An Inverted Pendulum Model of Walking for Predicting Navigation Uncertainty of Pedestrian in Case of Foot-mounted Inertial Sensors

This paper presents a simplified model for predicting navigation uncertainty of a pedestrian. The model simulates trajectories of a person's foot, and these trajectories are then used to generate simulated IMU readings. Eight different noise errors are considered for both the simulated accelerometer and gyroscope readings, including white noise, bias instability, random walk, scale factor error, misalignment, turn-on bias, limited full-scale range, and limited bandwidth. We conducted a series of pedestrian walking experiments to validate the proposed model. The experimental results showed that the position Root-Mean-Square-Errors (RMSEs) in the simulations and in the experiments had a discrepancy of 6% for about 40 [m] of walk. The model also predicted the bounds of the vertical position drift, which matched the trend of estimated vertical position uncertainties in the experiments. We concluded that the model could predict, with sufficient accuracy, the navigation uncertainty for foot-mounted IMU-based systems, and we suggested future research to enhance the model with additional details of foot motion to further improve the prediction accuracy.