Pneumatic artificial muscles: model, design, fabrication, sensing and control strategies for electromagnetic risk applications.

• Ano de defesa: 2023
• Autor(a) principal: Scaff, William
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Biblioteca Digitais de Teses e Dissertações da USP
• Palavras-chave em Português: Controle ótimo; Magnetic resonance imaging compatibility; Optimization algorithms; Otimização não linear; Pneumatic artificial muscle; Position control; Reabilitação; Rehabilitation; Ressonância magnética
• Link de acesso: https://www.teses.usp.br/teses/disponiveis/3/3152/tde-25052023-081533/

Artificial muscles are materials or devices that changes shape with a stimulus. These biological inspired actuators are getting popular because of their advantages over conventional actuators, such as electric motors, hydraulic and pneumatic cylinders. Pneumatic artificial muscles, for example, have several advantages over conventional actuators, such as the compliance, actuation flexibility and high power-to-weight ratio, and also have the flexibility to be constructed without conductive and/or ferromagnetic materials. These characteristics makes artificial muscles suitable for many applications where conventional actuators cannot be used or have limited performance, as in high electric and/or magnetic field environments such as inside magnetic resonance imaging or explosion risk environments. However, pneumatic artificial muscles usage is limited because of the complexity of its implementation. Furthermore, designing and controlling a system actuated by artificial muscles have never been done with totally Magnetic Resonance Imaging compatible materials and sensors. To improve the applicability of pneumatic muscles, this thesis develops a methodology for designing, sensing and controlling devices for electromagnetic risk applications. And, to address the control problem, an optimal control approach is used, considering several optimization algorithms to tune the controller, in a simulated environment or in an experimental environment. In this way, parameter tuning can be customized to each specific application, translating its requirements to an objective function. A new optimization algorithm is proposed and used to tune the parameters of the controller, resulting in a 48.15% shorter learning time and a 8% improvement on parameter quality compared to Bayesian Optimization, a state-of-the-art stochastic optimization algorithm.