| Summary: | Positioning systems are used to determine position coordinates in navigation (air, land, and marine). Statistical analysis of their accuracy assumes that the position errors (latitude—<i>δ<sub>φ</sub></i> and longitude—<i>δ<sub>λ</sub></i>) are random and that their distributions are consistent with the normal distribution. However, in practice, these errors do not appear in a random way, since the position determination in navigation systems is done with an iterative method. It causes so-called “<i>position random walk</i>”, similar to the term “<i>random walk</i>” known from statistics. It results in the empirical distribution of <i>δ<sub>φ</sub></i> and <i>δ<sub>λ</sub></i> being inconsistent with the normal distribution, even for samples of up to several thousand measurements. This phenomenon results in a significant overestimation of the accuracy of position determination calculated from such a short series of measurements, causing these tests to lose their representativeness. This paper attempts to determine the length of a measurement session (number of measurements) that is representative of the positioning system. This will be a measurement session of such a length that the position error statistics (<i>δ<sub>φ</sub></i> and <i>δ<sub>λ</sub></i>) represented by the standard deviation values are close to the real values and the calculated mean values (<inline-formula><math display="inline"><semantics><mrow><mover accent="true"><mi>φ</mi><mo stretchy="true">¯</mo></mover></mrow></semantics></math></inline-formula> and <inline-formula><math display="inline"><semantics><mrow><mover accent="true"><mi>λ</mi><mo stretchy="true">¯</mo></mover></mrow></semantics></math></inline-formula>) are also close to the real values. Special attention will also be paid to the selection of an appropriate (statistically reliable) number of measurements to be tested statistically to verify the hypothesis that the <i>δ<sub>φ</sub></i> and <i>δ<sub>λ</sub></i> distributions are consistent with the normal distribution. Empirical measurement data are taken from different positioning systems: Global Positioning System (GPS) (168’286 fixes), Differential Global Positioning System (DGPS) (864’000 fixes), European Geostationary Navigation Overlay Service (EGNOS) (928’492 fixes), and Decca Navigator system (4052 fixes). The analyses showed that all researched positioning systems (GPS, DGPS, EGNOS and Decca Navigator) are characterized by the Position Random Walk (PRW), which resulted in that the empirical distribution of <i>δ<sub>φ</sub></i> and <i>δ<sub>λ</sub></i> being inconsistent with the normal distribution. The size of the PRW depends on the nominal accuracy of position determination by the system. It was found that measurement sessions consisting of 1000 fixes (for the GPS system) overestimate the accuracy analysis results by 109.1% and cannot be considered representative. Furthermore, when analyzing the results of long measurement campaigns (GPS and DGPS), it was found that the representative length of the measurement session differs for each positioning system and should be determined for each of them individually.
|
|---|
