GelEvent—A Novel High-Speed Tactile Sensor With Event Camera

Abstract

Endowing tactile sensors with the ability to perceive contact status and contact force can enhance the precise perception and modeling of contact by robotic end effectors, thereby achieving more dexterous manipulation tasks. However, existing tactile sensors still fall short in rapidly responding to multimodal tactile information. In this work, we designed a novel tactile sensor called GelEvent, which is based on an event camera. It uses an event camera to collect contact information and is capable of determining contact status, estimating the area of the contact region, and the mechanical information of the contact surface online at an average speed of 180 Hz, with a latency of less than 5.5 ms. To address the issue that event cameras cannot capture the motion information of feature points in a stable state, we designed and fabricated a brand-new flexible contact structure. Using an event-based threshold judgment method, it achieves estimation of contact status and area, as well as continuous and stable mechanical output. Experiments have demonstrated that this sensor is capable of providing data on contact area, status of contact, and forces and torques (<inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$F_{x}$ </tex-math></inline-formula>, <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$F_{y}$ </tex-math></inline-formula>, <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$F_{z}$ </tex-math></inline-formula>, <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$T_{z}$ </tex-math></inline-formula>) between manipulated object and sensor at an average rate of 180 Hz during interactions. The estimation error of contact area is within ±9 mm2, the average accuracy of contact status judgment is 92%, the force estimation accuracy is approximately 0.80 N, and the torque estimation accuracy is about 8.3 N<inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\cdot $ </tex-math></inline-formula> mm. This sensor has great potential in promoting the dexterity and speed of robotic operations.