Powering a Six-Axis Robotic Arm with AI to Collect Plastic Bottles

The robotic arms industry is rapidly growing, with robotic arms being used actively in various industrial sectors to complete tasks such as pick-and-place, material handling, inspection, palletizing, and more. Most robotic arms are pre-programmed to perform repetitive tasks in a pre-determined movement path. However, this usually means that the robotic arm cannot account for different abnormalities that may be present in its application, such as a missing item in a pick-and-place application. An engineer has to program the robot to minimize the chances of errors. In this paper, we describe the process of powering a six-axis robotic arm with artificial intelligence to detect plastic bottles, determine the distance between the arm and bottle, and pick and place the bottle. The robotic arm is programmed in C++ while the image processing is done in Python programming language. Pytorch is used as the deep-learning framework and the YOLOv5 model is used for object detection and distance judgement. An intel realsense D435i depth camera is used for better depth judgment. All the code is run on a Raspberry Pi 4 connected to the robotic arm. With this setup, the robot is successfully able to detect plastic bottles up to a distance of 400 mm and retrieve them. A similar setup can be implemented in factories to aid robotic arms in pick-and-place, palletizing, material handling, and other applications.