Designing of Robot Grippers with Dynamic Active-Pixel Vision Sensor (Davis)

Abstract

In robotics, slip-motion detection and control is a challenging problem, and one of the main factors influencing effective grasping and manipulation. Currently, the grippers that are capable of detecting slip use sensors such as tactile sensors and frame-based cameras, which induce several limitations in the system's performance related to response time, flexibility, robustness, and multimodality. Therefore, the idea of using Dynamic and Active-pixel Vision Sensor (DAVIS) camera for vision-based gripper is investigated in this thesis. DAVIS camera is an event- based camera that considers events caused during the contact of an object with the sensing surface in detecting the motion of the object being grasped. Such an event-based sensor renders images quickly (i.e., in micro-seconds) and also does not generate redundant data unlike the frame-based sensors. This thesis focuses on proposing a new gripper design for slip detecting application using the DAVIS camera along with the fingertip modeling based on object contact simulation (FEM). The task includes designing both the gripper body and the fingertip. In this work, the gripper body is designed and modeled based on mirror reflection and Periscope Principle (optical method), where one optical sensor (i.e., the DAVIS camera) is used to reduce the workspace. Next, a flexible tactile sensor is developed to be used as a fingertip for the new vision-based gripper. Event-based cameras compared to classical frame-based cameras, offer high temporal resolution, wide dynamic range, and low latency. Therefore, the proposed gripper system is ideal for real-time grasping. Furthermore, the developed vision-based gripper consists of an optical channel that allows the placement of the event-camera at the gripper's center. Such a camera placement widens the span of the gripper thereby increasing its workspace. To the best of our knowledge, this is the first work on developing a visionbased gripper system equipped with an event-camera together with an optical channel to facilitate the detection of tactile activity at the fingertip. The eventcamera used in the developed gripper system enables it to perform grasping operations in real-time. Such a gripper system is capable of detecting incipient slip, which makes it suitable for industrial applications like pick and place and sorting applications.

Date of Award	Oct 2020
Original language	American English