Zero velocity update with stepwise smoothing for inertial pedestrian navigation

Zero velocity update (ZUPT) is an effective way to correct low cost inertial measurement unit (IMU) errors when it is foot-mounted for pedestrian navigation. The stance phase in steps provides zero velocity measurement for inertial sensor error correction. As the errors of IMU estimated position and velocity grow rapidly with time between each correction, ZUPT applied at each step leads to sharp corrections and discontinuities in the estimated trajectory. For motion analysis and visualization, these large corrections are undesirable. Consequently, the implementation of smoothing for ZUPT-aided INS is considered to eliminate the sharp corrections. In this paper, we propose a closed loop Rauch-Tung-Striebel (RTS) smoother using a 24 error states extended Kalman filter (EKF) implement on our previous pedestrian navigation systems. Unlike common RTS smoother which operates as off-line processing mode, a near-real-time stepwise smoother is implemented to eliminate the sharp corrections over the steps. The impact of the near real-time smoothing filter for different step manners (walk, run and climb stairs) combined with the Constant Velocity Update (CUPT) concept we proposed previously is illustrated and analysed. Experimental results show that the proposed method can dramatically improve pedestrian navigation smoothness