Walking Dynamics, User Variability, and Window Size Effects in FGO-Based Smartphone PDR+GNSS Fusion.

Area of Science:

• * Navigation and Positioning
• * Sensor Fusion
• * Human Motion Analysis

Background:

• * Smartphone-based pedestrian positioning relies on Global Navigation Satellite Systems (GNSS) and Pedestrian Dead Reckoning (PDR).
• * Factor Graph Optimization (FGO) shows promise for fusing GNSS and PDR data, but its optimal configuration concerning human motion dynamics is underexplored.
• * Existing methods like Kalman Filters (KF) may struggle with motion-dependent errors in PDR.

Purpose of the Study:

• * To investigate the impact of walking dynamics on the optimal sliding-window FGO (SWFGO) configuration for pedestrian positioning.
• * To compare the error mitigation capabilities of FGO against KF under varying motion conditions.
• * To determine the relationship between walking speed and the FGO window size for stable positioning accuracy.

Main Methods:

• * Data collection from ten pedestrians across four distinct motion types: slow walking, normal walking, jogging, and running.
• * Analysis of sliding-window FGO (SWFGO) performance with varying window sizes (e.g., 1, 10, 30 poses).
• * Comparative evaluation of FGO and KF in suppressing PDR outliers caused by motion irregularities.

Main Results:

• * An SWFGO window size of approximately 10 poses achieves a favorable balance between positioning accuracy and computational load.
• * This 10-pose window size offers significant accuracy improvements over a 1-pose window and approaches batch FGO performance at lower computational cost.
• * Increasing the window size to 30 poses provides minimal additional accuracy gains but increases computational demands, a trend consistent across all motion types.
• * Both FGO and SWFGO demonstrate superior outlier reduction compared to KF, enhancing robustness against gait variability and transient disturbances.

Conclusions:

• * SWFGO with a 10-pose window is an effective strategy for robust and accurate smartphone-based pedestrian positioning.
• * FGO significantly outperforms KF in mitigating motion-induced PDR errors, offering improved reliability across diverse walking dynamics.
• * The findings provide practical guidance for configuring FGO for pedestrian navigation systems, optimizing performance based on expected motion patterns.