A Novel ESKF-Based ZUPT Using Midpoint Integration Approach for Indoor Pedestrian Navigation

We present a novel pedestrian inertial navigation system (INS) using zero-velocity update (ZUPT) and error state Kalman filtering. Our system estimates pedestrian pose using only data from an inertial measurement unit (IMU). To improve accuracy, we use midpoint integration approach to obtain the discrete state and covariance propagation equations, which provide more accurate predictions than Euler integration. Furthermore, we model the slowly varying IMU biases over time to further improve the accuracy of the system. We also construct five observation constraints based on zero-velocity states, including zero acceleration, zero angular, zero velocity, heuristic drift reduction, and planar constraints. These observation constraints effectively control drift and divergence problems. Our foot-mounted IMU experiments show that our system achieves a positioning error of less than 1 m with a positioning error root-mean-square error (RMSE) of 0.23 m when circumnavigating a circle with a total distance of 100 m. Our system outperforms both naïve ZUPT-aided INS and the system using the same constraints as our system before.