Rapid Radio Map Construction via Vehicle-Based SLAM in GNSS-Denied Environments

This study proposes a practical framework for rapidly generating a fingerprinting-based radio map in GNSS-denied environments such as underground parking lots. The system employs vehicle-based simultaneous localization and mapping (SLAM) to estimate trajectories using data collected from onboard smartphones and an OBD2 interface. During normal driving, Bluetooth Low Energy (BLE) signals and inertial sensor measurements are recorded to construct spatial radio maps. To address device-dependent signal characteristics, Android and iOS smartphones are used simultaneously to generate platform-specific fingerprinting maps. To improve trajectory accuracy, a heading correction algorithm compensates for gyroscope drift during straight driving segments. Loop closures are then detected using both heading sequence similarity and BLE RSSI vector correlation within a graph-based SLAM framework. The optimized trajectory is integrated with BLE signal measurements to construct a spatial radio map. Experiments conducted in a large underground parking facility accommodating approximately 800 vehicles demonstrate that the entire radio map can be generated within approximately one hour of driving. The proposed approach significantly reduces the time and labor required for conventional site surveys while maintaining spatial consistency of the radio map. These results demonstrate the feasibility of scalable BLE-based indoor positioning in large GNSS-denied environments.