시퀀스 기반 위치추정 시스템을 위한 효율적인 노드 배치 및 양자기 설계 알고리즘에 관한 연구

Abstract

In this thesis, we study two algorithms for the sequence-based localization (SBL) systems which have been recently used for in-door positioning. The first is the node deployment algorithm for SBL systems. The existing node selection/deployment algorithms seek to place nodes in the centroid of the area where more targets can be found. However, it is observed that the boundaries separating these areas might be appropriate for node locations in SBL systems. From this observation, we propose an efficient node deployment algorithm that determines boundaries using K-means clustering and finds potential node locations based on the bisection method for a low complexity design. The second is a low-complexity algorithm that allows us to design quantizers at sensor nodes by using the reduced number of training samples which are obtained from distance samples between targets and centroids, each centroid assigned to one of the sequences constructed by localization process in SBL systems. In addition, we further improve performance of the design algorithm by generating more training samples for the sequences with higher probability. We demonstrate through experiments that the proposed algorithms for node deployment and quantizer design at local nodes for SBL systems show a significant performance gain as compared with typical and previous designs.