GPS Signal Offset Detection and Noise Strength Estimation in a Parallel Kalman Filter Algorithm

摘要

Measurements from Global Positioning System (GPS) satellites are subject to corruption by signal interference and induced offsets. This thesis presents two independent algorithms to ensure the navigation system remains uncorrupted by these possible GPS failures. The first is a parameter estimation algorithm that estimates the measurement noise variance of each satellite. A redundant measurement differencing (RMD) technique provides direct observability of the differenced white measurement noise samples. The variance of the noise process is estimated and provided to the second algorithm, a parallel Kalman filter structure, which then adapts to changes in the real-world measurement noise strength. The parallel Kalman filter structure detects and isolates signal offsets in individual GPS satellites. The offset detection algorithm calculates test statistics on each of the filters and makes decisions on whether to remove satellites from the solution based on these statistics. The two algorithms contain several user-defined parameters that have significant effects when adjusted. The various effects of parameter variation are described and a parameter set is chosen at which to evaluate the algorithms. The combined algorithm performs quite well in computer simulations.