Intelligent extraction of the informative features for UAV motion modelling: principles and techniques

The aim of the article is to develop a methodology for intelligent extraction of informative features of unmanned aerial vehicle (UAV) motion modelling using a Variational Autoencoder. Practical use of the developed methodology will allow to extract the most informative features of UAV motion modelling in the process of intelligent flight path planning. This fact will further reduce the complexity of calculations on their analysis and processing using the selected bases.In order to achieve this goal, a practical set of input data of the Variational Autoencoder is generated and investigations are carried out for examples of UAV velocity variation and manoeuvring. The autoencoder is used to reduce the dimensionality of the data to 3 indices, which will further reduce the complexity of the calculations to analyse and process them using the chosen bases. In the context of UAV motion modelling, this allowed up to 30% reduction in the volume of calculations.