Automatic Shape Control of Deformable Wires based on Model-Free Visual Servoing

Romain Lagneau (1) , Alexandre Krupa (1) , Maud Marchal (1)

1 RAINBOW - Sensor-based and interactive robotics

In this paper, we propose a novel approach to automatically control the 3D shape of deformable wires using robots. Our approach proposes a novel visual feature along with a novel shape servoing method to enable dual arm manipulation of deformable wires. The visual feature relies on a geometric B-spline model and the use of Sequential Importance Resampling (SIR) particle filtering to track the 3D deformed shape of a wire over time. The shape servoing method is an adaptive model-free method that iteratively updates the deformation Jacobian matrix using weighted least-squares minimization with sliding window and an eigenvalue-based confidence criterion. We performed several experiments on wires with different mechanical properties. The results show that our approach succeeded to control the 3D shape of various wires for many different desired deformations, while working at an interactive time. It has also been shown that the shape servoing method can be used to handle large deformations by subdividing the task in successive intermediary targets to reach. These promising results pave the way for automatic control of the 3D shapes of deformable wires in many fields such as catheter insertion in medicine or wire manipulation in industry.

Domaines: Robotique [cs.RO]

Romain Lagneau, Alexandre Krupa, Maud Marchal. Automatic Shape Control of Deformable Wires based on Model-Free Visual Servoing. IEEE Robotics and Automation Letters, 2020, 5 (4), pp.5252 - 5259. ⟨10.1109/LRA.2020.3007114⟩. ⟨hal-02890830⟩