Distributed approaches to multi-robot curve tracking with collision avoidance

• Ano de defesa: 2020
• Autor(a) principal: Gabriel Viana Pacheco
• Orientador(a): Luciano Cunha de Araújo Pimenta
• Banca de defesa: Antonio Ferramosca, Alexandre Santos Brandão, Vinicius Mariano Gonçalves
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Universidade Federal de Minas Gerais
• Programa de Pós-Graduação: Programa de Pós-Graduação em Engenharia Elétrica
• Departamento: ENG - DEPARTAMENTO DE ENGENHARIA ELÉTRICA
• País: Brasil
• Palavras-chave em Português: Vector fields; Backstepping control; Distributed predictive control; Collision avoidance; Multi-robot systems
• Link de acesso: http://hdl.handle.net/1843/38496

This work deals with the problem of convergence and circulation of closed curves by multi-robot systems, through vector field based strategies. The problem is addressed with two regards. In the first part, the problem of circumnavigating a time-varying curve in the three-dimensional space with a group of quadrotor aerial vehicles is approached. For that end, a cascaded system is proposed, in which the high-level, distributed, layer guarantees convergence and circulation of the aimed curve through a vector field based control law. The same control layer provides collision avoidance among robots by considering predefined priority laws that take into account the robots positions, which allow to modulate the vector field. The high-level control law generates a trajectory to be followed by a quadrotor vehicle, which is guided by the low-level controller. The design of the low-level controller is based on the backstepping nonlinear control, improved with an integral action and an additional control law based on the Lyapunov redesign approach, which provides robustness against bounded disturbances. In the second part of this work, the problem of convergence and circulation of curves is addressed under another perspective. Considering curves defined in the 2-D space, the objective is to design a distributed predictive control strategy in which the control laws based on vector fields are embedded in the optimization problem. Therefore, instead of finding a control sequence, the problem finds the parameters of each robot’s control law. From this, the optimum control problem is designed to ensure the convergence and circulation of the target curve and avoid inter-robot collisions. Then, the same problem is distributed by the alternating directions method of multipliers, which allows the negotiation of trajectories between neighboring robots. The effectiveness of the control strategies proposed in this work are evaluated with simulation results.