PropS: Towards Proprioception in Cyber-Physical Production Systems by Means of Collaborative Localization

Cyber-physical production systems (CPPS) are a key concept of Industry 4.0 in which heterogenous assets collaborate by sharing data via a network. In this paper we conceptualize and implement PropS, a proprioception system, which forms a global knowledge graph by gathering and interpreting distributed pose information from different assets of a CPPS. Pose information of the internal asset structure as well as detected objects of interests are shared in a standardized way to ensure seamless integration of new assets. We implement an asset administration shell submodel to define the necessary data structures. PropS enables collaborative localization by integrating distributing sensor information exceeding individual asset boundaries. The presented system can, e.g., reason about the docking pose for an autonomous mobile robot in a dynamically changing environment which is unknown to the robot and inferred from multiple cameras distributed over multiple assets. We implement PropS and the presented framework with real-world robots and sensors that use the Robot Operating System (ROS). The communication between the assets is realized with the Open Platform Communication Unified Architecture (OPC UA). The feasibility of the system is verified with a proof of concept experiment. We further provide current limitations and ideas for future work to unfold the full potential of PropS.