Bridging the Gap Between Semantic Planning and Continuous Control for Mobile Manipulation Using a Graph-Based World Representation

Abstract

We present our ongoing efforts to create a mobile manipulation database tool, a flexible multi-modal representation supporting persistent life-long adaptation for autonomous service robots in every-day environments. Its application to a prototypical domain illustrates how it provides symbol grounding to a reasoning system capable of learning new concepts, couples semantic planning with whole-body prioritized control, and supports exploration of uncertain and dynamic environments.