Improving the efficiency of visual navigation of an unmanned aerial vehicle by choosing the most informative route

Abstract. The problem of estimating the coordinates of unmanned aerial vehicles (UAVs) using visual navigation in the absence of satellite navigation signals is considered. The UAV has a camera pointing towards the underlying surface to adjust its position using correlation algorithms. It is assumed that the UAV can choose one of several alternative route options as part of the implementation of the target task. The aim of the study is to increase the effectiveness of visual navigation by choosing a route with maximum information content. When assessing the information content, it is proposed to take into account the size and shape of the correction areas that can be observed during flight along the appropriate routes, based on the predicted density of the distribution of errors in measuring the coordinates of the UAV. During the flight, this estimate can be updated at any time when new information is received. An adaptation of the algorithm for calculating information content for the task under consideration is proposed. As an example, the flight of a UAV with a known model accumulation of an error in determining its coordinates, loaded with a table of landmarks and a correction system is considered. Model calculations show that the proposed approach can significantly increase the probability of correctly estimating the coordinates of the UAV compared to a random choice of route. It should be noted that the estimated informative value of the routes is the average predicted value. Specific implementations can produce results that differ significantly from the calculated averages, however, the proposed algorithm allows to adapt the assessment of informativeness using new information.