Efficient Sensing of Correlated Spatiotemporal Signals: A Stochastic Gradient Approach

TL;DR

This paper introduces a low-cost, scalable method for monitoring correlated spatiotemporal signals using contour-based data compression and stochastic gradient algorithms to efficiently track signal variations in wireless sensor networks.

Contribution

It presents a novel progressive learning algorithm combining contour level compression with stochastic gradient methods for efficient spatial monitoring.

Findings

The method reduces sensor communication by focusing on contour levels.

It effectively tracks signal variations over time.

The approach is computationally efficient and scalable.

Abstract

A significantly low cost and tractable progressive learning approach is proposed and discussed for efficient spatiotemporal monitoring of a completely unknown, two dimensional correlated signal distribution in localized wireless sensor field. The spatial distribution is compressed into a number of its contour lines and only those sensors that their sensor observations are in a $\Delta$ margin of the contour levels are reporting to the information fusion center (IFC). The proposed algorithm progressively finds the model parameters in iterations, by using extrapolation in curve fitting, and stochastic gradient method for spatial monitoring. The IFC tracks the signal variations using these parameters, over time. The monitoring performance and the cost of the proposed algorithm are discussed, in this letter.