Electrical Impedance Tomographic Shape Sensing for Soft Robots
Wenci Xin, Fangmeng Zhu, Peiyi Wang, Zhexin Xie, Zhi Qiang Tang, Cecilia Laschi
Abstract
With infinite degrees of freedom, soft robots are expected to achieve dexterous and complex tasks, but this also puts forward higher requirements for their sensing capabilities. An important sensing task in soft robots is sensing their own deformation and current shape. Currently, most of the existing soft shaping sensors are limited by local perception abilities, stretchability, and fabrication difficulties. We propose a sensing method based on Electrical Impedance Tomography (EIT), which reconstructs conductivity patterns distributed on a surface, by considering the deformation-caused resistance changes. Comparison between the theoretical and experimental patterns reveals that even though the quality of the pattern is affected by a large amount of noise, the considered features are still able to reflect the change of shape. With the help of neural networks, the pattern is decoded to the physical data related to the deformation. Detection of the planar shape changes and proprioception of a sensor-integrated soft robot are presented to exhibit the capability of our method. Results show that the detected error ratios are mostly under 5% and 3% for 2D and 3D conditions respectively.