Previous Abstract Return to Session C6 Next Abstract

Session C6: Terrestrial Signals of Opportunity-Based Navigation Systems

Indoor Localization Based on Machine LearningAssisted PDR and Signals of Opportunity From Ambient Generic BLE Devices
Masakatsu Kourogi, Akihiro Sato, Satoki Ogiso, Ryosuke Ichikari, and Takashi Okuma, AIST
Location: Grand Ballroom IJ
Date/Time: Thursday, May. 1, 4:05 p.m.

In this study, we explored a method for indoor localization of smartphones equipped with Inertial Measurement Units (IMUs) and Bluetooth Low Energy (BLE) capabilities by combining Pedestrian Dead Reckoning (PDR) with machine learning and opportunistic signals from ambient, generic BLE devices. This method employs PDR with short-term integration of acceleration data, which is corrected using a machine learning-based speed estimation, to generate first-generation absolute trajectories with BLE scans in environments where the locations of BLE devices are unknown. The method estimates the locations of BLE devices solely based on these trajectories and BLE scans. Once the BLE device locations are estimated, subsequent smartphone walkthroughs not only enable absolute localization of the smartphones but also refine the BLE device locations. Early experimental results showed that the typical localization error for smartphones was less than 2 meters when using PDR and BLE scans with the estimated BLE device locations.
Keywords—Indoor localization, Pedestrian Dead Reckoning (PDR), Machine learning, Bluetooth Low Energy (BLE), Signals of opportunity

Previous Abstract Return to Session C6 Next Abstract