An APDR/UPV/Light multi-sensor-based indoor positioning method via Factor Graph Optimization

With the rapid development of Micro-Electro-Mechanical Systems (MEMS), smartphones become one of the most convenient tools to realize pedestrian positioning during daily life. Meanwhile, indoor positioning is a crucial part of pedestrian positioning. Facing the challenges associated with insufficient Global Navigation Satellite System (GNSS) coverage, landmark deployment costs, and complex and changeable obstacles in pedestrian indoor positioning, we propose a Factor Graph Optimization (FGO) based adaptive pedestrian dead reckoning (APDR)/up-view image (UPV)/Light integrated indoor positioning method. Firstly, to improve the positioning accuracy of the traditional pedestrian dead reckoning (PDR), we propose an APDR method. This method builds a height-based stride length estimation model and a unified heading error calibration model, improving the accuracy of stride length estimation and heading estimation. Secondly, we propose a light-aid up-view-visual-based indoor positioning method. The scenes in the up-view perspective are simpler and less changeable compared to side-view scenarios, reducing the image processing difficulty and labeling cost. Since lights are widely present in indoor areas, this algorithm utilizes lights as landmarks, eliminating the need for additional anchor deployment. During positioning, the handhold mode aligns more with the natural state of pedestrians using navigation apps on their phones. The up-view images are easily captured in this mode. We use edge operators to realize the light detection and extraction, thereby realizing the single-point position estimation. Then, we utilize a light-intensity-based light sequence matching module to extend the single-point position estimation to the kinematic positioning. Finally, we use the APDR factor, the UPV factor, and the light factor to construct the FGO function, realizing an APDR/UPV/Light multi-sensor-based indoor positioning method. To verify the performance of the proposed methods, we conducted pedestrian positioning experiments with a Huawei Mate40 Pro. The results indicated the feasibility of the APDR method and the light-aid up-view-visual-based indoor positioning method. Additionally, the superior positioning performance of the FGO-based APDR/UPV/Light integrated indoor positioning method is validated.