On the Quality and Timeliness of Fusion in a Random Access Sensor Network

Abstract

In this letter, we investigate the trade-off between the quality of multi-sensor fusion decision and the delay of making that decision in a random access sensor network in a particular setting where the correlation between two sensor observations decreases with the increasing distance between these sensors. We propose a system operation wherein the nodes located in a small neighborhood of a sensor node that has successfully transmitted its decision to the fusion center are deactivated to reduce the contention over the random access channel. We propose a utility function that involves the quality and timeliness of the decision and determine the optimal size of this neighborhood analytically under a Markovian system operation model and numerically by simulating the actual system operation. These experiments demonstrate how our model can be used to set this neighborhood size before data collection to achieve a desired trade-off between decision quality and timeliness.