Busca avançada
Ano de início
Entree


Dynamic modeling of a novel catamaran robotic system actuated by bio-inspired propulsion devices using physics-informed neural networks

Texto completo
Autor(es):
Savio, Thiago Liquita ; Geronel, Renan Sanches ; da Silva, Maira Martins
Número total de Autores: 3
Tipo de documento: Artigo Científico
Fonte: OCEAN ENGINEERING; v. 326, p. 11-pg., 2025-03-10.
Resumo

Industrial inspections or environmental awareness impose a demand for versatile and energy-efficient aquatic robots. This work proposes a novel boat robotic system composed of a catamaran actuated by bio-inspired tail- actuated propulsion devices. A catamaran is a type of boat featuring two parallel hulls of equal size. The combined hulls of a catamaran typically yield lower hydrodynamic resistance than comparable monohulls, requiring less power from the propulsion system. The propulsion system comprises two devices capable of mimicking body/caudal fin (BCF) based locomotion patterns. Macro-Fiber Composite (MPC) pairs actuate these devices via sinusoidal inputs. Servomotors can modify the orientation of the thrust forces generated by these devices, guaranteeing maneuverability to the robotic system. The catamaran robotic system can be modeled using Kirchhoff equations, but identifying some model parameters is challenging due to the scarcity of experimental data. A physics-informed neural network (PINN) is trained to tackle this challenge using the Kirchhoff equations and experimental data in the loss functions. A 3-fold cross-validation procedure is exploited, and six model parameters are identified. A good agreement between experimental data and the PINN's output of the PINN has been achieved. This device and its model can be further explored in autonomous tasks. (AU)

Processo FAPESP: 22/07119-2 - Melhorando a manobrabilidade de robôs macios aquáticos bioinspirados por meio de mecanismos multi-estáveis
Beneficiário:Maira Martins da Silva
Modalidade de apoio: Auxílio à Pesquisa - Regular
Processo FAPESP: 23/07834-6 - Gêmeos Digitais: uma Abordagem Integrada para a Modelagem e Controle de Manipuladores Flexíveis
Beneficiário:Renan Sanches Geronel
Modalidade de apoio: Bolsas no Brasil - Pós-Doutorado