A Pedestrian Positioning Method for Urban Canyon Environments Using Magnetic Field Matching/Inertial Odometry Fusion

In urban canyon environments, Global Navigation Satellite Systems (GNSS) frequently fail to provide reliable positioning services due to signal obstruction and multipath effects. This paper proposes a pedestrian positioning method that integrates magnetic field matching, learned inertial odometry and Inertial Navigation System (INS). Initially, the feasibility of magnetic matching positioning is validated by investigating the magnetic field characteristics in typical urban canyons. Subsequently, magnetic field profiles are constructed using relative position and attitude information generated by learned inertial odometry, rather than conventional Pedestrian Dead Reckoning (PDR), thereby enhancing the adaptability of the magnetic matching algorithm to transition between different phone holding modes and usage scenarios. Furthermore, to address the limitation of the Extended Kalman Filter (EKF) in integrating historical state information, the Multi-State Constraint Kalman Filter (MSCKF) is employed to fuse INS, relative displacement increments from learned inertial odometry, and magnetic matching positioning results, thus formulating a robust and reliable pedestrian positioning solution. The proposed method is evaluated through multiple field tests in urban canyon environment, covering an area of 5000 square meters with a total walking distance of 8.25 kilometers, achieving a positioning accuracy with a Root Mean Square (RMS) error of 2.98 m.