TacFinRay: Soft Tactile Fin-Ray Finger with Indirect Tactile Sensing for Robust Grasping

TL;DR

TacFinRay introduces a soft Fin-Ray finger with a hinge-based interface that transfers contact deformation to a rigid sensing crossbeam, enabling indirect 2D sensing of contact location and indentation depth. A CNN analyzes binarized marker-pin patterns to predict depth and position, removing the need for full mechanical modeling. Design optimization shows longer pins, base pins, and opposing hinges yield the best sensing accuracy, with robust generalization across indenters and improved pick-and-place performance under uncertainty. The work demonstrates a lightweight, scalable tactile sensing solution for soft robotics, expanding the capability of tactile feedback beyond direct fingertip sensing.

Abstract

We present a tactile-sensorized Fin-Ray finger that enables simultaneous detection of contact location and indentation depth through an indirect sensing approach. A hinge mechanism is integrated between the soft Fin-Ray structure and a rigid sensing module, allowing deformation and translation information to be transferred to a bottom crossbeam upon which are an array of marker-tipped pins based on the biomimetic structure of the TacTip vision-based tactile sensor. Deformation patterns captured by an internal camera are processed using a convolutional neural network to infer contact conditions without directly sensing the finger surface. The finger design was optimized by varying pin configurations and hinge orientations, achieving 0.1\,mm depth and 2mm location-sensing accuracies. The perception demonstrated robust generalization to various indenter shapes and sizes, which was applied to a pick-and-place task under uncertain picking positions, where the tactile feedback significantly improved placement accuracy. Overall, this work provides a lightweight, flexible, and scalable tactile sensing solution suitable for soft robotic structures where the sensing needs situating away from the contact interface.

Figures (9)

• Figure 1: Proposed tactile Fin-Ray finger and experimental setup for validation. When physical contact occurs on the front surface, both the finger structure and the hinges undergo deformation. This deformation is represented in the marker-pin array on the bottom crossbeam. A camera, fixed at the base of the finger, captures the marker movement (inset images), which is processed as a binarized array for input to a contact location and depth prediction model.
• Figure 2: Key variables for effective sensing in the Fin-Ray finger and its rendered image. To enable effective sensing of the contact location and depth of an external object, we considered the following variables of the bottom crossbeam: (A) the length of the long pin, (B) the presence of bottom pins, and (C) the orientation of the two hinges. The sensing of location and depth is achieved through the translation and deformation of (D) the bottom crossbeam, while an internal camera of the Fin-Ray finger (E) captures the marker movements.
• Figure 3: The deep neural network architecture inputs a 1-channel, 128$\times$128 image using four sequential convolutional blocks. Each block consists of a 2D convolution (32 filters; kernel sizes 11, 9, 7, or 5), batch normalization, a ReLU activation, and max-pooling. The 8$\times$8$\times$32 feature map is then flattened and passed through two 512-neuron fully-connected regression layers.
• Figure 4: Analysis of the motion of the bottom beam in the Fin-Ray finger induced by the added hinges. Considering the $x$, $y$ coordinate system defined in (C), when a perpendicular force $F$ is applied to the finger at a location offset by $y$, the resulting deformation is a combination of two effects: (1) the translation of the gripper due to the hinge mechanism (A–C) and (2) the inherent deformation of the Fin-Ray finger (D, E). These combined effects ultimately lead to the deformed state illustrated in (F).
• Figure 5: An illustration of the translation and deformation of the bottom crossbeam in response to the indentation depth $x$ and contact location $y$ of an object in contact with the Fin-Ray finger, along with the corresponding experimentally measured tactile images, which have been binarized from the original data. To facilitate comparison of marker movements under the four conditions, a yellow box has been placed at a fixed reference position.