Towards Contextual Robots for Collaborative Manufacturing

In the near future, rigid production lines will be likely complemented, if not completely replaced, with hybrid teams composed of humans and robots. Robots, in this case, that are collaborative, intelligent, being both stationary and mobile. Remarkably, hybrid teams will take over a variety of assembly operations on the same existing production lines. In this scenario, robots do not solely execute fully-automatized processes but instead become cooperative partners of the human worker, taking over non-ergonomic assembly operations and use contextual information to support selection of tools or adapt to ongoing certain tasks. Until now, companies had to decide in advance either for flexible but manual production processes or for highly-productive, repetitive automation with relatively low flexibility. In the future, there will be no need to decide whether a whole process needs to be fully automatized or not, but rather it will be possible to adapt specific tasks according to the skills of both the human and the robot. The goal of the Hybr-iT project [1] (funded by the German Federal Ministry of Education and Research (BMBF)) is the development and testing of hybrid teams composed of humans, robots and software-based assistant systems in real industrial environments. One essential component of the project is a resource-oriented architecture (ROA) acting as a middleware for the interconnection of heterogeneous cyber-physical systems and IT environments, and the hybrid teams. Another key component, subject of this paper, is a modular and robot-agnostic control software architecture which seamlessly interacts with the ROA and establishes the information flow for the control of robots in hybrid teams. Through the interplay of the ROA and the robot architecture, existing IT environments and robots should be allowed to be interconnected, so that robots receive additional and valuable information such as semantic task descriptions for hybrid teams, semantic information about the objects, current ongoing tasks and plans, location of objects and humans, to name a few. With the use of this additional contextual information and background knowledge from the IT systems, the aim is to grant robots with more flexible robot skills (such as navigation, interaction, perception or manipulation) which