Elsevier

Computers in Industry

Volume 64, Issue 7, September 2013, Pages 849-853
Computers in Industry

Optimizing communication in mobile ad hoc network clustering

https://doi.org/10.1016/j.compind.2013.05.005Get rights and content

Highlights

  • We consider typical communication workload of every mobile node in mobile ad hoc network clustering.

  • We also consider the additional communication workload of clusterheads in MANETs.

  • We propose an algorithm that optimizes communication workload, power consumption, clusterhead lifetime, and node degree.

  • Our clustering approach produces effectively balanced clusters over a diverse set of random scenarios.

Abstract

Mobile ad hoc networks (MANETs) are gaining popularity in recent years due to their flexibility, the proliferation of smart computing devices, and developments in wireless communications. Clustering is an important research problem for MANETs because it enables efficient utilization of resources, and must strike a delicate balance between battery energy, mobility, node degree, etc. In this paper, we consider the typical communication workload of every mobile node as well as the additional communication workload of clusterheads in MANET clustering. We propose an algorithm that optimizes communication workload, power consumption, clusterhead lifetime, and node degree. Experiment results show that our clustering approach produces effectively balanced clusters over a diverse set of random scenarios.

Introduction

Wireless mobile computing is gradually becoming mainstream nowadays. Executives from leading companies (e.g. Apple) have envisioned the post-PC era [1]. Some applications of mobile computing do not depend on a pre-existing infrastructure, such as routers in wired networks or access points in wireless networks, their communication can utilize a mobile ad hoc network (MANET). A MANET is a self-configuring, infrastructureless, wireless network of mobile devices. Many standardized technologies support MANET, such as Bluetooth [2], IEEE 802.11 (WiFi) [3], IEEE 802.15.3 (Wireless PAN) [4], Ultra-Wideband (UWB), etc. A mobile ad hoc network enables us to setup a temporary mobile network for instant communication without any fixed infrastructure. It has great potential in a variety of industrial applications such as emergency rescue, disaster relief, mobile conferencing, law enforcement, battle field communications, etc. It has been shown that a hierarchical network architecture will out-perform a flat structure for large MANET regardless of routing schemes [5], [6], [7]. A typical implementation of a hierarchical architecture is through a clustered structure. Choosing clusterheads optimally is an NP-hard problem [8].

In [9], [10], the authors propose a Highest-Degree clustering heuristic, which computes the degrees of a node based on its distance from other nodes. This approach is a big disadvantage because it does not put any restriction on the upper bound for the number of nodes in any cluster which severely impacts cluster throughput and stability. The Lowest-ID clustering heuristic was proposed in [11], [12], [13] which assigns a unique ID to each mobile node and chooses nodes with lowest ID's as clusterheads. It has better throughput than the Highest-Degree approach. However, nodes with smaller ID's tend to be picked as clusterheads repeatedly, which may quickly drain some of their batteries. Basagni proposed the distributed clustering algorithm [14] and the distributed mobility-adaptive clustering algorithm [15], which are also referred to as the Node-Weight heuristics. The heuristics assess the suitability of each node as clusterhead and assign a node-weight accordingly. However, a node has to wait for responses from all its neighbors to make its own decision as either clusterhead or clustermember.

Load-balancing clustering was studied in [16], [17], which believe that there is an optimum number of mobile nodes that a cluster can handle. The algorithm in [17] replace the current clusterhead with a new clusterhead if the current clusterhead cannot satisfy the node degree requirement. The algorithm in [16] merges neighboring clusters together or splits a cluster apart when the size of the cluster is too small or too large.

In [18], [19], the authors present a weighted clustering heuristic that combines various metrics for clustering, such as the number of nodes connected to a clusterhead, transmission power, mobility, and battery power of the nodes. This approach has been improved with genetic algorithm [20] and simulated annealing [21]. Surveys of various clustering schemes can be found in [22], [23].

The purpose of any mobile node to participate in a MANET is to communicate with other nodes. So there is a communication workload for every mobile node regardless of whether it is a clusterhead or not. In addition, the amount of communication for each node is different. Some nodes may generate a lot of network traffic, while other nodes may maintain very low amount of communication. If a node becomes a clusterhead, it will generate more communication with other nodes in the cluster, but that is only an overhead on top of its own communication workload. If all other criteria are equal, we should choose a node with smaller communication workload to serve as clusterhead. That will minimize the overall communication workload of the clusterhead node. All prior works did not address this load balancing aspect in terms of communication workload, because they did not optimize the communication needs of nodes in the network.

In this paper, we propose a new clustering scheme which strikes a balance between communication workload, node degree, power consumption, and remaining lifetime of clusterheads. Experiment results show that our clustering approach produce more balanced clusters over a diverse set of random scenarios.

Section snippets

Preliminaries

In this section, we introduce basic concepts and terminologies. Given a mobile ad hoc network, we use M to denote the set of mobile nodes in the network. Give a node vM, we use N(v) to represent the set of nodes in the neighborhood of node v:N(v)={v|vMdist(v,v)<r}where r is the transmission range of node v and dist(v,v) is the distance between nodes v and v. The mobile ad hoc network is organized by clusters. Each cluster has a clusterhead, and the rest of its nodes are clustermembers.

