Indoor localization is key to location-based services and applications beyond GNSS coverage. Solutions of different accuracy can be suited to diverse applications. Using infrastructures allows instant global localization and reasonable accuracy with low user cost, often desired by human users with resource-limited mobile devices. Shared infrastructures like LED lighting and WiFi networking, regularly required in buildings, are advocated for minimal infrastructure costs. These solutions have cost-sharing advantages and superior ubiquity. Specifically, visible light positioning (VLP) with rolling-shutter cameras achieves high-accuracy 6-degrees-of-freedom positioning based on LED lights with modulation; WiFi fingerprinting allows meter-level 2D positioning with existing hardware. However,...[ Read more ]

Indoor localization is key to location-based services and applications beyond GNSS coverage. Solutions of different accuracy can be suited to diverse applications. Using infrastructures allows instant global localization and reasonable accuracy with low user cost, often desired by human users with resource-limited mobile devices. Shared infrastructures like LED lighting and WiFi networking, regularly required in buildings, are advocated for minimal infrastructure costs. These solutions have cost-sharing advantages and superior ubiquity. Specifically, visible light positioning (VLP) with rolling-shutter cameras achieves high-accuracy 6-degrees-of-freedom positioning based on LED lights with modulation; WiFi fingerprinting allows meter-level 2D positioning with existing hardware. However, practical challenges remain due to the cost-performance trade-off.

This thesis contributes to shared infrastructure-based indoor localization systems, with a primary focus on improving cost-efficiency in real deployments while allowing competitive performance. The reduced cost covers several phases of the system deployment for VLP and WiFi fingerprinting. The outcomes can further their cost advantages and benefit their adoption. This work spans system design and implementation, testbed development, and experimental validation. The main technical contributions are listed below.

VLP with standard rolling-shutter cameras suffers from insufficient LED features under the regular lighting deployment. I propose a tightly-coupled inertial-aided VLP system with 2-point pose initialization. It maintains high accuracy reliably without increasing hardware and installation costs. To further reduce the amount of modulated LEDs for VLP, I propose a novel inertial-aided VLP system that exploits modulated LEDs and unmodulated lights. It allows competitive performance with reduced hardware costs. Building LED location maps for VLP or building WiFi signal maps for fingerprinting localization is costly through manual surveying. To relieve the workload and to save configuration costs, I propose an efficient LED mapping system and an automatic WiFi signal mapping system, respectively. All systems are evaluated with extensive experiments to verify their efficacy and usability.