Fogification of Industrial Robotic Systems: Research Challenges Shaik Mohammed Salman shaik.salman@se.abb.com ABB AB, Västerås, Sweden Mälardalen University, Västerås, Sweden Vaclav Struhar vaclav.struhar@mdh.se Mälardalen University, Västerås Alessandro V. Papadopoulos alessandro.papadopoulos@mdh.se Mälardalen University, Västerås Moris Behnam moris.behnam@mdh.se Mälardalen University, Västerås Thomas Nolte thomas.nolte@mdh.se Mälardalen University, Västerås ABSTRACT To meet the demands of future automation systems, the architec- ture of traditional control systems such as the industrial robotic systems needs to evolve and new architectural paradigms need to be investigated. While cloud-based platforms provide services such as computational resources on demand, they do not address the requirements of real-time performance expected by control applications. Fog computing is a promising new architectural par- adigm that complements the cloud-based platform by addressing its limitations. In this paper, we analyse the existing robot system architecture and propose a fog-based solution for industrial robotic systems that addresses the needs of future automation systems. We also propose the use of Time-Sensitive Networking (TSN) services for real-time communication and OPC-UA for information mod- elling within this architecture. Additionally, we discuss the main research challenges associated with the proposed architecture. CCS CONCEPTS · Computer systems organization → Cloud computing; Ro- botics; Robotic components. ACM Reference Format: Shaik Mohammed Salman, Vaclav Struhar, Alessandro V. Papadopoulos, Moris Behnam, and Thomas Nolte. 2019. Fogifcation of Industrial Robotic Systems: Research Challenges. In Workshop on Fog Computing and the IoT (IoT-Fog ’19), April 15–18, 2019, Montreal, QC, Canada. ACM, New York, NY, USA, 5 pages. https://doi.org/10.1145/3313150.3313225 1 INTRODUCTION Industrial robots have become an integral part of the industrial automation environment with traditional application areas such as spot welding, spray painting and machining [7]. Currently, each ro- bot comes with a dedicated controller that provides motion control, programming interfaces and physical interfaces for integrating feld devices such as sensors and actuators via industrial networks [4]. Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for proft or commercial advantage and that copies bear this notice and the full citation on the frst page. Copyrights for components of this work owned by others than ACM must be honored. Abstracting with credit is permitted. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specifc permission and/or a fee. Request permissions from permissions@acm.org. IoT-Fog ’19, April 15–18, 2019, Montreal, QC, Canada © 2019 Association for Computing Machinery. ACM ISBN 978-1-4503-6698-4/19/04. . . $15.00 https://doi.org/10.1145/3313150.3313225 The controller is designed to meet the real-time constraints de- manded by motion control algorithms as well as real-time require- ments of industrial networks. These controllers, however, have fairly limited computational resources restricting the integration of complex functionality such as image processing, multi-robot motion control and other complex applications [24]. Although multi-robot motion control within a single controller is possible with solutions such as ABBs multimove functionality, the number of robots that can be controlled is still limited. Additionally, fexible production re- quirements of future automation systems impose demands such as frmware updates and hardware maintenance without any produc- tion downtime [6]. Meeting such requirements within the existing architecture is non trivial. Supporting the required infrastructure for augmented reality based immersive human-machine interaction concepts as shown by Paelke et al. [17] and Guhl et al. [8] will also require signifcant computational capacity and communication bandwidth. While increasing hardware capabilities within the con- troller can be a presented as a solution, this only addresses some of the concerns, validating the need to investigate cloud and fog-based architectures. While cloud computing ofers signifcant computational resources on demand, it does not guarantee real-time performance as required by traditional control applications [3, 5, 9]. Fog computing [26], is a new paradigm that allows utilization of computational resources near the edge of the network close to the source of the data. It introduces an intermediate layer between the cloud and the end devices that consists of a number of devices, called fog nodes, that are interconnected to form a network and these devices ofer their computational resources (e.g., CPU, storage), for use by applications within this network. While the well established cloud computing paradigm provides services ranging from collection of historical data to big data analysis, fog computing complements the cloud functionality by providing local data processing. This capability, along with real-time communication mechanisms such as TSN [20], enables the fog-based architecture to provide predictable commu- nication times. Authors of [11, 15] have discussed fog-based solutions for general robotic systems and highlighted the advantages of using fog-based architecture for such applications. While Hao et al. [10] provided a generic software architecture for fog computing, Faragardi et al. [3] provided a time predictable framework for a smart factory integrating the fog and cloud layers. Skarin et al. [22] developed a test bed to study the feasibility of a fog-based approach for control applications, while Pallasch et al. [18] and Mubeen et al. [16] showed