Cyber-Physical System Integration of IoT Sensing and Machine Learning: A Cross-Domain Review of Decision Support and Control in Smart Buildings and Precision Agriculture

A new generation of smart buildings and precision agriculture is evolving through the integration of cyber–physical systems (CPS), which combine IoT sensors with machine learning (ML). As such, there is an implicit assumption made by researchers in most of these studies that the ML component represents the decision making mechanism within the overall system. Furthermore, most researchers do not articulate the full scope of the cyber–physical feedback loop linking prediction outputs, operational decisions based upon those predictions, actual actuation of the physical plant or farm operation, and subsequent performance evaluations. The outcome of this paper brings out transferable decision support patterns across domains such as the mechanisms which have proven to be effective in scenarios with low number or quality of data measurements. Specifically, we present a review for two CPS domains that benefit intensely from decision support: smart buildings and precision agriculture. We examined how sensing, data processing, ML, and control modules are combined in practice when creating decision support applications. This resulted in a review of the literature to identify architectural patterns, decision objectives, and feedback mechanisms in both domains. This combination insight paves the way for more flexible and more effective decision making applications compatible with different domains.