Abstract: In view that the currently traditional Zero Velocity Update (ZUPT) can theoretically correct the cumulative error, while the state error variable of 3-dimensional velocity, position and attitude is measured by the model which is established by the observations of 3D velocity error with poor stability and low precision results, an 18-dimensional zero-velocity update algorithm is proposed. The algorithm uses state information of adjacent time to calculate state observations that cannot be directly observed in the zero-velocity state, and then uses the Kalman filter to optimally estimate all state errors. To verify the accuracy of the improved ZUPT, the experiments were carried out by using a self-developed IMU. The results show that the improved algorithm has better stability than the traditional ZUPT and the navigation accuracy is improved by 1.7%.

Key words: IMU, zero velocity update, pedestrian self-navigation, observations of errors, Kalman filter