Design of a low-cost robotic manipulator for food handling

Abstract

Industry 4.0 improved production processes for many manufacturing processes however, this change was not as pronounced in many service industries including the food service industry. Considering the high cost of implementing technologies such as robotics, edge computing and machine learning there are not many options for establishments such as restaurants and fast-food places to also implement various parts of Industry 4.0 into their services. With the growing consumer demand especially regarding dietary differences, it has become more important for food service establishments to make their processes smarter to keep up with the demand. In this project, a low-cost, computer vision-controlled robotic manipulator is implemented to assemble a custom plate of food without human assistance. This was achieved by collecting and annotating 1000 images as well as training an object detection model with no less than 90% average precision for each of the possible classes as well as an average recall of 94.45% where there were no more than 10 objects in the image frame. A 3D printed robotic manipulator was then assembled with servo motors, a Raspberry Pi, and a low-cost web camera and the model, together with python scripts were then used to make real-time detections while picking and dropping the food items from one plate to another within the manipulator's workspace.