Agents operating in the real world have to deal with a constantly changing and only partially predictable environment and are nevertheless expected to choose reasonable actions quickly. One way to address this problem is to use behavior networks as proposed by Maes, which support real-time decision making. Robotic soccer appears to be one domain where behavior networks have been proven to be particularly successful. In this paper, we analyze the reason for the success by identifying conditions that make behavior networks goal converging, i.e., allow them to reach the goals regardless of which particular action selection scheme is used. In terms of STRIPS domains one could talk of self-solving planning domains. We finally show that the behavior networks used for different robotic soccer teams have this property.

PDF

In 16th European Conference on Artificial Intelligence 2004.

@inproceedings{nebel:ecai2004, title={Goal-Converging Behavior Networks and Self-Solving Planning Domains}, author={Bernhard Nebel and Yuliya Lierler}, booktitle={16th European Conference on Artificial Intelligence}, url="http://www.cs.utexas.edu/users/ai-lab?nebel:ecai2004", year={2004} }