Posted in November 2013

Feedback mechanisms in RoboTutor

As part of the HART Seminar master students Jaime Alonso Lorenzo and Henrik Hannemose implemented feedback mechanisms into the RoboTutor software. This allows the robot to interact with its audience in a new way, enhanching the learning experience. A video of this extension of the RoboTutor system can be found below:


YouTube Preview Image

The Robotutor System

The Robotutor is an NAO robot from Aldebaran Robotics that has been equipped with special software that enables it to give a presentation to a group of people. The type of presentation is one that most people are familiar with: the robot uses a PowerPoint slideshow and accompanies the slides with spoken text. An important difference however is that the robot can’t (yet) come up with the content of such a presentation. Therefore the software makes use of scripts. These are simple text files which contain the text for the robot to speak and the moments when the next slide is to be shown. Additionally the script can include a number of special commands, which provide interactivity in the presentation. After all, a presenter who just rattles out his text doesn’t really communicate with his audience, whether the presenter is a robot or a human being.

An important element of the interaction is the motion of the robot. The NAO is a humanoid robot with many degrees of freedom, and the Robotutor software uses its capabilities gratefully. There are generally two types of movements the robot can perform. First, there are the ‘big’ movements, such as dancing or doing a soccer kick. This is a good way to show off what the robot can do, but it is really a separate part of the presentation. On the contrary, the second type of movements is supposed to be executed during the regular presentation. These are movements such as hand gestures and looking around the room. The robot is capable of doing this on command, but through the script it is also possible to enable automatic execution of these movements throughout the presentation.

Another important way the robot interacts with its audience is by asking questions. For this purpose the Robotutor software makes use of the Turningpoint system. This system consists of ‘response cards’ that are handed out to the audience. These cards are small electronic devices with a number of buttons. These cards can be used to answer multiple choice questions. In the presentation, the robot will show a slide with a question and asks the audience to respond. The results can then be shown on the screen and the robot will reply accordingly. These are not the only features of the Robotutor, many others are implemented and even more are in development. In fact, extensibility is an important quality of the system. The software is written to support additional commands in the future to enable new ways of interaction. This way the capacity of the Robotutor can be continuously expanded and, based on feedback, developed into a complete educational experience.

The RoboTutor Project

What comes to mind when you think about education? Chances are that it involves an experienced teacher talking to a number of students. For thousands of years, this has been the main way people have been taught intellectual skills. But there are recent developments that might change the way we learn things. Using modern technology, teachers are no longer limited to the people that are physically near. Over the internet a great teacher can reach millions. Innovative people such as Salman Kahn have pioneered in this field, but the educational establishment is not far behind. Many famous names take part in the Open Courseware project, such as MIT, Yale and our own TU Delft.

But while this method of distribution provides the great advantage of scale, it often provides little engagement. A famous piece of Confucian wisdom says: “What I hear I forget, what I do I remember”. This emphasizes how important it is that a student plays an active role during lessons. This requires a level of interaction that cannot always be provided through internet lectures.
A different innovation that could provide a high level of interaction is the robot teacher. Modern humanoid robots are flexible creatures and it is far from unthinkable that robots will someday fulfill the role of teacher in the classroom. The chances of success for the robot as a teacher depend largely on its capability to interact. Saying words is not difficult for a robot; however, telling a story to an audience is.
The Robotutor project explores the issue of robot-classroom interaction in a practical way. Its aim is to develop prototype software that enables a robot to give an engaging presentation to an audience. Based on feedback from the audience the software can be improved and observations can be made about the subject.
© 2011 TU Delft