Research On Positioning Method Of Pedestrian Navigation System Based On MEMS Inertial Sensors

Pedestrian Navigation System (PNS) is the device that can provide real-time, accurate,and reliable location information of the specific individuals such as rapid reaction force,firefighters, police officers, etc. Thus the PNS is important and can provides important technical support for these personnel, and enable them to successfully complete the relative tasks.The current PNS are generally relying on the GPS techniques. However, GPS signals can easily be blocked in the environments like indoors, underground buildings, urban canyons,etc.,thus resulting in the unreliable positioning information out puts. In addition,in some applications, in order to ensure the personnel's own security and prevent their exposures to others, the PNS can not send or receive signals using devices like GPS receivers, radio frequencies, etc. Therefore, the research on the self-contained PNS is necessary.Based on the features that the inertial sensors are less susceptible to external environments, not dependent on any external signals, that can be used in the enclosed space,in this dissertation,we designed a strapdown inertial navigation system (SINS) algorithm based self-pedestrian navigation system, using the on-the-shelf micro inertial measurement unit, namely, MIMU, which include the micro electronical mechanical system (MEMS)based gyroscope, MEMS accelerometer, MEMS magnetometer. The MIMU is fixed on a shoe of the pedestrian,and we call this shoe "navigation shoe".Given the fact that when the pedestrian walks, his/her feet alternate between the stationary stance phase and the moving stride phase, When detecting the stance phase of the "navigation shoe", the foot mounted PNS will apply the zero velocity updates (ZUPT), using the velocity error as pseudo-measurements into the predesigned Kalman Filter (KF) error corrector, which will correct the inertial sensor error, navigation error.Due to the inaccurate detection problem of the zero velocity during ZUPT, we propose a Neyman-Pearson (N-P) criterion based detection method. To achieve accurate zero velocity detecting results in the multi-motion mode of the "navigation shoes". We convert the threshold analysis problem into the mathematical modeling problem utilizing the maximum likelihood estimation under the N-P criterion.Because of the recursive and integrative nature of the ZUPT based KF procedure, the error covariance increases through each ZUPT period, which is followed by obvious decrease and large state estimate corrections. This will result in the undesirable discontinuities in the navigation trajectory at the end of each step. For solving this problem, eliminating the discontinuities of the trajectory, in this dissertation, we apply the Rauch-Tung-Striebel (RTS)smoothing algorithm to the navigation data of the ZUPT based PNS.The validity and feasibility of the method are demonstrated and confirmed through establish a real foot mounted PNS and doing ground experiments.