Document Actions

You are here: Home Festtage Feiern Roboter … Do Robots Celebrate Christmas?Season’s …

Do Robots Celebrate Christmas?Season’s Greetings from Nao

With holiday trimmings and “Jingle Bells”: At the University of Freiburg even the robots are looking forward to the Christmas season. Junior Professor Dr. Maren Bennewitz is teaching the humanoid robot Nao to navigate through rooms and even how to play the xylophone.


factsheet2.png
profile.png
video.png
gallery.png

Aufmacher.jpg

Photo: Universität Freiburg 

factsheet2.png


Nao lifts the mallets and plays “Jingle Bells” from the first to the last note without making a single mistake. The student group of computer scientist Maren Bennewitz, junior professor at the University of Freiburg, needed about three weeks to teach the little humanoid robot the classic holiday tune. There are two cameras built into the musician’s face, which is connected to a computer over a wireless network. The computer converts the musical data for the Christmas tune, analyzing each note and its duration. Nao has to make one movement per note with its arm. A computer sends these commands, telling the robot how and in what tempo it needs to move to hit the right note on the xylophone, in the right rhythm and at the right angle.
 

  • Joints Are the Key to Flexible Movements

The idea for the Christmas video, explains the Freiburg researcher, came from her PhD candidate Daniel Maier. It was then realized by two students as part of their bachelor’s project. Nao’s dexterous handling of the instrument shows what flexible movements the robots designed at the University of Freiburg are capable of. Nao’s movements approximate those of a human being: The robot has 25 joints, enabling  to hold the xylophone mallets as well as to climb stairs and open cabinets.
 

  • The First Robot to Climb a Spiral Staircase

At the Collaborative Research Center “Spatial Cognition,” Bennewitz and her PhD candidate Armin Hornung are cooperating with researchers at the University of Bremen to study how humanoid robots can navigate in a multi-floor environment. They hope this will provide them insight into how humans and machines can function together in the same environment. Things like climbing spiral staircases and walking up and down ramps are no longer a problem for Nao: The built-in sensors provide images and laser data the robot can analyze to adapt its steps and body position to the curves of a spiral staircase. Nao still has difficulties climbing down high steps, however, as the robot tends to lose her balance when transferring the weight to the lower leg.

  • 3D Models for Navigating around Barriers

Robots are at a disadvantage in rooms: The machines don’t just have to perceive their environment but also recognize structures and use this information to interpret whether barriers are in the way. Healthy humans can do this effortlessly, whereas Nao first has to learn a 3D model of each room. On the basis of the height and shape of the furniture, the potential barriers, the robot literally calculates its way through the room.
 

  • Help from Robots for Humans in Daily Life

Bennewitz and her team have come another step closer toward their goal of making robots useful for humans: Nao can open doors to cabinets. This might not seem all that difficult at first, but a closer look at the necessary programming and the process of training the robot reveals how many calculations and motion sequences this involves. First, Nao has to determine how far from the cabinet it has to stand in order to open the door. Then the robot has to grasp onto the knob and choose the right speed to open the door. In addition, Nao has to bend her knee joints at a particular angle to maintain its balance. In comparison, the feat of playing Christmas tunes on the xylophone is a piece of cake for Nao.
 

  • Research Goals for Future Christmases

The Freiburg computer scientist still has big plans for the little robot, and she also sees potential for further research concerning her rendition of “Jingle Bells.” One of her goals is to get Nao to recognize the instrument with the help of a camera and position herself in front of it on her own. In addition, a built-in microphone will help Nao to determine whether it has hit a bar in the right way, enabling the robot to actively adapt and optimize its motions independently. After all, what the robot learns by playing can provide the research team with important insights for other areas.
 

Die Druckversion dieses Textes (pdf) finden Sie hier.

 

 Portait of the researcherprofile.png

maren1111-crop-small.jpg

Junior Professor Dr. Maren Bennewitz

Maren Bennewitz studied computer science and economics at the University of Bonn. In 2004 she completed her PhD under Prof. Dr. Wolfram Burgard in Freiburg with a dissertation on the topic “Mobile Robot Navigation in Dynamic Environments.” She then conducted research at the University of Freiburg in the area of robotics. Since October 2008 she has served in the position of junior professor and as head of the “Humanoid Robots Lab” at the Department of Computer Science of the University of Freiburg. In the new Cluster of Excellence BrainLinks-BrainTools she is collaborating with Wolfram Burgard on brain-machine interfaces, and she is also contributing to a research project involving motion analysis of Parkinson’s and epilepsy patients.
 

The idea for the “Jingle Bells” video came from her doctoral candidate Daniel Maier. The students Stefan Band and Jonas Delleske originally realized the idea as part of a bachelor’s project. The project is currently being continued by the computer science student Ramin Zohouri, who has taught the robot how to hit the bars more forcefully and is currently working on independent motion optimization for Nao.
 

Videovideo.png

 

 

 

 

Gallerygallery.png

This is what Nao lerns at the University of Freiburg

  cover-external-2.jpg   cover-octomap-2.png   door3.jpg
           
 

Inside rooms robotors are more challanged then humans: the machines not only have to sense their environment but they also have to recognise structures and to judge if obstacles are on their path.

  To navigate around obstacles Nao has to memorize a 3D model of the room. Using the hights and forms of objects hence of the possible obstacles, Nao calculates its path through the room.  The Robot Nao of the University of Freiburg is capable of opening a cupboard. This action needs complex underlying analysis and movements that Prof. Dr. Maren Brennewitz explains: It has to calculate in what distance it has to stop in front of the cupboard to open it. Then Nao has to grasp the handle and to use the right speed to open the door. At the same time it has to bend its knees in the correct angle to keep its balanced. 

 

^Top of the page


< Back