Pedestrian Navigation System And Multiple Systems Integration

Numerous solutions/methods to solve the existing problems of pedestrian navigation have been proposed in the last decade by both industrial and academic researchers.However,to date,there are still major challenges for a pedestrian navigation system(PNS)to operate continuously,robustly,and seamlessly in all indoor and outdoor environments.Since the Global Navigation Satellite System(GNSS)is reliable under most of the outdoor environments,in this thesis,novel methods for pedestrian indoor navigation applications through integrating information from different navigation techniques and systems are proposed.First,a PNS architecture based on a single device is proposed,which integrates the inertial navigation system(INS)mechanization,the pedestrian dead reckoning(PDR)mechanization,and the WiFi fingerprinting positioning method.In order to combine the merits of both the INS and the PDR,a PDR-aided INS(PDR/INS)algorithm is proposed.For fast and accurate generation of the WiFi fingerprinting database,a novel walking-survey method is given.Aiming at solving the WiFi ambiguity problem during the WiFi fingerprinting online navigation process,an improved K-nearest neighbors(KNN)method is presented.In order to make full use of the WiFi fingerprinting algorithm and the PDR/INS algorithm,a novel WiFi/PDR/INS approach is proposed for indoor navigation applications.The experimental evaluation shows that combined WiFi/PDR/INS algorithm not only tracks closest to the actual pedestrian walking trajectories but also has the navigation results with good continuity.Then,when multiple PNSs are used simultaneously by a specific user,novel information fusion methods for multiple PNSs integration to enhance the performance of each individual PNS are proposed.A nonlinear inequality distance constraint between any two PNSs is mathematically formulated.A novel filtering technique named Kalman filter(KF)with state constraint is used to explore such a constraint information,further diminishing the positioning errors of each PNS.Two different approaches based on the state-constrained KF for solving the multiple PNSs integration problems are proposed.The first approach incorporates a soft constraint into a normal KF to ensure that the state estimate almost satisfies the constraints rather than strictly satisfies the constraints;the second approach is based on solving a Quadratic Programming(QP)problem which incorporates a hard constraint into a KF to ensure that the state estimate should strictly satisfy the constraints.Simulation studies have been conducted on the integration of two and three different INS-based PNSs.The simulationresults show that both approaches can well bound the state estimates errors compared with the unconstrained state estimates.However,when a constraint's nonlinearity level is getting stronger,the performances of both approaches will decrease,especially for the soft constraint approach.Several real experiments with different navigation architectures and combinations are conducted to assess the proposed approaches for integrating multiple PNSs.The experimental results clearly indicate that through the proposed methodologies,using motion sensor data from multiple mobile devices could provide more accurate navigation solutions for a pedestrian in all indoor and outdoor environments.