Integration Of MEMS Sensors, WiFi, And Magnetic Features For Indoor Pedestrian Navigation With Consumer Portable Devices

Mobile location based services (LBS) is attracting the public attention due to their potential applications in a wide range of personalized services. Most LBS users spend 70%-90% of their time in indoor environments. Therefore, a demanding issue is to provide a trustworthy indoor navigation solution. This thesis provides a a continous and smooth navigation solution that has accuracy of 3-5 m (RMS) by using off-the-self sensors in consumer portable devices, local magnetic features, and existing WiFi infrastructures. The main novation points are:(1) It presents a real-time calibration method for gyro sensors in consumer portable devices. Through the use of multi-level constraints, this method happens automatically without the need for external equipment or user intervention, and reduced gyro biases from several deg/s to 0.15 deg/s indoors and 0.1 deg/s outdoors under natural human motions and in indoor environments with frequent magnetic interferences.(2) It introduces and evaluates two quality-control mechanisms for the integration of dead-reckoning (DR) and magnetic matching (MM), including a threshold-based method and an adaptive Kalman filter (AKF) based method. The DR/MM results were enhanced by 47.6%-67.9% and 43.9%-65.4% in two environments through the use of quality control.(3) It presents a profile-based WiFi fingerprinting algorithm by using the short-term trajectories from DR and geometrical relationships of various reference points (RPs) in the space. The Multi-Dimensional Dynamic Time Warping (MD-DTW) algorithm in introduced to match with inaccurate profile length for such a multi-dimensional system. The use of the profile-based approach reduced WiFi fingerprinting errors by 14.0%, and mitigated the WiFi mismatches when the user started navigation.(4) It proposes a WiFi-aided MM algorithm, which reduces both the mismatch rate and computational load. The WiFi-aided MM results were 70.8% and 74.5% more accurate than MM in two indoor environments, and 10.0% and 10.5% better than WiFi.(5) It provides designs for and evaluates two improved DR/WiFi/MM integration structures and corresponding quality-control mechanisms. Structure#1 utilizes the WiFi-aided MM algorithm, while Structure #2 uses the integrated DR/WiFi solutions to limit the MM search space. This mechanism in Structure #2 has at least one more level than those in previous DR/WiFi/MM structures. The difference between the DR/WiFi/MM Structure #2 results in two indoor environmrns were only 13%, and the difference between the DR/WiFi/MM Structure #2 results under four different motion conditions were only 16%.