Subjective Mapping

Extracting a map from a stream of experience is a key problem in robotics and artificial intelligence in general. We propose a technique, called subjective mapping, that seeks to learn a fully specified predictive model, or map, without the need for expert provided models of the robot’s motion and sensor apparatus. We briefly overview the recent advancements presented elsewhere (ICML, IJCAI, and ISRR) that make this possible, examine its significance in relationship to other developments in the field, and outline open issues that remain to be addressed.

Authors

• Michael Bowling University of Alberta

Keywords

No keywords are indexed for this paper.