JISE


  [ 1 ] [ 2 ] [ 3 ] [ 4 ] [ 5 ] [ 6 ] [ 7 ] [ 8 ] [ 9 ] [ 10 ] [ 11 ] [ 12 ] [ 13 ] [ 14 ] [ 15 ] [ 16 ]


Journal of Information Science and Engineering, Vol. 37 No. 5, pp. 1165-1176


The Use of Deep Reinforcement Learning for Flying a Drone


SANDRO DOMITRAN AND MARINA BAGIĆ BABAC
Faculty of Electrical Engineering and Computing
University of Zagreb
Zagreb, HR-10000 Croatia
E-mail: {sandro.domitran; marina.bagic}@fer.hr


Nowadays, unmanned aerial vehicles, commonly known as drones, are used for many different purposes. However, it is still a challenging task to fly a drone, which limits its potential for doing more useful things. This paper shows how to design, develop and test an ML-Agent simulation environment by using the Unity engine and the deep reinforcement learning algorithm. First, the drone model needs to be imported in a simulating environment where it should have an ability to fly, and then it should be made to fly using deep reinforcement learning. In addition, the drone can learn to perform a certain task to elaborate the benefits of this approach.


Keywords: unmanned aerial vehicle, flight control, VTOL Tailsitter UAV, machine learning, deep reinforcement learning

  Retrieve PDF document (JISE_202105_12.pdf)