Activities

No upcoming events

Intelligent Systems Lab Project: ASB - Autonomous Silent Butler

Participants

• Manuel Baum
• Philipp Dresselhaus
• Lukas Kettenbach
• Phillip Lücking
• Patrick Renner
• Alexander Wecker

Supervisors

• Andreas Kipp
• Florian Lier
• Ingo Lütkebohle
• Frederick Siepmann
• Torben Töniges

Motivation

• Nowadays robots are making their way into everday households, e.g., vacuum cleaners, lawn-mowers, etc.
• Interaction with robots should not be disturbing in any kind of way. Noises, dirt or smells are not appreciated by the user.
• Robots are very complex systems which is why our goal was to customize the Robot Operating System - ROS . It features lots of stable and flexible software tools for robotics.
• Our aim: Building a silent service robot with limited ressources and time using ROS.

Application Szenario

• A group of people meets to watch a movie on their couch.
• None of them is willing to stand up to get drinks/snacks for himself or the group though all of them are desperate for such kind of pleasure.
• However, they recently got themselves a SilentButler which is capable of getting the job done.
• They wave at the robot which is now heading towards the couch and is offering its services.
• Items having EAN-codes printed on them may now be held in front of the robots camera.
• After successfully identifying the object the user may choose to reorder the item or just get rid of it.

Objectives

The project goals are

• Assemble robot-base (Pioneer 2DX) and self-designed body (including WeeTab, Kinect, HD-Webcam, etc.).
• Install and configure ROS for autonomous navigation, 3D vision, inter-process communication and state management.
• Use gesture recognition and non-audible feedback for silent communication.
• Utilize EAN-Barcode scanner for product recognition (via webcam).

Description

The automomous silent butler is a service robot which fulfills simple services in a non-obtrusive way.
It is capable of serving drinks or snacks ordered by the user. It may also dispose objects like wrappings or empty mugs. In our limited scenario we could not care for any action taking place in the kitchen (moving objects on or off the robot could only be simulated).

Below you find an interaction example with our SilentButler

• After being turned on the robot automatically moves to its designated waiting position.
• A user sitting in the robots view field may wave at it.
• The robot drives towards a reasonable position near the detected person. Neither obtrusive nor too far away to interact with him/her.
• Now it is possible to ask for one of the three services:
  • An order may be placed. To do so objects with an EAN code may be held in front of the robots camera.
  • An object may be placed on top of the robot in one of the designated object-holders. This object will be disposed.
  • The ordering process is finished by pushing the 'Finish'-button on the robots multitouch display. If at least one order has been made and the robot is send away it will fulfill the given tasks. Afterwards it returns to the user serving the ordered itmes and awaiting further commands.
  • False alarm. The robot may be sent back to its waiting position.

In the following a simple sketch of our state-machine can be found:

Results

The system worked as planned. Some minor enhancement possibilities:

• Reduce starting time of the system.
• Faster barcode scanning/feedback of scanning progress.
• Use Kinect for a more sophisticated path planning and obstacle avoidance.

Before:

After:




A demonstration is available as interaction video:

Discussion and Conclusion

Our conclusion is that by using ROS it is possible to create a robust robotic system with few ressources and time from scratch. Although minor improvements are reasonable we were satisfied by the outcome of the project. The learning curve of the ROS framework was steep but we managed to customize it towards our needs. Since we heavily rely on this framework in most of the components this was the most challenging but also most rewarding aspect of our work.

Outlook

Since all our goals from winter term 2011 were met we will concentrate on the robustness of the work done so far. In short this means an improvement of the navigation by introducing the kinect for 3D-navigation. Also, the barcode scanner is going to be enhanced to quicker output results.
Furthermore, we will aim for improving the user-friendliness even more:

• The user shall be capable of telling the robot to automatically generate a map of its surrounding.
• Individual annotions for 'waiting-position', 'kitchen-position', and so on shall be freely definable by the user.
• The system should work in a 'out-of-the-box' manner.