Abstract
An important question for production engineers is, »How do I automate this task?» In traditional industrial automation, e.g., a car factory, robots perform a small set of tasks for long periods of time. Robots are selected because their kinematic structure and strength suit the task requirements, and their motions are preprogrammed by a skilled programmer. In flexible manufacturing environments, tasks can change daily or hourly. The classic approach to automation is less suitable here; buying a dedicated robot for each set of tasks is inefficient, and coding each task is very time-consuming. Furthermore, qualified programmers with the requisite knowledge may not be available. To solve these problems, we propose a framework for modular robots that determines their structure and program based on human demonstrations. © 2018 Elsevier B.V., All rights reserved.