Human interaction based Robot Self-Learning approach for generic skill learning in domestic environment

Unstructured domestic environments present a substantial challenge to effective robotic operation. Domestic environment requires service robots to deal with unexpected environment changes, novel objects, and user manipulations. We present an approach to enable service robots to actively learn high-level skills in an unstructured environment. This involves using a combination of processing environment changes, recording and learning user manipulation data, setting up meaningful hypothesis, proactively performing test actions and interacting with user feedback, and logic reasoning. We demonstrate our Robot Self-Learning (RSL) approach by using ROS (Robotic Operating System) and Care-O-bot (COB) 3. These methods enable service robots to learn generalized high-level skills in a sophisticated and changing environment. The RSL approach allows robots to learn new actions imposed by a human and action condition from perception changes from the environment. We also present logic based reasoning engine to speed up the self learning process.