The Complex Indoor Localization Technique Based on Ontology and SLAM-method

Abstract

One of the problems associated with methods using to determine the location of users of mobile devices within indoors, based on Wi-Fi radio signals, is the time-consuming procedure for setting up and placing equipment, which includes building a map of the room, creating a map of radio signals, or calibrating the radio signal propagation model. In solving this problem, it is planned to use complex indoor localization technique based on the usage of ontology and the SLAM method, which includes the phase of forming a training sample, as well as the phase of simultaneous navigation and mapping. The SLAM method is based on The Gaussian Process Latent Variable Model (GP-LVM) and includes requirements for correlation of the signal level values of the nearest points of the user's localization, for which the parameters of the correlation function are configured based on the training sample. The proposed method is based on solving the regression problem using machine learning methods to form a training sample, as well as solving the problem of reducing the dimension for simultaneous navigation and map construction. As a training sample, the smartphone's internal sensor readings (steps and rotation angles) and Wi-Fi received signal strength values obtained using crowd calculations are used. The resulting training sample is used to determine the parameters of the correlation function that sets the correlation between the user's localization points. The proposed ontology is intended to determining the user's entrance to the room and searching for Wi-Fi access points.

Keywords

• indoor localization
• ontology
• multilateration
• slam
• gp-lvm