Learning conditional effects of actions for robot navigation

GINKO, an integrated learning and planning system that has been applied to an autonomous mobile robot domain, is described. The goal of GINKO´s learning system is to partition the robot´s configuration space into regions in which actions exhibit a uniform qualitative behavior. This partitioning is performed by an inductive learning algorithm that classifies regions of the configuration space with regard to the effects of the robot´s actions when executed in those regions. GINKO´s learning is driven by its attempts to perform tasks. Thus, the learned effects of actions are directly applicable to normal system performance