From Localization to Floor Plan Identification

The Hong Kong University of Science and Technology
Department of Computer Science and Engineering


PhD Thesis Defence


Title: "From Localization to Floor Plan Identification"

By

Mr. Xiaonan GUO


Abstract

With the rapid development in pervasive computing and Location-based 
Services (LBS), location information play an important role in various 
applications, such as tracking, surveillance, advertising and etc. 
Further, such location information enable us to infer location-aware 
information. On the one hand, researches have proposed numbers of 
approaches that enable localization in indoor environments as GPS does not 
work well under such circumstances. These approaches mainly fall into two 
categories, based on RSS measurement and techniques that leverage inertial 
sensors.One the other hand, after obtaining location information, 
researchers aim to provide higher level information such as pathway and 
floor plan by leveraging accurate localization result.

However, all of above have several limitations in real practice. First, 
RSS-based ranging technique suffers from multipath phenomenon and it is 
not easy to localize multiple objects. Second, inertial sensor based 
techniques are far from accuracy due to the fact that they all based on 
step counting and step length vary from person to person. Third, based on 
location information, researchers focus on how to use sensor reading from 
smart phone to automatically reconstruct the walking pathway or floor plan 
and such information is insufficient to provide better services.

To fill in the gap, in this dissertation, I propose several novel 
techniques that overcome the above limitations. First, I leverage 
frequency diversity to differentiate Line-of-sight (LOS) path from 
multipath and then construct LOS radio map that can localize multiple 
objects based on finger print technique. Third, I improve the localization 
accuracy of inertial sensor based method by formulating the calculation of 
displacement as a curve fitting problem. Last, to provide better services, 
the proposed system achieve floor plan identification by leveraging smart 
phone sensor reading.


Date:			Tuesday, 6 August 2013

Time:			1:00pm – 3:00pm

Venue:			Room 3494
 			Lifts 25/26

Chairman:		Prof. Christopher Leung (CIVL)

Committee Members:	Prof. Lionel Ni (Supervisor)
 			Prof. Lei Chen
 			Prof. Huamin Qu
 			Prof. Furong Gao (CBME)
                        Prof. Weijia Jia (Comp. Sci., CityU)


**** ALL are Welcome ****