Machine Learning for Unmanned Aerial Vehicle Routing on Rough Terrain

Abstract

The paper considers the main methods of machine learning for unmanned aerial vehicle (drone) routing, simulates an environment for testing the flight of a drone, as well as a model with a neural network for the unmanned routing of a drone on rough terrain. The potential use of unmanned aerial vehicles is limited because today the control of drone flight is carried out in a semi-automatic mode on the operator's commands, or in remote mode using a control panel. Such a system is unstable to the human factor because it depends entirely on the operator. The relevance of the work is to use machine learning methods for drone routing, which will provide stable control of the unmanned aerial vehicle to perform a specific task. As a result of the work, a neural network architecture was developed, which was successfully implemented in a test model for routing an unmanned aerial vehicle on rough terrain. The test results showed that the unmanned aerial vehicle successfully avoids obstacles in the new environment.