehicles (UAVs) may communicate by flexible spectrum sharing. To avoid interference to primary users, UAVs require spectrum state prediction. Furthermore, prediction of spectrum status duration is also needed since the primary users may access just after the UAVs access. In other words, a UAV should predict the spectrum status duration and choose one with enough idle duration. Unfortunately, reported methods for duration prediction, such as the non-stationary hidden Markov model (NSHMM), may not work for UAV communications. These methods predict the duration of the next moment at the current location, while the UAV is probably to have moved to the next location when obtaining the prediction result. Therefore, the UAV should predict the spectrum status duration of the next moment at the next location. The main challenge is that it is difficult to obtain the historical data of the next location in advance during the UAV flight. In this paper, we propose a new prediction approach that firstly estimates the historical data of the next location by using a non-linear homotopy relationship, then implements the NSHMM to predict the duration, referring to the non-linear homotopy estimation based HMM (NLH-HMM). Experimental results show that the NLH-HMM based method can efficiently predict the spectrum status duration of the next location.