Scalable Real-Time Control in Industrial Cyber-Physical Systems.

Real-time communication and control performance are the precursor of industrial cyber-physical systems that employ Wireless Networked Control System (WNCS) in critical industrial applications including process control and smart manufacturing. Control performance and real-time communication are interdependent in a WNCS. Hence, optimizing control performance under limited resource of network requires a cyber-physical codesign approach. A codesign approach needs to be online to take into account both the current network condition and the current control behavior. Leading industrial wireless standards such as WirelessHART and ISA100 adopt software-defined networking as a centralized routing mechanism. Hence, in current approach, transmission schedules of the entire network are created centrally at a network manager in advance and are then disseminated to the nodes. Control performance optimization usually requires to update sampling rates and/or priorities of the control loops, thereby requiring to re-create the schedules. Thus, it becomes highly inefficient under the current fully centralized scheduling approach. In this paper, we propose to optimize control performance in an industrial WNCS through a scheduling-control codesign based on a local and online scheduling approach proposed in a recent work. We formulate the scheduling-control codesign problem to optimize the control performance based on model predictive control theory. Unlike existing offline solution, our codesign complies with online scheduling and entails a rolling optimization to take into account the current control performance to dynamically update the rates and priorities of the control loops. We propose a highly scalable solution of the codesign problem based on a local search approach used as an anytime algorithm.