Large-Scale Vision-Based Tactile Sensing for Robot Links: Design, Modeling, and Evaluation

Abstract

The sense of touch allows individuals to physically interact with and better perceive their environment. Touch is even more crucial for robots, as robots equipped with thorough tactile sensation can more safely interact with their surroundings, including humans. This article describes a recently developed large-scale tactile sensing system for a robotic link, called TacLINK, which can be assembled to form a whole-body tactile sensing robot arm. The proposed system is an elongated structure comprising a rigid transparent bone covered by continuous artificial soft skin. The soft skin of TacLINK not only provides tactile force feedback but can change its form and stiffness by inflation at low pressure. Upon contact with the surrounding environment, TacLINK perceives tactile information through the three-dimensional (3-D) deformation of its skin, resulting from the tracking of an array of markers on its inner wall by a stereo camera located at both ends of the transparent bone. A finite element model (FEM) was formulated to describe the relationship between applied forces and the displacements of markers, allowing detailed tactile information, including contact geometry and distribution of applied forces, to be derived simultaneously, regardless of the number of contacts. TacLINK is scalable in size, durable in operation, and low in cost, as well as being a high-performance system, that can be widely exploited in the design of robotic arms, prosthetic arms, and humanoid robots, etc. This article presents the design, modeling, calibration, implementation, and evaluation of the system.

IoTouch: whole-body tactile sensing technology toward the tele-touch

“…This relation deserves more investigation in the near future, therefore, it is not within the scope of this paper. Actually, a similar elaboration was shown in our previous paper (see [7]) on the analytical method based on an FE model.…”

“…The objective sensing system in this paper is different to the TacLINK (cylindrical shape) [7], but resemble to our first version (barrel shape) mentioned in [8]. The core design includes two fish-eye cameras (sampling rate of 30 fps) located at two ends of a transparent link (acrylic pipe), which work as stereo camera for tracking white markers (dots) on the inner wall of a soft skin.…”

“…It was considered challenging to implement in a larger scale. In our previous research [7], an elongated cylindrical robotic link (called TacLINK) was proposed for accomplishment of whole-arm tactile sensing. We utilized two co-axial cameras at two ends of the link for tracking markers on a cylindrical soft skin, in which 3-D (three-dimensional) movements of markers could be obtained.…”

“…Then, by synthesizing 3-D movement of trackers and the stiffness matrix of soft skin, the information of applied force can be extracted in real-time. However, this analytical technique in [7] faces several problem. First, the force synthesizing method requires large burden of computation, resulting in low sample rate (about 15 Hz w.r.t 30 frameper-second (fps) of the camera).…”

“…In this paper, we introduced IoTouch, an artificial intelligence (AI)-based tactile sensing approach on the previously proposed vision-based sensing system. The aim of this research is to increase the sampling rate of the system compared with the analytical approach in [7], and reduce the dependence in computation subjected to scalability. We showcased the approach on a barrel-shape tactile sensor, which is considered challenging for implementation of large-scale tactile sensing, thanks to its inherited geometrical shape.…”