Personal navigators (PN) have been studied for about a decade in different fields and applications, such as safety and rescue operations, security and emergency services, and police and military applications. The common goal of all these applications is to provide precise and reliable position, velocity, and heading information of each individual in various environments. In the PN system developed in this dissertation, the underlying assumption is that the system does not require pre-existing infrastructure to enable pedestrian navigation. To facilitate this capability, a multisensor system concept, based on the Global Positioning System (GPS), inertial navigation, barometer, magnetometer, and a human pedometry model has been developed. An important aspect of this design is to use the human body as navigation sensor to facilitate Dead Reckoning (DR) navigation in GPS-challenged environments. The system is designed predominantly for outdoor environments, where occasional loss of GPS lock may happen; however, testing and performance demonstration have been extended to indoor environments.;DR navigation is based on a relative-measurement approach, with the key idea of integrating the incremental motion information in the form of step direction (SD) and step length (SL) over time. The foundation of the intelligent navigation system concept proposed here rests in exploiting the human locomotion pattern, as well as change of locomotion in varying environments. In this context, the term intelligent navigation represents the transition from the conventional point-to-point DR to dynamic navigation using the knowledge about the mechanism of the moving person. This approach increasingly relies on integrating knowledge-based systems (KBS) and artificial intelligence (AI) methodologies, including artificial neural networks (ANN) and fuzzy logic (FL).;In addition, a general framework of the quality control for the real-time validation of the DR processing is proposed, based on a two-stage Kalman Filter approach. The performance comparison of the algorithm based on different field and simulated datasets, with varying levels of sensor errors, showed that 90 per cent success rate was achieved in detection of outliers for SL and 80 per cent for SD. The SL is predicted for both KBS-based ANN and FL approaches with an average accumulated error of 2 per cent, observed for the total distance traveled, which is generally an improvement over most of the existing pedometry systems.;The target accuracy of the system is +/-(3-5)m CEP50 (circular error, probable 50%). This dissertation provides a performance analysis in the outdoor and indoor environments for different operators. Another objective of this dissertation is to test the system's navigation limitation in DR mode in terms of time and trajectory length in order to determine the upper limit of indoor operations. It was determined that for more than four indoor loops, where the user walked 261m in about 6.5 minutes, the DR performance met the required accuracy specifications. However, these results are only relevant to the existing data. Future studies should consider more comprehensive performance analysis for longer trajectories in challenging environments and possible extension to image-based navigation to expand the indoor capability of the system.
展开▼
