Técnicas de inteligencia artificial aplicadas al sector de las aeronaves pilotadas por control remoto

Resumo

The main objective of this Doctoral Thesis is to study the use of the techniques for the control of heterogeneous swarms of Remotely Piloted Aircraft (RPA) or Unmanned Aerial Vehicles (UAV), colloquially known as drones. This thesis is supported by three scientific publications indexed in the Journal Citation Report system. One of them is the study of the application of these and other techniques in the field of UAV swarms. The remaining two propose models for their application in simulated maps without obstacles and with fixed obstacles. The importance of the study of these techniques for UAV swarm control demonstrates that using a heterogeneous group of UAVs with full freedom of movement allows tasks to be performed faster than using only one. In addition, Reinforcement Learning techniques prove that they are able to adapt to the environmental situation and its obstacles. Reinforcement Learning is a set of Artificial Intelligence techniques that seek to solve certain types of tasks based on interaction with an environment. All this is done based on the reward or reinforcement caused by performing different actions in that environment. Thus, if an action is correct, the reinforcement is positive and, if it is incorrect, the reinforcement is negative. By being able to use a single system to control UAVs, the need for one operator per aircraft is reduced, reducing the costs associated with the operation. To improve the capability of these techniques, Artificial Neural Networks have been used for their ability to extract knowledge from patterns. Thus, it is possible to improve the adaptability of the proposed models to the different environments in which they are tested.