Kinematic Control for Crossed-Fiber-Reinforced Soft Manipulator Using Sparse Bayesian Learning

Muscular antagonism caused by muscle fibers and muscle layers is vital for the hydrostatic skeleton of soft creatures. This article introduces a crossed-fiber arrangement and a hybrid material structure to a hydrostatic-skeleton-inspired soft manipulator that mimics the bending motion of worms. Deformation tests on radial expansion, torsion, and bending motion verify that the proposed design is controllable under large deformations to the manipulator. To achieve accurate control, this article proposes a sparse-Bayesian-learning-based piecewise constant curvature kinematic model and designs a feedback controller based on the kinematic model. The proposed controller demonstrates superior accuracy and reduces vibration compared with the linear least-squares-based controller and the neural-network-based controller in tracking desired movements during experiments.

Automated summary

This article introduces a new soft manipulator inspired by the hydrostatic skeleton of soft creatures, mimicking the bending motion of worms through a crossed-fiber arrangement and hybrid material structure that creates muscular antagonism. The proposed controller, based on a sparse Bayesian-learning kinematic model, demonstrates superior accuracy and reduced vibration compared to other controllers during experiments.