The Development of an Articulated Robot Arm and and its Effectiveness in Performing Household Dishwashing Tasks

Abstract

Dishwashing is a vital household task, necessary for maintaining hygiene and a clean kitchen environment. However, constant dishwashing is repetitive, time-consuming and in the long term can lead to symptoms of arthritis. This project aims to test, develop and evaluate different model architectures to determine which is the most effective at classifying cups and plates. This study also aims to develop an articulated robotic arm to pick up cups and plates, perform a washing motion under the sink, and place them in a designated drop-off zone. A CNN model with 3 convolutional layers, a ResNet18 model, a Resnet50 model and a VGG16 model were tested and compared based on accuracy, precision, F1-score, recall and inference time to determine which model is the most suitable to my classification model. Then, the robotic arm used this inference model to test the pick up, washing motion and drop off success rates on 25 trials for each dish object. The robotic arm successfully picked up cups 72% of the time and completed the washing phase 60% of the time, whereas plate handling success was significantly lower, with only 40% pickup and 12% washing success. These results suggest that the articulated robotic arm is more effective with cylindrical objects than flat objects with larger surface areas, highlighting the need for a new gripper design and new handling strategies for plates.