Аннотация:Serious requirements for the accuracy of strapdown inertial navigation systems (INS) are set now. In particular, the navigation errors in determining the latitude and longitude should not exceed 1 mile (1852 meters) in the first hour of a flight. These requirements are applied to the INS developed by PAO Moscow Institute of Electromechanics & Automatics (MIEA). The tolerance corridor uniformly expanded at a rate of 1 mile per hour on each subsequent hour of a flight. The tolerance on INS velocity errors is 1 m/s during the flight. One of the criteria for INS quality functioning is the time during which the errors in position and velocity exceed the set limits. In addition, one widely spread statistical criterion is used for the conclusion when INS satisfies accuracy conditions. In a series of flights for some time samples, for example, at the end of each 10-minute from the beginning of the flight statistical parameters are calculated: mean and standard deviation (M,) of INS positioning error for latitude and longitude. The |M|+2 values in meters are calculated for these time samples. The boundaries of tolerance for INS coordinate error are computed by the above method. It is assumed that if in a plurality of flights the magnitude |M|+2 at some point falls out of the tolerance corridor, the criteria requirements are not met. It is obvious that mentioned above statistical criterion assumes that INS error in Latitude and longitude are distributed according to a normal (Gaussian) law at fixed time points from the beginning of a flight. If this condition is not met, a situation is possible when INS errors in position and velocity do not exceed the limit values, but the test requirements are not met. This leaves open the question of how many flights must be considered in order to develop a statistical criteria.This paper presents the results of testing the hypothesis of normal distribution of INS error in latitude and longitude for one hour flight of highly maneuverable aircraft. Analysis is based on data derived from INS flight tests. We consider data of 170 test flights done in 2015-2016. Histograms (geometric image of the empirical density distribution function) were computed. They were compared with the graph of the probability density of the normal law. So the visual inspection of the hypothesis was done. We compared also the graphics of the distribution functions for errors of latitude and longitude.The results of testing the hypothesis of normal distribution for the position error after one hour of flight are presented. We use the Kolmogorov-Smirnov goodness of fit test and Lilliefors criterion.Test was done using a standard software package used for engineering calculations. The features of the application of criteria for the small sample sizes are pointed out. It is shown that the hypothesis of normal distribution of INS positioning error after one hour flight was not confirmed in any of the tests.The results of the study show that the statistical test requirements to assess the quality of INS navigation solutions are not well-founded. So for INS installed on maneuverable object, on the basis of a different approach one should formulate more logical requirements for the estimate of accuracy of navigation solutions. As such a criteria we propose the ratio computation between the sum of time intervals during which the system is out of tolerance, to the total duration of the flight.
