MATHEMATICAL METHODS OF COMPLEX PROCESSING OF RTC NAVIGATION DATA

Abstract

Nowadays, the navigation systems of robot-technical complexes (RTC) use heterogeneous sensors
of primary information, which can provide redundancy of navigation data. This allows to increase
the accuracy of calculation of motion parameters, as well as allows to determine them with greater
reliability in case of failure of one or more sensors. The paper gives a review and classification of lowlevel
mathematical methods of processing overridden state parameters of RTC navigation systems. It is
noted that the problem of combining is a subfield of the problem of system identification and therefore
has common approaches to the construction of the solution. In the vast majority of methods based on the
optimization approach, the quadratic error function is used as the optimality criterion. All mathematical
methods of combining (complex processing or fusion) any data are divided into low-, medium- and
high-level methods. In navigation systems, low-level methods such as recursive, nonrecursive, and covariance-
based methods are the most used. Non-recursive methods are rarely used directly. Recursive
ones are usually constructed using a Kalman filter scheme. Recursive ones, as a rule, are constructed
according to the Kalman filter scheme. Not all methods are robust to non-Gaussianity and correlation
dependence of the original data, which is often encountered in navigation systems with overdetermined
data. In addition, not all methods can be used to address the relevance of data from navigation instruments.
It is noted that the key for combining methods is the approach of fusion data in an information
space, understood as the inverse of covariance, since the vast majority of methods, including Bayesian
methods, are reducible to it. In this regard, covariance-based methods are of most interest. However, for solving the problem of data relevance in navigation, the existing methods are poorly suited to the problem
of data relevance because they require computationally intensive optimization problem solving at
each step, and navigation systems are real-time systems. Thus, there is a problem of developing new
approaches to solve this problem