Network Selection Scheme Using Positioning Information for Heterogeneous Wireless System: Generation Method of Map Information Using Deep Learning

摘要

We have already proposed traffic navigation, which is high-accuracy network selection scheme using map information and highly accurate positioning information. In this paper, we discuss the generation method of map information. For making map information, it is a main issue to reduce the network load (i.e., the number of collected data for creating map information) with keeping the system selection accuracy. In order to reduce the network load with keeping the accuracy, we propose a map generation method using deep learning in this paper. At first in the proposed method, the data of received signal strength indicator (RSSI) linking with high precision position information aggregate into heterogeneous network control server (HCS). Based on the collected data, the HCS creates a deep neural network (DNN). The generated DNN is delivered to each terminal. The terminal can select the optimum network by inputting the position information to the DNN. As a result of evaluation by computer simulation, it is shown that optimum network selection is possible with higher precision even when the number of collected data is small.