Modeling Time-critical Tasks for Heterogeneous Robotic Systems in Programming by Demonstration

Abstract

Programming by demonstration has been introduced in recent years as a rapid and efficient way to impart skills to robots. In programming by demonstration, a robot learns a new skill by having an end-user perform demonstrations of the skill, bypassing the need for traditional programming. As robotic systems can often be considered as composed of multiple heterogeneous components, learning skills for these systems requires capturing and preserving concurrency and synchronization requirements in addition to task structure. Furthermore, learning time-critical tasks depends on the ability to model temporal elements in demonstrations. This thesis proposes a modeling framework in programming by demonstration based on Petri nets capable of modeling these aspects. In this approach, models of tasks are constructed from segmented demonstrations as task Petri nets, which can be executed as discrete controllers for reproduction. The implementation details of a complete prototypical system are given, showing how elements of time-critical tasks can be mapped to those of Petri nets. Finally, the approach is validated by an experiment in which a robot learns and reproduces a musical keyboard-playing task.