Topological Nodes Generation Using Anomaly Scores of WiFi Signal Strengths

Abstract

The essential information for location based service (LBS) is, of course, the location of a mobile robot providing the services. In outdoor environments, GPS is definitely the solution for LBS, however it is difficult to use it in indoor environments due to its weak signal strength. The two main radio-based approaches to the indoor localization are trilateration and fingerprinting, which require either many signal sources or heavy calibration. In this paper, we focus on the fact that the purpose of the localization is to provide services not at the whole locations but at certain locations. For this purpose, we propose a novel topological node generation method using anomaly scores of WiFi signal strengths, enabling to localize the robot under a single WiFi access point. Learning of sequential RSSIs with a hierarchical temporal memory model is able to detect a distinct node. We show the experimental results on node generation and recognition of a node using only one WiFi access point.