Ubiquitous positioning based smartphone level GNSS, camera, and INS for smart city applications

Abstract

In recent years, the surge in smart mobility has been driven by advancements in intelligence, sensors, and cutting-edge networks. This trend has ushered in diverse applications in mobile positioning to provide real-time location information for users. Smartphones, with their varied sensor capabilities, hold significant potential in enhancing smart mobility by delivering real-time location insights and enabling a wide range of applications and services that enhance efficiency, safety, and user experiences. However, urban mobile positioning encounters challenges from Global Navigation Satellite Systems (GNSS), such as signal reception due to complex environmental conditions. There are non-light of sight (NLOS) signals and multipath effects due to the buildings and obstacles in dense urban areas. Efforts focus on enhancing GNSS accuracy by countering NLOS signal errors. Visual-inertial navigation system (VINS), improving accuracy in GNSS-deprived urban scenarios. Recent studies explore VINS integration, improving pose estimation by combining GNSS, INS, and visual sensors. Real urban tests showcase enhanced accuracy compared to standalone systems. However, GNSS/VINS integration persists with challenges like signal loss, complexity, and robustness. Multipath errors, signal obstruction, and drift add to the hurdles. Solutions include 3D models for urban positioning, Doppler integration for smoother outcomes, and IO detection for compromised GNSS settings indoors. To enhance indoor positioning, an Indoor/Outdoor (IO) detection is incorporated while Visual-Inertial Odometry (VIO) bridges GNSS gaps in GNSS-denied areas. In the experiment, a complex environment dataset is collected in Hong Kong, a high-density urban canyon. This thesis integrates the GNSS and VINS on smartphones. It is a cost-deducted way to acquire a precise loosely coupled pedestrian positioning in an urban canyon. Support Vector Machine (SVM) is used for Indoor/Outdoor (IO) detection. By integrating adaptive 3DMA GNSS/VINS using Factor Graph Optimisation (FGO) in a loosely coupled manner, promising sub-meter accuracy.