Algorithm for Finding a Descriptive Path for Delivering Cargo to Hard-to-Reach Areas

Abstract

At the moment, due to the growing interest in unmanned aerial vehicles (UAV), as well as due to the rapid and constant development of the UAV market and high technologies in the field of artificial intelligence, this branch of robotics has begun to occupy more and more sectors of the economy. With the increasing popularity of fast and light vehicles such as UAVs, the demand for efficient flight path planning algorithms has also increased to achieve a variety of goals: overflight, obstacle detection and collision avoidance. One of these goals is achieved by our algorithm for flying around static obstacles in 3D space in a sparse weighted graph. A hybrid method has been developed for finding the safest and fastest route based on Dijkstra's algorithm, which works with satellite images of various terrain classifications. The novelty of the proposed algorithm lies in the combination of artificial neural networks for categorizing terrain classes and using these data when constructing a flight task. The idea is to build a route on top of the safest terrain so that in the event of an emergency fall or landing, the UAV will not be lost. It is more memory efficient because it does not require storing all the vertices in an open list, as do algorithms that perform a similar task. In addition to the software architecture, the most suitable hardware architecture for the intended purpose - delivering cargo to difficult terrain - is presented.