Acquiring a robust case base for the robot soccer domain

This paper presents a mechanism for acquiring a case base for a CBR system that has to deal with a limited perception of the environment. The construction of case bases in these domains is very complex and requires mechanisms for autonomously adjusting the scope of the existing cases and for acquiring new cases. The work presented in this paper addresses these two goals: to find out the "right" scope of existing cases and to introduce new cases when no appropriate solution is found. We have tested the mechanism in the robot soccer domain performing experiments, both under simulation and with real robots.