- This event has passed.
SMART Seminar Series
May 22 @ 11:30 am - 12:30 pm
Presenter – Yan Li
Ms Yan Li is currently a PhD student at the Department of Electrical, Electronics, Engineering, University of Melbourne, working on a project to track mobile user in a wireless indoor environment. Before that, she completed her master by research degree at the Centre for Autonomous Systems (CAS), University of Technology, Sydney (UTS), working on a shoe-mounted INS based pedestrian navigation system. Yan’s research interests include wireless sensor network, mobile sensing, inertial navigation and sensor fusion.
A Probabilistic Approach to WiFi based Indoor Localisation
The aim of the project is to develop a practical prototype model of a reliable room location awareness system for mobile users in a real public wireless indoor environment. Most existing wireless indoor positioning systems have been developed and validated in small scale environments or manually set-up test environments, where limited numbers of APs are installed, taking account of the geometry and coverage of the area. This leaves an important and practical question as to whether they perform well in large-scale campus wireless environments, with large numbers of APs deployed in an unplanned way.
Localisation in a more complex scenario in which three typical problems arise. Firstly, in a large-scale public wireless environment where hundreds of APs can be sensed leading to a high-dimensional classification problem. Secondly, in terms of practical applications, users may carry different devices for training data collection, heterogeneous devices record different received signal RSS levels, thus device heterogeneity has to be taken into consideration when constructing the cross-device radio map. These challenges are exacerbated by differences in the device populations between the survey and online client phases. Thirdly, the RSS samples collected from APs are unstable and interfered by limited coverage range and unstable network communication, APs are not necessarily visible in every scanning cycle leading to missing data issue. The proposed system works properly with an average matching accuracy of 97% in a real-world wireless network, it can also localize a new device new to the environment with an average of 94% accuracy.
The proposed system is designed to handle practical problems such as device heterogeneity, signal reliability and environment complexity. The system can handle data uploaded by diverse devices using a simple linear regression model, which can be widely applied in any existing system that needs implicit user participation or free calibration. It also provides a prototype of the indoor localization system deployed in any potential public spots and localization applications such as guiding customers in a shopping mall or monitoring patients in a hospital. The system can be applied to a wide range of localization applications in a practical indoor environment regardless of the quality of the signals, the number of the APs, the heterogeneous devices, the interference from other channels, the time-varying phenomena or the complexity of the environment.