An energy-aware forecasting-based framework using a novel moving coordinate scheme for sensor real-time communications

In order to save on the energy expended by a sensor node in its communications with the sink, forecasting-based frameworks have recently been proposed. Those frameworks approximate at the sink most of the node measurements, saving the energy that would otherwise be wasted to transmit them. However, it has been the case that many of those frameworks bump into one or more of three problems. First, the framework may be so computationally sophisticated that the energy needed to run it at the node can surpass the energy it was intended to save. Second, the framework may insert excessive delay which degrades the performance of any real-time sink-resident applications. Third, the framework may backtrack, harming applications that make use of measurements immediately after their construction at the sink. In this paper, we present a linear regression forecasting-based framework that saves impressively on the communications energy of the node while avoiding all three problems. Using a novel moving-coordinate scheme, the framework makes the node transmit less data than is done in existing frameworks. This data reduction saves on energy, as evidenced by the experimental results carried out on three real world datasets.