The Design and Development of Robotic Arms to Assist in Manufacturing Processes Using Machine Vision Techniques

Abstract

This article presents the design, development, and prototyping of a dual-arm robot equipped with a camera for visual perception. The dual-arm robot is designed to replicate human-like activities, including object sorting, picking up items, examining objects for decision-making, and performing other tasks typically reliant on human decision-making processes. The study employs the camera-equipped dual-arm robot to autonomously detect objects with varying positions and orientations. The system can select objects based on predefined models or specific characteristics and examine object quality. This research includes two main components: the creation of two independent 6-axis articulated robot arms using 3D printing and the development of a perception program in Python. This program utilizes machine vision techniques and image processing with Convolutional Neural Network (CNN) and the OpenCV library for the robot's response in sorting tasks, based on the position and direction of the object. These techniques determine the coordinates of contours, which are crucial for object detection. Control is facilitated through a Graphical User Interface (GUI) on the Raspberry Pi 4 microprocessor board and Arduino microcontroller board via I2C communication, managing the stepper motors' rotation degree and direction for each joint of both robotic arms. Based on the research findings, it is recommended to use both robotic arms for gripping workpieces with tolerance coordinates. For the x, y, and z positions, a deviation of 3±1.22 centimeters is recommended, with a mean range of –1 to 1.125 centimeters for the x, y, and z axes. For orientation angles, a compensation value of –7±11.5 degrees is suggested, with a mean range of –3.65 to 0 degrees during gripping.