Control-based auto-scaling of compute nodes in a fog cluster with service time guarantees

In this paper, we adopt a control-theoretic approach for navigation of a differential drive mobile robot (DDMR) while offloading vision data to be processed in a scalable cluster of fog computing nodes. The approach comprises developing a linear parameter varying (LPV) framework for modeling and a linear matrix inequality (LMI) based control design for navigation of DDMR and auto-scaling compute nodes in fog cluster independently. We validate the developed theory for mobile robot navigation in the application environment with service time guarantees in processing offloaded vision data for object detection.