Robot Perception for Motion Control of Robotic Arm

How, Choan Xian (2024) Robot Perception for Motion Control of Robotic Arm. Final Year Project (Bachelor), Tunku Abdul Rahman University of Management and Technology.

Text HOW CHOAN XIAN_Full Text.pdf Restricted to Registered users only Download (5MB)

Abstract

As human-robot collaboration is gaining more and more popularity in the industry, there is an urgent need to improve human safety and human-robot interactive experience under such mode of collaboration. In this project, a framework for integrating facial expression recognition (FER) system into the control of robots is developed. This project involves the identification of face detection algorithm and the facial expression recognition (FER) algorithm based on both categorical and dimensional emotion model and develop a framework for the robot to respond to emotions using the algorithms identified. Experiments are conducted to identify the most suitable algorithm, where an algorithm with high accuracy and low processing time is more desired. From the experiments conducted, it is identified that SSD would be the face detection algorithm used for the frameworks, whereas the FER algorithm for the categorical and dimensional frameworks are DeepFace Emotion Model and Edu’s Model respectively. The FER system of the frameworks are developed based on the identified algorithm. The robot’s speed is identified as the parameter of the robot to be modified based on the emotion detected. The computing of the robot speed based on the emotion for the categorical framework is achieved through a lookup table, whereas a fuzzy inference system is implemented for the same purpose in the dimensional framework. Simulation was carried out to validate the functionality and performance of the framework before implementing it onto an industrial robotic arm for testing. Based on the testing carried out, the dimensional framework is profound as it is able to provide a smoother change and better adaptability of the robot’s behaviour towards the emotion as compared to the categorical framework. The performances of the proposed frameworks are further validated by implementing them onto the Kuka K4 AGILUS R600 industrial robot, and the experimental results are found to be align with the simulation results. Through testing, it is found out that the response time of the robot is 0.56s.