Indoor Localization with Machine Learning Based on Interpolated Received Signals of Phased Array

This paper presents indoor localization with machine learning, which estimates the position of a target three-dimensionally based on the interpolated received signals of a phased array. In a fundamental study, retro-directive signals of a small 2-by-2 phased array received from the target are input to a 3-layered neural network (NN) to estimate the location of the target. To reduce the time and cost for the training of this system, especially in the 3D case, the signals from a tag on a coarse grid are interpolated to prepare training data. The higher accuracy of localization with interpolated signals is demonstrated.