VTacArm. A Vision-based Tactile Sensing Augmented Robotic Arm with Application to Human-robot Interaction

Abstract

Endowing artificial sense of touch comparable to human's has been challenging, yet significant to enabling adaptive and collaborative interaction in contact-rich tasks. This work is dedicated to proposing a novel vision-based tactile sensor augmented robot arm (VTacArm) design with full surface coverage and developing algorithms to retrieve contact information which is essential for down-stream feedback control. We first introduce the robot arm design and its accompanying fabrication process. Then, to convert the contact signals in the image space to the arm coordinate system, a calibration procedure and method are proposed. Finally, the tactile robot arm and the contact information extraction algorithm are integrated into a control system for collaborative interaction tasks. Bumping detection/reaction and contact motion following experiments are presented to justify that the designed tactile robot arm and proposed contact sensing method are beneficial and give robot capabilities to adapt to human contacts, which is vital for workers' safety. Our work can be informative for developing novel full-body vision-based tactile sensing on robots as a new concept with significantly lower cost and manageable fabrication complexity.