Coordination on Systems of Multiple UAVs
David Alejo Teissière · Dialnet (Universidad de la Rioja) · 2015
espanolEsta tesis trata acerca de metodos para coordinar las trayectorias de un sistema de Vehiculos Aereos no Tripulados y Autonomos (en adelante UAVs). El primer conjunto de tecnicas desarrolladas durante la tesis se agrupan dentro de las tecnicas de planificacion de trayectorias. En este caso, el objetivo es generar planes de vuelo para un conjunto de vehiculos coordinadamente de forma que no se produzcan colisiones entre ellos. Ademas, este tipo de tecnicas puede usarse para modificar el plan de vuelo de un subconjunto de UAVs en tiempo real. Entre los algoritmos desarrollados en la tesis podemos destacar la adaptacion de algoritmos evolutivos como los Algoritmos Geneticos y el Particle Swarm (Enjambre de Particulas), la incorporacion de nuevas formas de muestreo del espacio para la aplicacion del algoritmo Optimal Rapidly Exploring Random Trees (RRT*) en sistemas multi-UAV usando tecnicas de muestreo novedosas. Tambien se ha estudiado el comportamiento de parte de estos algoritmos en situaciones variables de incertidumbre del estado del sistema. En particular, se propone el uso del Filtro de Particulas para estimar la posicion relativa entre varios UAVs. Ademas, se estudia la aplicacion de metodos reactivos para la resolucion de colisiones en tiempo real. Esta tesis propone un nuevo algoritmo para la resolucion de colisiones entre multiples UAVs en presencia de obstaculos fijos llamado G-ORCA. Este algoritmo soluciona varios problemas que han surgido al aplicar el algoritmo ORCA en su variante 3D en sistemas compuestos por vehiculos reales. Su seguridad se ha demostrado tanto analiticamente, como empiricamente en pruebas con sistemas reales. De hecho, durante esta tesis numerosos experimentos en sistemas multi-UAV reales compuestos hasta por 4 UAVs han sido ejecutados. En dichos experimentos, se realiza una coordinacion autonoma de UAVs en las que se asegura la ejecucion de trayectorias libres de colisiones garantizando por tanto la seguridad del sistema. Una caracteristica resenable de esta tesis es que los algoritmos desarrollados han sido probados e integrados en sistemas mas complejos que son usados en aplicaciones reales. En primer lugar, se presenta un sistema para aumentar la duracion del vuelo de planeadores aprovechando las corrientes ascendentes de viento generadas por el calor (termicas). En segundo lugar, un sistema de deteccion y resolucion de colisiones coordinado para sistemas con multiples UAVs reactivo ha sido disenado, desarrollado y probado experimentalmente. Este sistema ha sido integrado dentro de un sistema automatico de construccion de estructuras mediante multiples UAVs. EnglishThe aim of this thesis is to propose methods to coordinately generate trajectories for a system of Autonomous Unmanned Aerial Vehicles (UAVs). The first set of proposed techniques developed in this thesis can be defined as trajectory planning techniques. In this case, the objective is to generate coordinated flight plans for a system of UAVs in such a way that no collision are produced among each pair of UAVs. Besides, these techniques can be applied online in order to modify the original flight plan whenever a potential collision is detected. Amongst the developed algorithms in this thesis we can highlight the adaptation of evolutionary algorithms such as Genetic Algorithms and Particle Swarm, and the application of Optimal Rapidly Exploring Random Trees (RRT*) algorithm into a system of several UAVs with novel sampling techniques. In addition, many of these techniques have been adapted in order to be applicable when only uncertain knowledge of the state of the system is available. In particular, the use of the Particle Filter is proposed in order to estimate the relative position between UAVs. The estimation of the position as well as the uncertainty related to this estimation are then taken into account in the conflict resolution system. All techniques proposed in this thesis have been validated by performing several simulated and real tests. For this purpose, a method for randomly generating a huge test batch is presented in chapter 3. This will allow to test the behavior of the proposed methods in a great variety of situations. During the thesis, several real experimentations with fleets composed by up to four UAVs are presented. In these experiments, the UAVs in the system are automatically coordinated in order to ensure collision-free trajectories and thus guarantee the safety of the system. The other main topic of this thesis is the application of reactive methods for real-time conflict resolution. This thesis proposes a novel algorithm for collision resolution amongst multiple UAVs in the presence of static obstacles, which has been called Generalized-Optimal Reciprocal Collision Avoidance (G-ORCA). This algorithm overcomes several issues that have been detected into the algorithm 3D-ORCA in real applications. A remarkable characteristic of this thesis is that the developed algorithms have been applied as a part of more complex systems. First, a coordinated system for flight endurance extension of gliding aircrafts by profiting the ascending wind is presented. Second, a reactive collision avoidance block has been designed, developed and tested experimentally based in the aforementioned G-ORCA algorithm. This block has been integrated into a system for assembly construction with multiple UAVs.