Perception-enhancement based task learning and action scheduling for robotic limb in CPS environment

Abstract

As an important application of a Cyber–Physical–Social System, a smart service robot has enormous potential to facilitate service for humans. However, there are still some challenges especially in performing daily life tasks because of the complicated physical environment and the various demands of humans during daily activities. Therefore, it is essential for a service robot to have high adaption to a changing physical environment and the capability of performing new-task learning. In this paper, a perception-enhanced smart robotic limb is presented to realize deeper incorporation of information in physical layers by deploying a three-channel perception and a semantic reasoning model to allow comprehensive perception of related objects, human actions and commands, for better adaption and robustness to environment changes. Based on the perception system, the robotic limb is capable of self-learning skill of new tasks from human demonstrations, The learning times and robustness to environment change were evaluated by a 50-task experiment.