Hierarchical global localization with radio signals and spatial maps

Abstract

This dissertation is regarding a method to estimate the global location of mobile robots and pedestrians by fusing Wi-Fi’s received signal strength indication (RSSI), dead-reckoning position information, and spatial map data in indoor environments. Recently, Wi-Fi routers, as typical pervasive network devices, have been widely installed in indoor spaces such as homes, offices, and shopping malls owing to the wide spread of ubiquitous computing. Wi-Fi provides communication functions

furthermore, its RSSI can be used as an invisible marker for indoor localization. In this study, I proposed two methods to improve issues in conventional localization methods of mobile robots and pedestrians by utilizing Wi-Fi RSSI based localization. The first method, hierarchical global localization framework, estimates the position of mobile robots by integrating the Wi-Fi RSSI-based fingerprint analysis method and an indoor grid map-based particle filter. This method determines the initialization region of a particle filter using fingerprint analysis results and estimates position by optimizing the sampling distribution. This method can solve the local minimum problem of particle filters and optimize the sampling size more effectively. The second method, sequential motion tracking, estimates a pedestrian’s position stably by combining the Wi-Fi RSSI-based fingerprint analysis method and indoor topological map motion information. This method estimates position by matching the moving trajectory of a topological map with the motion information accumulated using a dead-reckoning sensor in a global region discovered through fingerprint analysis results. This method can improve the continuous localization’s accuracy and the technique’s stability compared with the conventional methods such as the fingerprint analysis and the particle filter. Finally, the two proposed methods were applied to a mobile robot and a smartphone, separately, and their performances were investigated through actual localization test results. Furthermore, their effectiveness was verified based on comparative analysis with other methods.