Elsevier

Computer Networks

Volume 91, 14 November 2015, Pages 425-437
Computer Networks

Using central nodes for efficient data collection in wireless sensor networks

https://doi.org/10.1016/j.comnet.2015.08.041Get rights and content

Abstract

We study the problem of data collection in Wireless Sensor Networks (WSN). A typical WSN is composed of wireless sensor nodes that periodically sense data and forward it to the base station in a multi-hop fashion. We are interested in designing an efficient data collection tree routing, focusing on three optimization objectives: energy efficiency, transport capacity, and hop-diameter (delay).

In this paper we develop single- and multi-hop data collection, which are based on the definition of node centrality: centroid nodes. We provide theoretical performance analysis for our approach, present its distributed implementation and discuss the different aspects of using it. Most of our results are for two-dimensional WSNs, however we also show that the centroid-based approach is asymptotically optimal in three-dimensional random node deployments. In addition, we present new construction for arbitrary network deployment based on central nodes selection. We also present extensive simulation results that support our theoretical findings.

Introduction

A wireless sensor network (WSN) consists of small autonomous low-cost low-power devices that carry out monitoring tasks. Initially developed for military use, WSNs can nowadays be found in many civil applications, such as environmental monitoring, biomedical research, seismic monitoring, and precision agriculture [1]. The devices are called sensor nodes and the monitored data is typically collected at a base station, following a specific collection pattern of activated wireless links.

As these networks have no hard-wired underlying topology, one of the most fundamental issues when a WSN is deployed is the formation of an efficient communication backbone, or in other words, answering the question which links to use in order to collect the data from the sensor nodes?

Efficiency can be defined in many ways, for example it can be maximizing the rate at which data is collected [23], [45], [47] from the sensor nodes, prolonging the network lifetime by reducing the energy consumption [6], [9], [34], [36], [40], minimizing the number of hops from the sensor nodes to the collecting base station [16], [22], and other optimization objectives. It is apparent that the topological structure of the communication backbone plays a vital role in its efficiency. However, it is also important to note that a communication backbone which has good performance in some of the criteria can have a bad one in others. For example, using the minimum spanning tree (MST) as the backbone provides an optimal network lifetime performance for same initial battery charges [5], however it can have a very poor hop-diameter, which is critical for delay minimization.1 Thus, the network designer has to take special care when deciding which links to activate for the purpose of data collection, as different optimization objectives may be have a negative effect on each other.

The problem of data collection can be divided into two major paradigms. Data collection with aggregation [25], [43] allows each sensor node to accumulate the messages of its descendants and then pass only one fixed-size message towards the base station. The second paradigm, is data gathering without aggregation [29], [30] which requires that all messages initiated by the sensors will eventually reach the base station.

Our main objective in this paper is to construct efficient communication backbones for single- and multi-hop data collection with aggregation in WSNs for random sensor node deployments, while measuring the efficiency based on the next three metrics.

  • The transport capacity metric represents the sum of rate-distance products over all the active links. It is measured in bit-meters and was first introduced by [20]. The idea behind this measure is to capture both the notion of the overall rate and distance that the information travels in a network.

  • Hop-diameter is another important metric which reflects the depth of the data gathering tree, i.e., the maximum number of hops from any of the sensor nodes to the base station.

  • Total energy consumption is probably one of the most important parameters of a WSN as the sensor nodes are often deployed in areas where battery replacement is infeasible [8]. Wireless communication is a major contributor to the energy budget of a node. In this paper we focus on minimize the total energy consumed by all nodes for communication purposes.

We propose a novel approach for the construction of communication backbones by identifying central locations in the deployment area and routing all data through these regions. The general idea is that these locations would serve as aggregation points both on a local and global scale. In particular, we use an interesting geometrical notion of centroids, which is defined as the central geometrical position of a collection of nodes, which are used as a guide for the construction of hierarchical aggregation trees.

