Accepted at: 25th IEEE International Conference on Emerging Technologies and Factory Automation (ETFA 2020), pp. 1–8 Enabling Fog-based Industrial Robotics Systems Mohammed Salman Shaik V´ aclav Struh´ ar, Zeinab Bakhshi Van-Lan Dao, Nitin Desai Alessandro V. Papadopoulos Thomas Nolte M¨ alardalen University, Sweden {name.surname}@mdh.se Vasileios Karagiannis Stefan Schulte TU Wien, Austria {v.karagiannis, s.schulte}@dsg.tuwien.ac.at Alexandre Venito Gerhard Fohler TU Kaiserslautern, Germany {venito, fohler}@eit.uni-kl.de Abstract—Low latency and on demand resource availability enable fog computing to host industrial applications in a cloud like manner. One industrial domain which stands to benefit from the advantages of fog computing is robotics. However, the challenges in developing and implementing a fog-based robotic system are manifold. To illustrate this, in this paper we discuss a system involving robots and robot cells at a factory level, and then highlight the main building blocks necessary for achieving such functionality in a fog-based system. Further, we elaborate on the challenges in implementing such an architecture, with emphasis on resource virtualization, memory interference management, real-time communication and the system scalability, dependability and safety. We then discuss the challenges from a system perspective where all these aspects are interrelated. I. I NTRODUCTION Industrial robots are widely used in different automation applications such as painting and welding in automotive facil- ities and packaging in the food industry [1]. More recently, the domain of robotics has evolved to support warehouse automa- tion with mobile robots and, at the same time, emphasis on collaborative robots has also gained significant attention [2]. However, existing solutions are limited in addressing the demands of such applications due to limited computational resources and the strong coupling of the software and the computing hardware [3]. The on-demand availability of re- sources and reduced communication latency as offered by the fog computing paradigm [4] makes an interesting case for investigating the use of fog computing for addressing existing limitations. For example, the localization and mapping tasks of mobile robots which may be computationally intensive, have been successfully implemented using fog computing resources, improving the computation time significantly [5]. Furthermore, an optimized offloading algorithm which targets fog computing resources has been designed for robotic mission planning to show the benefits of fog computing [6]. While these approaches show the benefit of using fog computing in industrial robots, practical implementation of fog-based robotics systems remains a challenge. To facilitate further discussions on fog-based industrial robotics systems, and to identify the challenges that should be addressed to enable fog-based control of robots, in this paper, we provide an overview of the different technical aspects that are necessary for a practical realization of a fog network based on the OpenFog reference architecture [7] in Section II. In Section III, we describe a robotic cell-based fac- tory automation environment and a robot control application. In Section IV we use the factory automation environment to contextualize the technical aspects of the fog system and identify the potential challenges that should be addressed to enable a fog-based industrial robotics system. Here, we focus primarily on virtualization, resource orchestration, multi- core memory management and real-time communication along with a discussion on challenges from the dependability, safety and scalability perspectives. Finally, Section V concludes the paper. II. SYSTEM ARCHITECTURE While several fog computing architectures have been pro- posed in the literature, [7], [8], [9], for our discussion, we adopt the IEEE OpenFog reference architecture and deploy- ment model for fog computing [7]. Here, we assume that the participating devices are distributed between three layers (see Fig. 1) consisting of (i) cloud layer, (ii) fog layer, and (iii) device layer. The cloud layer provides a high computing capacity, but offers limited time predictability due to varying data transmission latencies. The fog layer provides an elastic environment in the vicinity of the origin of the data. It consists of a number of interconnected physical hardware devices (fog nodes) that are capable of hosting software applications on the shared node resources. While the processing power of the fog layer is less than that of the cloud layer, the network latency, however, is improved. The device layer consists of resource limited devices such as sensor and actuator devices that typically pre-process data from sensors and transmit it to the fog nodes, but also smart sensors and actuators, that are capable of communicating directly with the fog nodes. Fog Software Components and Services: We assume a fog computing architecture to be composed of a network of fog nodes, which can be viewed as a single logical entity [10]. The network is assumed to be hybrid of wired and wireless networks to exploit the benefits of both advanced wired and wireless technologies under the practical constraints of reliability, timeliness, and security for industrial environments. We briefly discuss some of the