Hybrid Indoor Positioning Method of BLE and Monocular VINS-Based Smartphone

The visual simultaneous localization and mapping (VSLAM) method has a significant advantage in indoor positioning methods. Visual–inertial navigation system (VINS)-Mono is one of the state-of-the-art works widely used in many augmented reality (AR)/virtual reality (VR) applications on a smartphone. However, in indoor positioning based on a smartphone, the low accuracy of the inertial measurement unit (IMU) leads to a disappointing estimate of the initialization scale of VINS-Mono, which makes the assignable error in subsequent positioning. In addition, the time consumption of the loop detection module in VINS-Mono is high, which can be replaced if there are other global measurements based on smartphone sensors, for example, Bluetooth low energy (BLE) broadcast frame. Indoor positioning methods based on BLE have been implemented in Wechat mini program because of the lightweight performance in memory and complexity. However, almost all related methods process BLE information in a synchronous way for triangulation or fingerprint positioning. Such processing does not consider the ranging error caused by the asynchronous reception of the BLE broadcast frame, which leads to nonnegligible positioning errors. Therefore, we propose a lightweight VINS-Mono-BLE framework using only the smartphone’s monocular vision, low-precision IMU, and BLE. We design the BLE back-propagation algorithm to eliminate the influence of asynchronous reception and then fuse it with the vision and IMU measurements. Furthermore, we use BLE as a global observation instead of the loop detection module to improve the accuracy of VINS-Mono and reduce the time cost of the algorithm. Meanwhile, we have conducted experiments in the laboratory and commercial scenarios, respectively. The experimental results show that our proposed BLE back-propagation algorithm can effectively reduce the ranging error and improve the positioning accuracy of the fusion algorithm.