As the field of Cyber-Physical Systems continues to advance, new and interesting changes regarding its capability, adaptability, scalability, and usability [1] have come about. The most notable change has been the aggressive expansion of the variety of entity types that can be deployed in these systems (i.e. the entity eco-system). With this expansion, the complexity of handling vastly different communication protocols, processes, and data becomes a significant challenge since these data sources possess many more characteristics compared to the scalar data acquired by traditional IoT inputs. These additional characteristics severely limit the usefulness of such data sources to a CPS/IoT control system, especially in determining what controls should be enforced, effectively rendering these growing areas of the IoT out of the scope of many current CPS/IoT control systems. New CPS/IoT control systems would therefore need to leverage a great deal of computing power on demand to be able to deal with these devices and their data at scale while still being able to keep up with the core demands of real-time control, adaptability, and usability that users of these systems expect. Typically, the response to these issues has been to rely heavily on cloud computing resources. But this reliance has come at the cost of additional latency that would prove disastrous in application domains such as Industry 4.0, Internet of Healthcare Things, and defense, to name a few. In our previous work [2]–[5], we developed the cloud-based, Command Messaging Policy Enforcement Service (CoMPES). CoMPES is the result of an initiative towards a policy enforcement platform for large scale, massively distributed Cyber-Physical Systems (CPS) with a core objective to offer an elastic architecture for managing and controlling physical devices using discrete control schemas that are interconnected via our novel CPS/IoT control system. However, CoMPES has a significant performance bottleneck in its policy enforcement cycle as it relies on a collection of processing hubs to forward entity telemetry to the cloud for processing.

In this paper, we propose a novel edge computing framework that relies on the established Pub/Sub communication paradigm to build a complex event processing framework that allows users to push critical computations to the edge of the IoT. We then, with supporting results, show that when we increase the utilization of edge computing resources, CoMPES outperforms the cloud-centric variant by a 20-43x when comparing the latency of transitional state changes and supports systems of significantly larger scale (10-100x) while maintaining the same level of service as the cloud-centric version.