Multiple hypothesis positioning algorithm for robust GPS-denied navigation

Abstract

In the past few decades, GPS has become the dominant source of precision navigation and is often required for many modern systems to operate. However, recent exposure of GPS vulnerabilities have called into question its overall resiliency and shown necessity for robust alternatives. Precision celestial navigation using Draper's Skymark technique can be used to replace GPS. However, these systems rely on prior position knowledge for system initialization. In GPS-denied scenarios, prior position knowledge may not be available or trustworthy. Similarly, other GPS-denied navigation techniques such as landmark navigation or vision-aided navigation can be difficult when there is limited prior position information. Therefore, the Multiple Hypothesis Positioning algorithm is developed in this thesis to provide robust positioning in GPS-denied navigation scenarios where little or no prior position knowledge is available.

The proposed robust positioning algorithm makes use of Multiple Hypothesis Tracking techniques to develop an object identification and observer positioning framework. The Multiple Hypothesis Positioning framework is developed broadly in this thesis to encompass multiple applications of the proposed algorithm. The Multiple Hypothesis Positioning framework is applied to two separate applications including a Lost-at-Sea positioning algorithm and a Lost-in-a-Forest positioning algorithm. The Lost-at-Sea application serves as an initialization process for Draper's Skymark technique in situations where no prior position knowledge is available. The Lost-in-a-Forest positioning algorithm uses pattern matching techniques to identify trees near an observer and compare these locally observed trees to a global map of all tree locations. The pattern matching techniques are combined with the Multiple Hypothesis Positioning framework to determine the observer's global position.

Both applications were tested in robust Monte Carlo simulations with positive results. Overall, the proposed Multiple Hypothesis Positioning algorithm and framework prove effective tools for robust positioning in GPS-denied navigation applications where prior position information is unavailable.