SICK robot day 2014
Team Attempto places second at the SICK robot day 2014
This year we were able to continue our successful participation at the SICK robot day, placing second out of 14 teams from Germany, the Czech Republic, England and Italy. The competition was again organized by SICK AG in Waldkirch near Freiburg, Germany.
The task
This year's task was to collect cubes at a collection station at the center of a circular arena. These cubes were labeled with barcodes, which represented one of four delivery stations at the outer border. These deliveries were marked with the same signs already in use back in 2010. The robot got one point for every correctly delivered cube, and one negative point for each wrong delivery.
Our robot was able to deliver 6 cubes in 10 minutes in the first round, just as many as the winning team PARMA delivered in their better run. In the second run we had a hardware problem, leading to a 5 minute pause in which our robot did not move at all. Afterwards the robot was able to collect a few cubes, but not enough to achieve the first place.
Our approach
Our robots Arnie and Sly are based on a former, omnidirectional RoboCup robot that has attended the SICK robot day twice in the past. They were heavily revised in order to meet the requirements for the tasks of this year's event. The most important revision was to make the robots as low as possible to be able to drive under the pick up and delivery rings.
A combination of multiple methods was implemented to solve the sign and target detection problem. These includes artificial neuronal network based machine learning methods as well as Hough line detection. Furthermore we implemented a RANSAC based map analysis to detect the central area, which was supposed to be square, but ended up being an octagon.
We used a customized version of slam_karto for localization and mapping, which was optimized for faster grid map generation. Furthermore we employed laser scan segmenation to only map long segments for a more robust system. Path planning was done using A* with an omnidirectional motion model and path following was implemented using orthogonal projection and an exponential law.
Team Attempto
Attempto Tübingen is a team of students and faculty staff with background in Computer Science, Bioinformatics, Automation, Control and Cognitive Science. Our two robots Arnie and Sly are based on former RoboCup models that have attended SICK Robot Day twice.
Team Members
- Sebastian Buck (sebastian.buck at uni-tuebingen.de)
- Richard Hanten (richard.hanten at uni-tuebingen.de)
- Goran Huskić (goran.huskic at uni-tuebingen.de)
- Alina Kloss
- Jan Leininger
- Gerald Rauscher (gerald.rauscher at uni-tuebingen.de)
- Eugen Ruff
- Felix Widmaier