Deep Learning Approach for Unmanned Aerial Vehicle Landing

Abstract

It is one of the biggest challenges to land an unmanned aerial vehicle (UAV). Landing it by making its own decisions is almost impossible even if progress has been made in developing deep learning algorithms, which are doing a great job in the Artificial Intelligence sector. But these algorithms require a large amount of data to get optimum results. For a Type-I civilization collecting data while landing UAV on another planet is not feasible. But there is one hack all the required data can be collected by creating a simulation that is cost-effective, time-saving, and safe too. This is a small step toward making an Intelligent UAV that can make its own decisions while landing on a surface other than Earth's surface. Therefore, the simulation has been created inside gaming engine from which the required training data can be collected. And by using that training data, deep neural networks are trained. Also deployed those trained models into the simulation and checked their performance