Problem solving and learning for a humanoid robot

Abstract

Based on our previous work that integrated a cognitive architecture, ICARUS, and a humanoid robot, MAHRU, we extend our use of the architecture to higher-level capabilities of problem solving and learning. During this process, we found the need to address a known limitation of ICARUS's problem solver. An extension is made to allow handling of underspecified goals in the system, and we demonstrate this new ability in a blocks world task that involves alphabet writing with several identical blocks.