Neural Network Navigation System Time Model of the Sensors Errors Errors Interconnection Function Unobservation Model Adequacy Verification Neural Network Algorithm Increasing the Accuracy
Institute of Information Theories and Applications FOI ITHEA
Application of neural network algorithm for increasing the accuracy of navigation systems are showing.
Various navigation systems, where a couple of sensors are used in the same device in different positions and the
disturbances act equally on both sensors, the trained neural network can be advantageous for increasing the
accuracy of system. The neural algorithm had used for determination the interconnection between the sensors
errors in two channels to avoid the unobservation of navigation system. Representation of thermal error of two-
component navigation sensors by time model, which coefficients depend only on parameters of the device, its
orientations relative to disturbance vector allows to predict thermal errors change, measuring the current
temperature and having identified preliminary parameters of the model for the set position. These properties of
thermal model are used for training the neural network and compensation the errors of navigation system in non-
stationary thermal fields.