Algorithms And It’s Realization Of Indoor 3D Localization With Multi-Sensor Information Fusion

With quick development and popularization applications of navigation and location technology, Location based services and applications under complex indoor environment has become a research focus in field of navigation. Considering indoor 2D positioning can not to meet demand, this thesis systematically studies algorithms of indoor 3D location based on multi-sensor information fusion, including sensors errors calibration and compensation, attitude solution of inertial sensors, algorithms of pedestrian dead reckoning, barometric altimeter etc. Based on the theoretical research, the applications of sensors data acquisition has been developed with Android platform, and 3D positioning program has been achieved using Fortran, which has been tested and analyzed. The main research contents of this thesis are summarized as follow:(1) Error characteristics of sensors which are mounted on experiment platform are analyzed. According to the existed error sources, the quasi-linear vector module method for the error calibration of MEMS inertial sensors is presented based on existing sensor error mathematics model. The calibration parameters are estimated with least square method, and experiment results show this method is verified after calibration parameters was substitute into the model.(2) Pedestrians walking gait in the process of the model is researched, a hybrid detection method which is made up of dynamic window, peak detection, zero crossing is presented based on existing methods. And the projection variable of three axis acceleration in the vertical direction is analyzed. In addition, test indicates that this method can improve accuracy of step detection.(3) This article summarizes the current step length estimation model and analyzes the relationship between different acceleration characteristics variables and step length during walking. Based on this, step length estimation model of stride frequency and acceleration’s minimum is put forward. Two kinds of estimation model experiments show step length estimation precision of this model.(4) Aiming at the existing problem of pedestrians walking in the process of heading estimation, based on inertial sensor attitude algorithm theory and quaternions kalman filter, attitude updating algorithm is put forward in this article. Experimental results show that this algorithm can improve the precision of direction estimation effectively.(5) Two of the factors affecting the pressure measurement precision are analyzed in comparison, results shows that influence of temperature and gravitational acceleration to height measurement precision can ignored; Considering that elevation can’t get in the indoor environment, relative height can be calculated based on absolutely barometric pressure height measurement. Experimental results show that calculation accuracy of relative height can reach the dm level, which meets the request to judge number of floor levels exactly.(6) Based on EKF, the indoor three-dimensional localization algorithm is put forward which fuses the PDR and height measurement with atmospheric pressure. Experiment’s result show that the indoor three-dimensional localization algorithm can restrain the drift caused by error accumulation of PDR algorithm and improve the accuracy of indoor 3D positioning.