Event-based control as a cloud service

Event-based control has gained significant interest from the research community in recent years because it allows better resource utilization in networked control systems. In this thesis, we propose an architecture for offering event-based control as a service from the cloud, which not only improves resource utilization but also reduces the cost and setup time of large-scale industrial automation systems. Providing event-based control from the cloud, however, poses multiple research challenges. We address two of the main challenges, which are mitigating long and variable network delays and handling controller failures introduced because of moving the controller far away from the plant. We propose novel methods to solve the delay and failure problems and we show that these methods maintain the stability and performance of the control system. We implemented the proposed methods and deployed them on the Amazon cloud. Our results show that our delay mitigation technique can handle large communication delays up to several seconds with practically zero effect on the main performance metrics of the system. Moreover, the proposed fault tolerance approach can transparently handle controller failures even if the controlled system is thousands of miles away from its cloud controllers.