Programming by demonstration: A machine learning approach to support skill acquision for robots

Abstract

Programming by Demonstration (PbD) is a programming method that allows to add new functionalities to a system by simply showing the desired task or skill in form of few examples. In the domain of robotics this paradigm offers the potential to reduce the complexity of robot task programming and to make programming more ”natural”. In case of programming an assembly task PbD allows with the help of a video or a laser camera and a data glove the automatic generation the necessary robot program for the assembly task. In addition, the demonstration of the task with few different assembly situations and strategies may achieve a generalized assembly function for all possible variants of the class. In order to realize such a PbD system at least two major problems have to be solved. First, the sensor data trace of a demonstration has to be interpreted and transformed into a high-level situation-action representation. This task is not yet well understood nor solved in general. Second, if a generalization is required, induction algorithms must be applied to the sensor data trace, to find the most general user-intended robot function from only few examples. In this paper mainly the second problem is focused. The described experimental PbD environment consists of an industrial robot, a 6D space mouse used as input device, and some sensors. Various data can be recorded during a demonstration for further processing in the PbD system implemented on a workstation. The objective is to exploit the possibilities of integrating learning and clustering algorithms for automated robot programming. In particular it is investigated how human interaction with the PbD system as well as user-initiated dialogs can support inductive learning to acquire generalized assembly programs and skills.KeywordsProgramming by Demonstration (PbD)Robot Programming by Demonstration (RPD)TeachingRobotsManipulatorsUser IntentionsDialog-Based Learning (DBL)