Description: It is shown that navigational systems of technical objects provide the generation of information based on the purpose of the object: processing of the navigation parameter vector, which is used later in motion control systems, information on the deviation of the object from a given point; information about the position of the object relative to other objects and the like. It is substantiated that the requirements to the accuracy of determining navigation parameters can differ significantly for different objects, determining the volume and complexity of navigation equipment. The purpose of the article is to develop a method for estimating the multistruc-tural error signal of inertial navigation systems. In the case of an inertial system, the algorithm works in the form of a system of dif-ferential equations for the vector of navigation parameters, includes in its composition the coordinates of the object and the compo-nents of the speed of its movement. For the error in the development of navigation parameters, the nonlinear differential equation is valid, where the components of the parameter vector are the accelerometer measurement errors by the accelerometer and the errors in their angular orientation. At the same time, the components of the error vector of navigation system sensors that are associated with the errors in the development of navigation parameters (their accumulation) in time are most dangerous. It is proposed to ex-tend the multistructural description of the error vector by taking into account the measurement ambiguity at consecutive discrete-time moments. One of the options for simplifying the problem is to use the Gaussian approximation of a posteriori density at each step of the discrete time. The proposed Kalman filter relations for each of the models give partial error estimates and their covari-ance. In this case, the formula for calculating the a posteriori probabilities of navigational system models forms a sub-optimal algo-rithm for filtering a multistructural signal. An important feature of the proposed algorithm is that particular estimates and covari-ance’s do not require separate memorization. They can be calculated successively (within one step of discrete time) by means of a reconstructed Kalman filter and averaged. The additional load of RAM in the realization of the pulsating filter in comparison with the Kalman filter is related only to the recalculation of a posteriori probabilities.
Keywords: vector of errors, inertial navigation systems, navigation parameters, structural analysis