Traffic prediction scheme for resource assignment of satellite/terrestrial frequency sharing mobile communication system

The recent development of large aperture on-board multi-beam antennas enables a small-size, low-powered and cellular phone like hand-held terminal as a satellite earth terminal. A single mobile terminal can communicate to both mobile satellite systems and terrestrial systems depend upon his location, QoS and availability of resources among satellite and terrestrial communication systems. In this paper, we propose a new traffic prediction scheme for the integrated satellite/terrestrial frequency sharing mobile communication system. The system shares a common frequency bandwidth in order to enhance the total capacity by a dynamic bandwidth allocation. A key for this allocation depends on a traffic prediction scheme of a few hundreds of terrestrial cells under a footprint of a satellite with a large aperture onboard multi-beam antennas. We propose three traffic predictors based on neural networks for dynamic resource allocation. The performances of the proposed schemes are evaluated in terms of the Relative Traffic Prediction Error and Maximum Traffic Prediction Error by the computer simulation. For the evaluation, we adopt the actual traffic statistic published by Ministry of Internal Affairs and Communications of Japan with population density of terrestrial cells based on the actual population of Aichi, Japan. As results, average traffic prediction error of less than 0.25 is achieved for the prediction interval of one hour, enough for dynamic resource allocation.