Real-Time Control Selection over the Computing Continuum

The computing continuum is an architecture where a set of interconnected computing nodes, like cloud, edge, and IoT devices, can be used for data processing and decision making. On one hand, edge and cloud devices can provide significant support for addressing complex control problems. They can exploit their high computational capabilities to implement more sophisticated control algorithms than the onboard units, thus achieving higher performance. On the other hand, the key drawback of edge and cloud lies in the limitations of the real-time performance of the communication networks. While the onboard computing units can provide real-time feedback, the delays induced by the communication network might be large, possibly jeopardizing the benefits of the more advanced computational units. For this reason, it is fundamental to select the computing unit to implement the controller based on the actual delay experienced to communicate with the plant, possibly switching among different computing units in real-time. In this note, we address this problem and we propose a solution that selects online which controller (either the cloud, the edge, or the onboard controller) to use based on the network status. Simulations show the potential of the proposed method.