Clustered Zigbee networks with data fusion: Characterization and performance analysis

Abstract

In this paper, we analyze the performance of clustered Zigbee wireless sensor networks (WSNs) with data fusion. Performance indicators at both physical (probability of decision error) and network (network transmission rate, throughput, aggregate throughput, delay, and network lifetime) layers are considered. Data fusion is carried out at the access point (AP) and, in clustered configurations, also at the clusterheads, which act as intermediate fusion centers (FCs). The goal of this paper is to shed light on the joint impact of topology and data fusion on the network performance. The presented results, mainly obtained through Opnet-based simulations, show clearly that the operational point of a Zigbee WSN with data fusion lies over a characteristic multi-dimensional surface, whose shape remains the same regardless of the number of nodes in the network. The existence of this peculiar surface highlights fundamental performance trade-offs in Zigbee networks.

Introduction

Wireless sensor networks (WSNs) have recently become an interesting research topic, both in military [1], [2], [3] and civilian scenarios [4], [5]. In particular, remote/environmental monitoring, surveillance of reserved areas, etc., are important fields of application of wireless sensor networking techniques. These applications often require very low power consumption and low-cost hardware [6], and clustering has been proposed as a possible strategy for saving energy. In [7], the authors present a system-level design methodology, based on a semi-random communication protocol, for clustered WSNs. In [8], after the derivation of an energy consumption model of WSNs, the transmission range is optimized. In [9], the authors investigate how the energy efficiency of a clustered WSN is affected by the transmit power distribution, the numbers of sensors in the clusters, the required end-to-end packet error rate, and the relative lengths of intra-cluster and inter-cluster distances.

Another important aspect of wireless sensor networking is the presence of an embedded data decision strategy, often involving data fusion. In [10], the authors present theoretical results on optimal decision rules and their application to data fusion. In [11], the authors follow a Bayesian approach for the minimization of the probability of decision error at the access point (AP). The data fusion mechanism has also a strong impact on practical applications. In [12], the authors analyze several methods of multi-sensor data fusion, such as Bayesian estimation, Kalman filtering, and Dempster–Shafer evidence theoretical methods, in order to design a move-in-mud robot. In [13], the impact of source–destination placement and communication network density on the energy costs and delay associated with data aggregation are evaluated.

In complex systems, such as WSNs, basic design approaches may no longer be sufficient to effectively improve the performance. Therefore, it is preferable to resort to joint optimization strategies which involve more than one layer of the communication/networking protocol stack, in order to significantly improve the performance. In [14], the authors present a cross-layer design framework, based on the use of an optimization agent to exchange information between different protocol layers, to improve the performance in WSNs. In [15], however, the authors show that unintended cross-layer interactions can have undesirable consequences on the overall system performance. Therefore, care has to be taken in developing cross-layer frameworks for WSNs.

In this paper, we shed light, through a simulation-based study, on the impact of clustering and data fusion on the performance of Zigbee networks. We first characterize the behavior of the network transmission rate as a function of the network tolerable death level, denoted as χ_net and representative of the network lifetime. We then evaluate the network transmission rate and the delay as functions of the packet generation rate. In addition, we provide a complete simulation-based characterization of Zigbee WSNs through the evaluation of network and physical layer performance indicators, such as network transmission rate, throughput, delay, network lifetime, and probability of decision error. The goal of this paper is to analyze the performance of clustered Zigbee networks with or without a simple data fusion mechanism, clearly understanding the interplay between network configuration and data fusion and, thus, deriving useful network design guidelines.

The structure of this paper is the following. In Section 2, preliminaries are given: Section 2.1 contains an overview of the Zigbee standard; in Section 2.2, the network tolerable death level is clearly defined; in Section 2.3, the clustering configurations of interest are introduced; in Section 2.4, we summarize the considered data fusion mechanisms from an analytical point of view; finally, in Section 2.5 the implementation characteristics of the used Opnet simulation model are accurately described. In Section 3, simulation-based performance results are shown and discussed: in Section 3.1, network transmission rate and delay are studied as functions of the network tolerable death level; in Section 3.2, the behavior of the same performance indicators is analyzed, for fixed values of network tolerable death level, as a function of the packet generation rate; in Section 3.3, we analyze the impact of data fusion on the throughput, the delay, and the probability of decision error. In Section 4, a few guidelines for the design of Zigbee WSNs, on the basis of the performance analysis in Section 3, are given: in Section 4.1, we show that the operational point of a Zigbee WSN lies over a characteristic multi-dimensional surface, whose shape is independent of the number of nodes and clearly shows the existence of fundamental performance trade-offs between χ_net, network transmission rate, and delay; in Section 4.2, we derive simple analytical approximation of the obtained multi-dimensional Zigbee performance surfaces. Section 5 concludes the paper.