The rest of this paper is organized as follows. In Section 2 we present our system settings and state our objective. Related works are surveyed in Section 3. Sections 4 and  5 are the technical sections of the paper and show the construction of data collection communication backbones for three scenarios, single-hop general network and multiple-hop random network. We present additional construction for arbitrary network deployment based on central nodes selection in Section 6. Simulation results for various types of networks are presented in Section 7 and compared to the results of similar spirit obtained in [14]. Finally, we conclude and discuss future work in Section 8.

Section snippets

System settings

A WSN consists of n wireless sensor nodes, S={s1,,sn}, distributed in some area A. These nodes perform monitoring tasks and periodically report to a base station r which is located somewhere within the area A (we consider different locations throughout the paper). During the report phase, the sensor nodes propagate a message to the base station through a data collection tree, TS=(S{r},ES), rooted at r. We consider data collection with aggregation, where every node sS forwards a single unit

Related work

This work is a continuation of works [14], [16] which take into account all of the above three performance measures simultaneously. Below we discuss some of the related work on data collection, energy efficiency, bounded-hop communication, and transport capacity.

In terms of total energy consumption measure, it was proved in [39] that using the minimum spanning tree for data collection (gathering) with aggregation achieves optimality. A different criterion used to measure energy efficiency is

Single-hop collection

We start be defining the notion of geometric centroids and then analyze the performance bounds of single-hop communication backbone which is centroid-based. In the end we discuss the possible pitfalls of using a single-hop collection tree.

For n points P={p1,p2,,pn}, n ≥ 2, placed in the Euclidean plane, with coordinates (xi, yi), i=1,,n, and assuming general position, the centroid c(P) is a point defined as c(P)=(x¯,y¯), where x¯=i=1nxi/n and y¯=i=1nyi/n, which conceptually represents the

Multi-hop collection for random deployments

Our first construction of multi-hop data collection is for randomly deployed sensor networks. In this scenario we assume that n sensor nodes are randomly and independently placed in the area A with uniform distribution. We also assume that A is a unit square. We show an efficient communication backbone construction which is based on centroid networks, which are hierarchical geometrical structures on top of a point set P which represents the sensor nodes. As in the single-hop scenario, we assume

Multi-centers for arbitrary deployments

The construction of our data collection tree is based on finding a Hamiltonian circuit in the network. The existence of Hamiltonian circuit is evident since graph G represents a full graph (all the nodes potentially capable transmitting to each other). The Hamiltonian circuit is built on top of a minimal spanning tree of the graph, based on an algorithm presented by Andreae and Bandelt [2]. Following the algorithm, we build in linear time a Hamiltonian circuit with the following properties:

Theorem 6.1

[2]

Let

Simulation results

In this section we show some simulation results of the k-centroid network constructed for the multi-hop random scenario as described in Section 5. As we show, the simulation results fully support and even slightly outperform our theoretical analysis. In what follows we compare the k-centroid network topology with the optimal one, in terms of both energy consumption, hop-diameter and total link distance, which is achieved by using the minimum spanning tree and balance nodes partition based

Conclusions

In this paper we developed various data collection topologies that were based on the location theory notion of centroids and central nodes. We have shown that a centroids based hierarchy provides good approximation factor solutions for energy, transport capacity, and hop-diameter measures, in 2D, and performs asymptotically optimal in 3D for random sensors locations. Our simulation results verify our theoretical findings and, in fact, suggest that a possible tighter analysis for two dimensional

Milyeykovski Vitaly, finished M.Sc. degree in mechanical engineering from Tavria State Agrotechnological University, Ukraine, and M.Sc. in indusrial engineering from Tel-Aviv University, Israel, in 2000 and 2008, respectively. From 2009 to 2012 worked on obtaining Ph.D. in communication systems at Ben-Gurion University.

References (51)

  • D. Berend et al.

    Energy and lifetime efficient connectivity in wireless ad-hoc networks

    Ad Hoc Sens. Wirel. Netw.

    (2010)
  • G. Calinescu et al.

    Network lifetime and power assignment in ad hoc wireless networks

    Proceedings of the ESA’03

    (2003)
  • A. Chandrakasan et al.

    Design considerations for distributed microsensor systems

    Proceedings of the CICC’99

    (1999)
  • J.-H. Chang et al.

