Exposing Robot Learning to Students in Augmented Reality Experience

This paper considers a learning process in which the student teaches the robot new tasks, such as lifting unknown weights, via reinforcement learning procedure. Using CAD software, we ran virtual trials using the robot's digital twin in place of physical robot trials. When performing the task, the robot measures and sends the value of the weight to an IoT controller implemented on the ThingWorx platform and receives parameters of the optimal posture found through the virtual trials. When we presented the robot learning process to high school students they had difficulty fully understanding the robot's dynamics and selection of posture parameters. To address this difficulty, we developed an augmented reality interface which allows students to visualize robot postures on the digital twin and monitor the change in parameters (such as the center of gravity) measured by virtual sensors. The student can select a weightlifting posture and control the robot to implement it.