Reinforcement Learning Approaches in Social Robotics

This article surveys reinforcement learning approaches in social robotics. Reinforcement is a framework for decision-making problems which an agent interacts through trial-and-error with its environment to discover optimal behavior. Since interaction key component both and robotics, it can be well-suited approach real-world interactions physically embodied robots. The scope of the paper focused particularly on studies that include physical robots human-robot users. We present thorough analysis In addition survey, we categorize existent based used method design reward mechanisms. Moreover, since communication capability prominent feature robots, discuss group papers medium formulation. Considering importance designing function, also provide categorization nature reward. includes three major themes: interactive learning, intrinsically motivated methods, task performance-driven methods. benefits challenges evaluation methods regarding whether or not they use subjective algorithmic measures, discussion view proposed solutions, points remain explored, including have thus far received less attention given paper. Thus, this aims become starting point researchers interested using applying particular research field.