Robotic Arm with Obstacle Detection Designed for Assistive Applications

Abstract

With the growing popularity in manufacturing, medical or aerospace applications, Robotic Arm machinery is a growing industry and field of focus. In this paper, a low-cost robotic arm, HiWonder xArm 2.0, provides a proof-of-concept assistive system for use towards individuals with visual impairment. The system takes the camera and ultrasonic sensor input to collect objects of interest while maintaining user safety, automatically detecting obstacles or humans nearby and providing a requested object as close to a human as possible. This system correctly identifies and retrieves the objects based on the YOLOv4-tiny (You only look Once) deep-learning object detection network, Intel's pyrealsense2 library for the Intel RealSense D435i camera, and four HC-S04 ultrasonic sensors connected to an Arduino Uno.