In RoboCup, although the fields are standardized and color coded, the area outside the fields often contains many objects of various colors. Sometimes objects off the field may look very similar to balls, robots, or other objects normally found on the soccer field. Robots must detect all of these objects, and then differentiate between the true positives and false positives. This paper presents a new method using Gaussian fitness scores to differentiate between true positives and false positives for balls, robots, and penalty crosses. We also present some other improvements in our code base following our 2012 championship, such as our usage of a virtual base for forward kinematics calculations, our ability to flexibly transition player roles given dynamic numbers of teammates, and our ability to quickly integrate new kicks of varying speeds into our strategy. With these improvements, our UT Austin Villa team finished third in the Standard Platform League at RoboCup 2013.

PDF, PS, HTML

In The Eighth Workshop on Humanoid Soccer Robots at Humanoids 2013, October 2013.

@inproceedings{HUMANOIDS13-menashe, title={UT Austin Villa 2013: Advances in Vision, Kinematics, and Strategy}, author={Jacob Menashe and Katie Genter and Samuel Barrett and Peter Stone}, booktitle={The Eighth Workshop on Humanoid Soccer Robots at Humanoids 2013}, month={October}, url="http://www.cs.utexas.edu/users/ai-lab?menashe:humanoids13", year={2013} }