الفهرس | Only 14 pages are availabe for public view |
Abstract The demand for navigation systems is rapidly increasing, especially in GNSS-denied environments. The ubiquitous use of smart mobile devices equipped with various sensors encouraged many researchers to investigate their use in improving indoor navigation, where GNSS is not available. Dead reckoning (DR) application depends on measuring the traveled distance and the orientation from the known position to the unknown one. Inertia navigation sensors installed in mobile devices are normally low-cost and drift significantly. Consequently, there is a need for auxiliary systems to aid the navigation process, which can be achieved using external sensors or additional information extracted from, for example, base maps. This thesis represents a new technique for distance estimation during dead reckoning navigation depending on the proximity sensor. The new technique is based on the proximity sensor for step counting in case of pedestrian dead reckoning (PDR) or cycle counting in cycle dead reckoning (CDR). The new technique has accuracy in distance measurement equal to 1.34% and 0.53% in PDR and CDR, respectively. As for the orientation (heading), the common technique is the sensor fusion concept through the use of gyroscope rate by integration with accelerometer and magnetometer data through Extended Kalman Filter (EKF). But this thesis is targeted towards improving the accuracy of heading estimation for pedestrian navigation in GNSS-denied environments by innovatively using the maps. The map directions were used in dead reckoning to improve the low-accuracy directions derived from portable device sensors. This method is significantly computationally efficient compared to traditional geospatial map-matching algorithms. The new approach replaces the conventional geospatial database with a list of street directions and paths used as Map Heading Constraints (MHC) when navigating in straight directions. The applied algorithm improved the navigation solution with an average positional error of 1.23%, where the drift had been reduced with a percentage exceeding 80%. |