Design and hybrid control of a two-axis flexure-based positioning system

Petri, Elena; Hao, Guangbo; Kavanagh, R.C.

doi:10.1007/s41315-021-00162-7

Cited by 1 publication

(1 citation statement)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Flexible mechatronics (incorporating compliant elements to eliminate relative motions between rigid elements) plays an important role in soft robotics, which has a spectrum of applications ranging from traditional engineering designs (Guo et al 2012;Wang and Lan 2014;Petri et al 2021) to emerging biomedical applications (Thomas et al 2021). The advantages of flexible mechatronics are under-exploited for soft robotics, however, because their designs involve complicated forward and inverse models.…”

Section: Introductionmentioning

confidence: 99%

Physics informed neural network for parameter identification and boundary force estimation of compliant and biomechanical systems

Lee

2021

Int J Intell Robot Appl

View full text Add to dashboard Cite

This paper presents the framework of a physics-informed neural network (PINN) with a boundary condition-embedded approximation function (BCAF) for solving common problems encountered in flexible mechatronics and soft robotics; both forward and inverse problems are considered. Unlike conventional PINNs that minimize a lumped loss function including the errors contributed by the initial or boundary conditions (ICs or BCs), the BCAF-PINN completely satisfies the ICs and/ or BCs while minimizing a loss function for parameter identification and boundary force estimation, overcoming a common erroneous-convergence problem due to unbalanced gradients in training a PINN. The formulation and implementation of a BCAF-PINN are illustrated with three practical applications, including a nonlinear system where solutions are available for numerical verification and for comparing with conventional PINNs, and a biomechanical system where a BCAF-PINN uses multiple cycles of natural foot flexion to identify its dynamic parameters experimentally. While overcoming several problems associated with traditional studies based on perturbation models with certain level of muscle contraction, the damping ratio identified by the BCAF-PINN indicates that the ankle joint is an overdamped system during flexion, consistent with that observed in published experiments.

show abstract

Section: Introductionmentioning

confidence: 99%

Physics informed neural network for parameter identification and boundary force estimation of compliant and biomechanical systems

Lee

2021

Int J Intell Robot Appl

View full text Add to dashboard Cite

show abstract

Design and hybrid control of a two-axis flexure-based positioning system

Cited by 1 publication

References 26 publications

Physics informed neural network for parameter identification and boundary force estimation of compliant and biomechanical systems

Physics informed neural network for parameter identification and boundary force estimation of compliant and biomechanical systems

Contact Info

Product

Resources

About