M.Sc Thesis


M.Sc StudentMailhot Nadav
SubjectConstellation Design for Regional Space-Borne Geolocation
DepartmentDepartment of Aerospace Engineering
Supervisor PROF. Pinchas Gurfil
Full Thesis textFull thesis text - English Version


Abstract

Time difference of arrival-based geolocation allows for the position estimation of a terrestrial emitter by measuring its signals with 3 or more receivers. This type of geolocation, also known as a hyperbolic fix, is used by a variety of satellite and aircraft systems worldwide. Existing systems do not provide continuous coverage of a target area, instead relying on infrequent passes of satellites or keeping aircraft in flight for extended periods of time. In this study, we design constellations of satellites which can provide continuous geolocation services in a specific target region.

The problem of time difference of arrival-based geolocation is described in detail, and several estimation algorithms are proposed for its solution. These estimators include the extended Kalman filter, and two versions thereof which incorporate the Earth surface constraint. The lower bounds of the estimation errors are developed in the form of the Cramér-Rao lower bound and the position dilution of precision.

The constellations are then designed using global optimization techniques of exhaustive search and genetic algorithms with the aim of minimizing the position dilution of precision, which is a variant of the Cramér-Rao lower bound. The constellations resulting from the initial optimization are then analyzed in terms of the main constellation cost drivers - the number of orbital planes and the number of satellites. The regional coverage of the constellations is also tested, as the initial optimization does not take into account the entire region of interest.

The design methodology and the efficacy of the estimation algorithms are demonstrated with Monte Carlo simulations of geolocation with one of the resulting constellations. The constellation is shown to provide consistent coverage with a low estimation error, and the estimator is shown to be efficient.