Robust and accurate localization algorithms for indoor positioning and navigation
Date of Issue2016-12-02
School of Electrical and Electronic Engineering
The explosive proliferation of mobile devices and the popularity of social networks have spurred extensive demands on Location Based Services (LBSs) in recent decades. Global Positioning System (GPS) is not capable of providing indoor LBS with sufficient localization accuracy due to the lack of line of sight (LoS) transmission channels between satellites and receivers. Hence, developing Indoor Positioning System (IPS) to provide reliable indoor LBS has been a hot research topic in recent years. IPS has been recognized as a crucial component in numerous applications such as asset tracking, logistics, tourism and security. Furthermore, IPS for occupancy detection can also play an important role in energy saving in buildings. Various wireless communication technologies have been exploited for indoor positioning and navigation services in the past two decades. Since the existing IEEE 802.11 (WiFi) network infrastructures, such as WiFi routers, have been widely available in large numbers of commercial and residential buildings and nearly every commercial off-the-shelf (COTS) mobile device is WiFi enabled, WiFi based IPS has become the primary alternative to GPS for indoor positioning. Nevertheless, there are still several bottlenecks that restrain them from large-scale implementation. In this thesis, we aim to propose systematic solutions to overcome the longstanding challenges of existing WiFi based IPSs, and develop novel algorithms and systems that outperform the existing ones in terms of accuracy, reliability, robustness and efficiency. Firstly, in order to overcome the device heterogeneity issue, we propose to standardize WiFi fingerprints by a statistical shape analysis method (i.e. Procrustes analysis), and define Signal Tendency Index (STI) to measure the similarity between such standardized location fingerprints. Secondly, we address the issue of robustness of the WiFi fingerprinting-based IPS against environmental dynamics by proposing an online sequential extreme learning machine (OS-ELM) based localization algorithm. The fast learning speed of OS-ELM can reduce the time and manpower costs for offline site survey. Meanwhile, its online sequential learning ability enables the proposed localization algorithm to adapt to environmental dynamics in a timely manner. Moreover, we also propose a novel online mutual information (OnlineMI) based access point (AP) selection strategy that is able to select the optimal subset of APs to reduce the computational burden and improve the indoor localization accuracy of the IPS. Furthermore, in order to remove the needs of tedious and laborious offline site survey process for WiFi based IPS, we design and develop WinIPS, a WiFi based non-intrusive IPS that enables automatic online radio map construction and adaptation for calibration-free indoor localization. It is able to capture data packets transmitted in the WiFi traffic and extract the received signal strength (RSS) and MAC addresses of both WiFi access points (APs) and mobile devices in a non-intrusive manner. Owing to this unique advantage, the online RSS measurements of APs are obtained and used as online reference points for radio map construction and adaptation in real-time. Our WinIPS received the 3rd Place Award in the IPSN 2014 Microsoft Indoor Localization Competition (the 3rd most accurate system in the Infrastructure-Free Category). In addition, in order to seamlessly integrate the proposed IPS with GPS, we propose BlueDetect, an accurate, fast and energy-efficient scheme for indoor-outdoor (IO) detection and smooth LBS in all environments running on a mobile device based on the emerging low-power iBeacon technology. By leveraging the portable BLE beacons and Bluetooth module on mobile devices, BlueDetect provides precise IO detection results to turn on/off on-board sensors (such as WiFi and GPS) smartly, improve their performances and reduce the power consumption of mobile devices simultaneously. Furthermore, seamless LBS, such as positioning and navigation service, can be realized by BlueDetect, especially in semi-outdoor environments, which cannot be achieved easily by either GPS or IPS. By integrating BlueDetect with on-board motion sensors on smartphone, including accelerometer, magnetometer and gyroscope, the system achieved indoor localization accuracy of 1.37m in the infrastructure-based category in the 2015 Microsoft indoor localization competition. With the aid of the aforementioned algorithms and IPSs, numerous LBSs such as indoor navigation on wearable device (Google Glass), indoor geofencing for a smart lighting control system and seamless indoor outdoor navigation, have been successfully developed and implemented in a wide variety of indoor environments.