Comprehensive Correction Of Errors In Inertial Sensor Based Pedestrian Navigation System

With the precisionimprovement of the micro electromechanical system(MEMS)based inertial devices,it becomes possible to develop accurate long term pedestrian navigation systems based on such devices.Due to their low cost,portability and strong anti-interference performance,systems of this type are suitable for applications in complex environment especially in cases where the satellite signals are not available.However,due to the drift and random noise of the inertial sensors,navigation errors will accumulate during recursive calculation and often need to be corrected with methods such as the zero velocity updatingstrategy.Nevertheless,this method cannot reduce the position error effectively because of the residual motion in the static phase and also the lack of yaw angle observation.To resolve this problem,wefocus ourselves in this thesis in developing comprehensive error correction methodsfor MEMS inertial sensor based pedestrian navigation system.The main work and contributions are as follows:Firstly,in traditional zero velocity update method,due to the lack of position error observation,the inertial pedestrian system cannot estimate the position drift in the static phase accurately,which usually leads to the divergence of the recursive calculation.To tackle such a problem,a novel static phase position drift detection method is developed.In this method,the motion between adjacent static points in the same static phase is first calculated and then regarded as an observationfor position error estimation.Secondly,to tackle the divergence problem due to the weakness in the observation for the yaw angle in the zero velocity update method,a novel magnetic disturbance detection and yaw angle construction method is developed.In this method,the disturbance is detected with the magnitude of the magnetic field intensity and the inner product between the magnetic field and the specific force.Meanwhile,we combine the disturbance detection and the straight trajectory detection strategies to provide more accurate yaw angle error observation.Finally,different from the Kalman filter in typical zero velocity update,we propose to integrate together not only observations of errors in the velocity,but also that of attitude and the position drift in one Kalman filter and amend them simultaneously.Then the estimated error in position,velocity and attitude are feedback to the inertial navigation algorithm,and the estimated sensor bias is utilized to correct sensor output.Extensive experiments in outdoor,indoor straight line,indoor square and staircase circumstance results show that this method is able to reduce the accumulated errors in the inertial pedestrian navigation extremely.