Theory

The MANET clustering problem is to partition M into a set of clusters, C = {c1, c2, …, cN}, such thatc1c2cN=MFor each cluster c  C, the head node (clusterhead) is denoted by h(c). The relative distance between each node and its clusterhead must lie within a transmission range r, i.e.cCncdist(n,h(c))<r

Let kvM(t) denote the normal communication of node v at time t with other nodes in the MANET M when it is not a clusterhead. This communication workload is the typical information exchange

Algorithm

We present an algorithm based on simulated annealing [24], [25], combined with the idea of dominated solution to handle multiple optimization objectives. The pseudo code is shown in Algorithm 1. In line 1, we randomly generated an ordered list of clusterheads, which is a solution s. The assignment of the rest of mobile nodes into their respective clusters is performed using a notion of average neighborhood. Give a node v, we use N¯(v) to represent the set of nodes in the neighborhood of node v:N

Implementation

One node P is picked (with large enough battery power) to run the MANET clustering algorithm, and then broadcast the result to the rest of the network. There are several methods to collect the relative distance between nodes:

  • 1.

    Every node carry a device to periodically measure the distance of other nodes. The local storage of each node can keep track of the distances for the last N measurements. When it comes to reclustering, each node will send the distances to the node P computing the clustering

Experiments

We simulate a mobile ad hoc network with N nodes to evaluate the performance of various clustering algorithms. The simulation is carried out in a 500 km × 500 km shaped region. We assume a normal distribution in the probabilistic communication volume between pairs of nodes. The communication workload kvM(t) for every node v is proportional to its probabilistic communication volume. If a node v is a clusterhead, its additional communication overhead kvC(t) is proportional to the size of the cluster.

Conclusion

In this paper, we consider the typical communication workload of cluster members as well as the additional communication workload of clusterheads in Mobile Ad Hoc Networks. We propose a new clustering scheme which strikes a balance between communication workload and other clustering objectives. Experiment results show that our clustering approach produces effectively balanced clusters over a diverse set of random scenarios.

Acknowledgements

The paper was supported in part by the Bagui scholarship project, the National Medium and Long-term Development Plan (Grant No. 2010ZX01045-002-3), the 973 Program of China (Grant No. 2010CB328000), the National Natural Science Foundation of China (Grant Nos. 61073168 and 61133016) and by the National 863 Plan of China (Grant No. 2012AA040906).

Xibin Zhao received the Ph.D. degree from Jiangsu University, China, in 2004. From 2007, he is an assistant professor in School of Software at Tsinghua University, China. His research insterests include reliability analysis of heterogeneous network and information system security.

References (27)

  • C. Cooper

    Apple's Cook: 172 Million ‘Post-PC’ Devices in the Last Year

    (2012)
  • Bluetooth Official Website (2011)....
  • IEEE Computer Society

    IEEE 802.11 Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications

    (2007, June)
  • IEEE 802.15 Working Group for WPAN (2011)....
  • P. Gupta et al.

    The capacity of wireless networks

    IEEE Transactions on Information Theory

    (2000)
  • X. Hong et al.

    Scalable routing protocols for mobile ad hoc networks

    IEEE Network

    (2002)
  • K. Xu et al.

    An ad hoc network with mobile backbones.

  • S. Basagni et al.

    A generalized clustering algorithm for peer-to-peer networks

  • A.K. Parekh

    Selecting routers in ad-hoc wireless networks

  • M. Gerla et al.

    Multicluster, mobile, multimedia radio network

    Wireless Networks

    (1995)
  • D.J. Baker et al.

    A distributed algorithm for organizing mobile radio telecommunication networks

  • D.J. Baker et al.

    The architectural organization of a mobile radio network via a distributed algorithm

    IEEE Transactions on Communications

    (1981)
  • A. Ephremides et al.

    A design concept for reliable mobile radio networks with frequency hopping signaling

    Proceedings of the IEEE

    (1987)
  • Cited by (20)

    View all citing articles on Scopus

    Xibin Zhao received the Ph.D. degree from Jiangsu University, China, in 2004. From 2007, he is an assistant professor in School of Software at Tsinghua University, China. His research insterests include reliability analysis of heterogeneous network and information system security.

    William N.N. Hung received the B.S. and M.S. degrees from the University of Texas, Austin, Texas, in 1994 and 1997, respectively, and the Ph.D. degree from Portland State University, Portland, Oregon, in 2002, all in electrical and computer engineering.

    He has worked at Intel, Synplicity and Synopsys over a variety of areas such as formal verification, FPGA synthesis, and Constrained Random Verification. He has published more than 60 refereed papers and has 2 patents with 3 more patents pending.

    Dr. Hung has served as Session Chair or Program Committee for many conferences, such as Design Automation Conference (DAC), Design, Automation and Test in Europe (DATE), Computer-Aided Verification (CAV), Formal Methods in Computer-Aided Design (FMCAD), IEEE World Congress on Computational Intelligence (WCCI) and International Conference on Computer Design (ICCD), etc.

    Yafei Yang obtained his master of engineering in Electrical and Computer Engineering at Portland State University in 2012. His research interests include mobile ad hoc network, digital signal processing, etc.

    Xiaoyu Song received his Ph.D. degree from the University of Pisa, Italy, 1991. His current research interests include formal methods, design automation, embedded system design, and emerging technologies.

    View full text