Detection and Resistance of the Outlying Observation in GNSS/INS Integrated Navigation System
Li K · Hydrographic Surveying and Charting · 2015
Outliers exist almost inevitably in GNSS positioning results under adverse conditions,such as tunnels,canyons,and those with multipath effect. These outliers will violate the assumption about the Gaussianity of the observation noise,and hence will severely degrade the performance of the Kalman filter for GNSS/INS integration. Robust Kalman filter is employed in this paper to detect and resist observation outliers. The χ2test is performed for the innovation vector. If the null hypothesis is to be rejected,outliers are deemed to be present.Then a scaling factor,which can be calculated in closed form,is introduced to inflate the covariance matrix of the innovation vector,and hence the impact of the outliers is effectively restrained. Simulation results validate the efficacy of the algorithm.