Interactive Natural Programming of Robots: Introductory Overview

Natural, multimodal interaction is a key issue of humanoid robots since a human-like appearance suggests human-like properties and behavior. Although human communication in all its facets seems to be hardly reachable today, interaction via verbal inand output or via gestures is a big step to meet this requirement. Besides these channels, recognition of the interaction context is important because e.g. human references are often incomplete and sloppy. From human behavior, several context types like advising, commenting and teaching can be derived. In this article, we present a system that supplies context dependant interaction with robots in these contexts. Interaction may be established with verbal and gestural channels. Furthermore, an interactive teaching technique called ”programming by demonstration” is presented for native and intuitive transferring human skills and behavior to human-like robots. Since manipulation skills from humans are very well developed but not easy to describe with standard programming methods, it is expected to ease robot programming significantly.