Features of the Data Modeling Method of the Composite Filter Algorithm

Methods for modeling input data for the filter of statistical processing of target motion parameters are considered. The inputs are measurements some coordinates of the phase vector and variance or standard deviations of these measurements. When debugging the filter in laboratory conditions, these the inputs are commonly referred to as the measurement model and the error model of those measurements. It is very important that these models are as close as possible to real measurements and their standard deviations. The closer the model is to real data, the more accurate the filter will work, although of course, no model can match the actual data perfectly. For this reason, always after testing on model data, the transition to real data entails some deterioration in the filter's performance. To this circumstance it is necessary to be ready and perhaps it makes sense in laboratory tests to lay the features of real work, when situations are possible when measurements do not come at all, and the standard deviations take on abnormally high values. In this article discusses measurement models in an additive form - an exact value plus a normally distributed random variable with zero mathematical expectation and some standard deviations. These random variables are independent as the coordinates of one vector and independent with respect to the coordinates of another vector one random process. Such a model of errors is due to the presence of many factors independently introducing an error (interference) and their summation. The article discusses the methodology for obtaining the root-mean-square measurement by the true coordinates and obtaining a measurement model based on these standard deviations. The standard deviation model for each the radar station can have its own, and therefore it is important to know which station will track the target. When designing an escort filter, it is important also test it with different standard deviation models. The modeling features of the input data are therefore very important in the development of statistical processing filters.