A Client-Assisted Approach Based on User Collaboration for Indoor Positioning

Abstract

Fingerprint indoor positioning technology is one of the most attractive and promising techniques for mobile devices positioning. However, it is also time consuming for building the radio map offline and cannot provide reliable accuracy due to the changing environment. In response to this compelling problem, a client-assisted (CA) approach is proposed for radio map construction based on multi-user collaboration. In this method, multi-dimensional scaling (MDS) approach is used to transform the distance to two dimensional data. MDS as an set of analytical technique has been used for many years in fields like economics and marketing research. It is a suitable for reducing the data dimensionality to points in two or three dimensional space. In CA, this can be used where only distances between users are known which are used as an input data. All the client data is collected at one point because of the centralized nature of the MDS. It is advantageous in that, MDS can reconstruct the relative map of the network even when there are no anchor clients (clients with a priori known location). Given a sufficient number of known client locations, MDS generates accurate position estimation enabling local map to be transformed into an absolute map.

Based on gradient features of users’ walking speed, solve-stuck (SS) method is adopted to improve the efficiency by reducing calculation complexity and solving “data drift” problem. Radio map with a small number of labeled fingerprints can be self-updated by iterating the distance between users. Kalman filter (KF) method is used to remove the noise to make the trajectory closer to the ideal trajectory. We further demonstrate the influence of density distribution and time-cost of different number of clients. The experimental results show that CA approach can improve positioning accuracy with acceptable time-cost.