Planejamento de caminhos suaves e seguros baseados em espuma probabilística para Sistemas Robóticos Autônomos

• Ano de defesa: 2021
• Autor(a) principal: Nascimento, Luís Bruno Pereira do
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal do Rio Grande do Norte; Brasil; UFRN; PROGRAMA DE PÓS-GRADUAÇÃO EM ENGENHARIA ELÉTRICA E DE COMPUTAÇÃO
• Palavras-chave em Português: Espuma probabilística; Planejamento de caminho; Robótica assistiva; Segurança do caminho; Suavização do caminho
• Link de acesso: https://repositorio.ufrn.br/handle/123456789/44894

Planning a path for a robot to navigate between two points in a given environment and avoiding collision with obstacles is one of the main issues for autonomous robotics. The search for short paths and reduced search time are aspects of most planning methods, but in applications in which the robots interact directly with human beings, such as assistive robotics, ensuring a greater degree of safety associated with the movements is an essential requirement. In this context, this thesis presents a set of new strategies for the planning of safe paths for autonomous robots. The developed methods are essentially based on the concepts of the probabilistic foam method (PFM). PFM is a sampling-based path planning method capable of generating paths bounded by a set of connected bubbles, which guarantees a volumetric region in the free space for safe maneuverability. In order to compute the bubbles, PFM requires an explicit representation of the obstacle region in the configuration space, which is computationally impracticable considering its application for most problems. Thus, a novel strategy to compute bubbles without representing these obstacles regions is introduced. Besides, we present an analysis to reduce the number of PFM parameters. To improve the quality of the paths, two optimization procedures were implemented to reduce the path length and increase the path smoothness, maintaining a high clearance. Novel PFM-based techniques are developed to explore different mechanisms for the propagation of the foam, aiming to ensure the planning of shorter paths with a short search time, also guaranteeing paths with high clearance. In order to evidence the main contributions of this thesis, some simulated experiments are performed considering the path planning for two assistive robots: The first one is a lower limb exoskeleton, which aim at overcoming obstacles, walking up and down a stair, resulting in smooth movements with a more anthropomorphic pattern. These results illustrate the ability of PFM to plan safe and smooth paths for open kinematic-chain robots without the explicit representation of the obstacle region in configuration space. Smart Walker is the second robot considered in this work, and it was possible to represent the planning of safe paths for a mobile robot with differential drive, in addition to show some aspects of the paths planned by the new introduced methods.