A Multifaceted User Study for the Teaching-Learning-Prediction-Collaboration Framework in Human-Robot Collaborative Tasks

As the implementation of robotics systems in modern industries becomes more commonplace, the desire to streamline and simplify humans' interaction with them is highly needed. Human-robot collaboration frameworks have made strides towards the goal to facilitate shared tasks in human-robot teams. Such methods as Learning from Demonstration (LfD) show great potential in well performing collaborative tasks. To boost LfD's capacity, our previous study has developed a novel Teaching-Learning-Prediction-Collaboration (TLPC) framework for robots to learn from human demonstrations, customize their task strategies according to humans' personalized working preferences, predict human intentions, and assist humans in collaborative tasks. In this work, we conduct a multifaceted user study to evaluate it in real-world human-robot collaborative tasks. Participants of this user study are from diverse age groups with varying educational backgrounds and genders. Seven assessment metrics are developed to comprehensively evaluate the performance of TLPC through t-tests. A controlled human-robot collaborative experiment without TLPC is also conducted. This study seeks to observe and analyze the subjective feelings and feedback of the participants using TLPC when they perform collaborative tasks with a robot via periodic surveys given throughout the experiment. Our research outcomes help us gather insights into and create catalysts for the construction and optimization of human-robot interactive systems in advanced manufacturing contexts. They can be leveraged to improve human-robot collaboration quality, manufacturing productivity, human safety, and ergonomics.