Estratégias de guiagem e cooperação de robôs aéreos sujeitos a restrições nas entradas e/ou nos estados

• Ano de defesa: 2013
• Autor(a) principal: Tales Argolo Jesus
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais; UFMG
• Palavras-chave em Português: Engenharia Elétrica; Robótica; Engenharia elétrica
• Link de acesso: http://hdl.handle.net/1843/BUBD-9GAHBW

In this thesis the problem of multiagent systems coordination with emphasis on fixed-wing Unmanned Aerial Vehicles (UAVs) systems is addressed. Two descentralized coordination strategies are proposed. In the first strategy the task to be accomplished is divided in three stages: (i) each UAV has to track a circle parallel to plane xy in a given altitude (different altitudes for different UAVs); (ii) the UAVs must distribute themselves in such a way that their projection onto plane xy are evenly spaced around a circle; (iii) each UAV converge to the same altitude to track the same circle. The main contribution of this work is in the mathematical proofs for the proposed strategy for a given UAV model. Minimum and maximum limits in the forward speed command, maximum limit in the absolute value of the yaw rate, maximum limit in the absolute value of the altitude speed, second-order dynamics, and a non-holonomic constraint are considered. Stability and collision avoidance guarantees throughout all the task accomplishment are provided, including the stage transitions, without violating the constraints. Then, a control law inspired in the artifical vector field based approach developed in the first stage of the first strategy is proposed. Such a design takes into account the presence of model uncertainty in the task of making an UAV track a circle. Mathematical guarantees that the robot is able to converge to a bounded region of known size that contains the target curve are given. In the second strategy, the problem of consensus for double-integrator dynamics with velocity constraints and constant group reference velocity is addressed. In comparison to previous works, this approach is different in three aspects: (i) the control law of an agent does not depend on the local neighbors velocities or accelerations, but only on the neighbors positions and on the own velocity; (ii) the constraints are non-symmetric; (iii) the velocity is constrained but not the acceleration. A decentralized control strategy with the neighboring topology described by an undirected connected graph is proposed. Mathematical guarantees of convergence without violating the constraints are given. Such an approach can be applied to the problem of making n VANTs, each one flying in a straight line and distributed in parallel and adjacent lanes, span a rectangular area with constant velocity and paired side by side, without violating the velocity constraints. All the results presented in this thesis are illustrated by means of numerical simulations.