Geração de trajetórias para veículos aéreos autônomos não-tripulados

Abstract

In this work a study on the problem of the trajectory generation for unmanned aerial vehicles is carried through. The main objective is to provide tools of path planning for aerial robots, taking into account some of its main physical constraints of movement. For this, some of the more used techniques for motion planning of grounded robots are argued initially, whose the navigation space is bidimensional. Two techniques in special (Dubins' Path and the Pythagorean Hodograph) are analyzed in details, once these take in consideration the main studied kinematic constraints in this work. The main focus of this text is the trajectory generation in the three-dimensional space, and therefore, some of the techniques more recently used for this end are also analyzed. Two new boardings are proposals in this work. The first one constitutes an extension of the Dubins' optimal path for the 3D space. The second promotes the unification of the two techniques previously cited for the 2D case, aiming at to produce curves in the space that are realizable for a specific air vehicle. It describes still in this work, the implementation of a Hardware-in-the-Loop system, used for the accomplishment of tests with intention to validate the methodologies proposals. This system uses a flight simulator as virtual platform for the study of the control and planning modules of the real autonomous air vehicle. Such modules are implemented in an embedded computer, that in turn is connected to the flight simulator via net interface. Thus, a (mathematician) aerodynamic model of a virtual vehicle is used as aircraft of tests for the navigation and trajectory planning tasks in the embedded hardware. Other tests still are carried through using the mathematical model of a real aerial robot developed by the Universidade Federal de Minas Gerais.