    Energy conserving routing in wireless ad-hoc networks

    Proceedings of the INFOCOM’00

    (2000)
  • Y. Chen et al.

    Per connection performance analysis of a frame-based TDMA/CDMA MAC protocol containing both reservation and contention slots

    Queueing Syst.

    (2005)
  • A.E.F. Clementi et al.

    The minimum range assignment problem on linear radio networks

    Proceedings of the ESA’00

    (2000)
  • A.E.F. Clementi et al.

    On the power assignment problem in radio networks

    Electron. Colloq. Comput. Complex.

    (2000)
  • A.E.F. Clementi et al.

    The power range assignment problem in radio networks on the plane

    Proceedings of the STACS’00

    (2000)
  • J. Crowcroft et al.

    Improved structures for data collection in wireless sensor networks

    Proceedings of the 2014 IEEE Conference on Computer Communications, INFOCOM 2014, Toronto, Canada, April 27 - May 2, 2014

    (2014)
  • S. Dolev et al.

    Bounded-hop energy-efficient liveness of flocking swarms

    IEEE Trans. Mob. Comput.

    (2013)
  • M. Elkin et al.

    Novel algorithms for the network lifetime problem in wireless settings

    ACM/Springer Wireless Networks

    (2011)
  • P. Fraigniaud et al.

    Assigning labels in an unknown anonymous network with a leader

    Distrib. Comput.

    (2001)
  • S. Funke et al.

    Bounded-hop energy-efficient broadcast in low-dimensional metrics via coresets

    Proceedings of the STACS’07

    (2007)
  • N. Garg et al.

    Faster and simpler algorithms for multicommodity flow and other fractional packing problems

    Proceedings of the FOCS’98

    (1998)
  • P. Gupta et al.

    The capacity of wireless networks

    IEEE Trans. Inf. Theory

    (2000)
  • Cited by (8)

    • Finding bounded diameter minimum spanning tree in general graphs

      2022, Computers and Operations Research
      Citation Excerpt :

      The communication topology of such networks is determined by the physical topology, transmission range, terrain shape, obstacles, and more. Originally, WSNs were used in military deployed networks, but nowadays, WSNs have spread to many other fields, such as environmental monitoring, precision agriculture, habitat monitoring, transportation, and biomedical purposes as well as in the growing field of the Internet of things (IoT) (Milyeykovski et al., 2015; Elson and Romer, 2002). Computing BDMST for WSN is important in order to improve the performance of WSN in terms of communication delay between the sensors.

    • Collective multi agent deployment for wireless sensor network maintenance

      2021, Engineering Applications of Artificial Intelligence
    View all citing articles on Scopus

    Milyeykovski Vitaly, finished M.Sc. degree in mechanical engineering from Tavria State Agrotechnological University, Ukraine, and M.Sc. in indusrial engineering from Tel-Aviv University, Israel, in 2000 and 2008, respectively. From 2009 to 2012 worked on obtaining Ph.D. in communication systems at Ben-Gurion University.

    Michael Segal finished B.Sc., M.Sc. and Ph.D. degrees in computer science from Ben-Gurion University of the Negev in 1994, 1997, and 1999, respectively. During a period of 1999–2000, Prof. Michael Segal held a MITACS National Centre of Excellence Postdoctoral Fellow position in University of British Columbia, Canada. Prof. Segal joined the Department of Communication Systems Engineering, Ben-Gurion University, Israel in 2000 where he served as departmenś Chairman between 2005–2010 and currently is a Full Professor. Prof. Segal serves as the Managing Editor of the Journal of Computer and System Sciences. He published more than 130 journal and conference papers on topics including algorithms (sequential and distributed), data structures with applications to optimization problems, mobile wireless networks, scheduling and efficient networking.

    Vladimir Katz received B.Sc. with honors in Communication Systems Engineering from Ben Gurion University of the Negev at 2008. Currently he is doing his M.Sc. degree in Communication Systems Engineering at Ben Gurion University, under the supervision of Prof. Michael Segal. His research interest are distributed algorithms and optimization problems in wireless networks.

    A preliminary version of this paper has appeared in [31].

    View full text