LEO Aiding for a Barometric Map-Matching Pedestrian Dead Reckoning Solution in a GNSS Inhibited Environment

Abstract

The current pedestrian navigation space has become increasingly reliant on Global Nav- igation Satellite Systems (GNSS) as their sole form of navigation. However, GNSS signals are quite vulnerable to intentional and accidental interference. When a pedestrian loses their only navigation solution the can be left in very confusing and potentially dangerous scenarios. Thus, there is a growing need to develop pedestrian navigation systems that operate entirely independent of GNSS information. One way to circumvent the need for GNSS navigation is to used a localized inertial navigation system (INS). Specifically, a step detection approach to pedestrian dead reckoning (PDR) allows for the propagation of a pedestrian’s position solution using only the measurements of an inertial measurement unit (IMU). In this algorithm, angular velocity and specific force measurements are used to calculate a heading and step length which are transformed into a position over time. However, the noise inherent to IMU measurements creates a drift on this PDR position estimate. Depending on the magnitude of noise in the system, this drift can significantly degrade the quality of the navigation solution. This thesis presents a series of sensor fusion algorithms that reduce the errors in the PDR system. To start, low Earth orbit (LEO) satellite information is used to estimate the errors on the calculated heading measurement. In this methodology, LEO Doppler shift measurements are used to derive the pedestrian’s course. By comparing this course to the heading output by IMU measurements, a heading bias is estimated when at least two Doppler shift measurements are available. This heading bias estimation technique is tested and evaluated for both a static and dynamic heading bias. The other sensor fusion algorithm presented in this thesis is a novel implementation of a barometric elevation map-matching algorithm. This map-matching system uses the differential between a barometer measurement and an elevation pulled from a digital elevation map (DEM) to constrain position errors. Traditionally, this algorithm has experienced computational issues on low size, weight, and power (SWaP) hardware due to the chosen esti- mation framework and DEM. The barometric elevation map-matching algorithm in this thesis, however, is evaluated with a variety of nonlinear estimation filters and open source DEMs to maximize the system’s computational efficiency while maintaining position accuracy. Addi- tionally, this map-matching algorithm does not estimate a heading bias which helps to alleviate some of the algorithm’s computational burden. Each of these two sensor fusion algorithms are individually evaluated using a series of Monte Carlo analyses in a simulated environment. The results of these studies validate the sensor fusion algorithms for a variety of pedestrian navigation systems. These methodologies are then combined into a singular system that aims to constrain the errors on the PDR position output. This combined sensor fusion algorithm is evaluated in both a simulated and more realistic testing environment. In both the simulated and pseudo-live data testing scenarios, the combined sensor fusion algorithm is shown to reduce the position error to within only a few meters of error. An accurate position solution for a pedestrian is determined without the use of any GNSS information.