Global Navigation Satellite Systems (GNSS) are widely used for most navigation applications. However, GNSS quality and availability suffer greatly in certain environments, such as urban canyons, or indoors due to signal blockage. This thesis investigates estimation algorithms to integrate data from multiple low-cost MEMS sensors in a personal navigation system to bridge those signal gaps.
MEMS-based accelerometer, gyroscope, magnetometer, and barometer sensor technologies are surveyed in depth. The main sensor design parameters and their connection to navigation performance are presented. Furthermore, this thesis presents a way of decomposing the sensor error terms then applying proper stochastic and deterministic error models. Subsequently, navigation estimation states and online calibration methods are elaborated.
Several key sensors-based positioning algorithms are explored. First, a nine-axis fusion engine of accelerometers, gyroscopes, and magnetometers is formulated into an attitude Kalman filter for orientation determination. Then a Pedestrian Dead Reckoning (PDR) algorithm is developed based on the accelerometer’s step detection and stride length estimation with the heading determined from the attitude fusion filter. In addition, Wi-Fi positioning is investigated for indoor environments based on received signal strengths. Finally altitude integration of the barometer and GPS height measurements is introduced to improve vertical position accuracy.
The complete navigation system is constructed using an Extended Kalman Filter (EKF) to perform the data fusion from multiple positioning above. This thesis also introduced the observability analysis for quantitative analysis about the degree of observability of each estimated state in EKF.
Field tests are presented to verify the system and developed algorithms using three different portable navigation prototypes. The first prototype explores optimal integration of the PDR and GPS for a continuous positioning solution. The second prototype is focused on Wi-Fi assistance when GPS is not available in deep indoor environments. The third prototype is a more compact form factor design that mimics the smartphone experience in real-life applications. The results show that the prototype systems can effectively deal with short GPS signal outages using EKF. Thus this thesis shows a cost effective design for a mobile, reliable and accurate system that enables continuous navigation anywhere.