Infrastructureless Pedestrian Positioning

Abstract

Many methods for pedestrian positioning exist. In outdoor environments, global satellite navigation systems such as GPS can give satisfactory positioning performance in many circumstances encountered by pedestrians. Pre-installed outdoor communication infrastructure, such as cellular networks or TV broadcast signals, can be leveraged for pedestrian uses. Specialized RF, ultrasound or light ranging beacons can also be installed indoors for positioning in spaces as small as individual rooms and networks of transponders can cover large installations. However, all these systems use transmitted signals that are subject to attenuation, blocking, reflection and diffraction effects, all of which can greatly reduce the accuracy and availability of range information. In contrast, Inertial Navigation Systems (INS) are "sourceless" in that they do not rely on any external transmitted signals. This explains their great utility in highend land, air, marine and space guidance, navigation and control systems, where dependingon external signals for aiding purposes might be impractical or risky. Unfortunately, for pedestrian navigation, unaided traditional INSs are of limited use. If the upper limit to the position error is set to some reasonable value, say a few metres after some 10s of minutes of self-contained navigation, either a very accurate navigation-grade INS or very frequent zero velocity updates (ZUPTs) with a tactical grade system are required. These realities, plus the fact that navigation-grade INSs will remain large, costly and power-hungry for at least another 10 years, means that traditional mechanization schemes for self-contained, personal navigation are currently impractical. The overall objective of this thesis is to investigate how low-grade, low-cost, and low-power INSs can be exploited for pedestrian positioning and in particular for first responder scenarios. To begin, a thorough bibliography of past research permits the identification of the relative merits of various technologies that have been proposed for emergency, rescue and military operations. Next, an extension to the well-studied occurrential pedestrian dead reckoning (PDR) technique using headgear-mounted motion sensors is described and good distance over ground (DoG) estimation performance is demonstrated. Since it is not a simple matter to apply occurrential techniques to a large class of locomotion patterns, the foot-inertial technique is then explored as an alternative. With an IMU (Inertial Measurement Unit) attached to (and in the future, mounted in) footwear, simplified strapdown inertial navigation techniques allow for omnidirectional motion patterns, very good DoG estimates, and vertical excursion characterization. Unfortunately, large heading jumps occur indoors, caused by magnetic disturbances and by the use of a generic orientation filter. It is shown how these heading errors can be modeled and then mitigated via map filtering techniques running over minimal a priori building geometry information.