P. Kielland et T. Tubman, ON ESTIMATING MAP MODEL ERRORS AND GPS POSITION ERRORS - (APPLYING MORE SCIENCE TO THE ART OF NAVIGATION), The International hydrographic review, 71(2), 1994, pp. 47-67
In order to decide whether a desired manoeuver can or cannot be safely
undertaken, a prudent navigator must be aware of both the current spa
tial uncertainty of his vehicle's positioning system and the spatial u
ncertainty of the navigational map model being used to depict the thea
tre of operations. From this safety to navigation perspective, knowled
ge of data accuracy is as important as the data itself. This paper dis
cusses the Electronic Chart (EC) implications of both GPS vehicle posi
tioning errors and the relatively large data modeling errors specific
to bathymetric map models (charts). It proposes and demonstrates softw
are solutions which statistically evaluate both of these spatial uncer
tainties and graphically integrates the two stochastic models within a
n EC environment. The paper also documents to an experiment carried ou
t by the Canadian Hydrographic Service, designed to insure that real-t
ime DGPS users compute statistically valid position error estimates. T
he experiment ground truthed the position error estimates obtained usi
ng a conventional real-time error analysis of pseudo-range redundancy.
Using this ground-truth information, an improved pseudo-range error m
odel was empirically determined. The new pseudorange error model is co
ntinually updated using the estimated pseudo-range variances computed
by the Novatel GPS receiver rather than applying the constant a priori
pseudo-range variance typical in least-squares adjustments. This dyna
mic range error model effectively reduced the statistical bias between
the observed errors and their predicted error estimates. The improved
range error model also significantly improved the performance of the
position solution. All DGPS positions computed by the modified softwar
e had a positional accuracy of better than 0.5 metres.