Elsevier

Computer Networks

Volume 48, Issue 4, 15 July 2005, Pages 549-566
Computer Networks

A study on the optimal time synchronization accuracy in wireless sensor networks

https://doi.org/10.1016/j.comnet.2004.10.018Get rights and content

Abstract

Time synchronization may play a key role in wireless sensor networks to meet real-time and energy-saving requirements and improve data-fusion and multiplexing efficiency. In this paper, under given constraints of hardware and mathematical models, we have studied performance limitations of the time synchronization for wireless sensor networks in terms of synchronization accuracy. First, sources of synchronization errors have been identified and a mathematical model has been introduced to analyze time synchronization schemes. Second, error distributions and the accuracy limitations have been formulated according to different error-source parameters. Third, a lightweight protocol is proposed, which is capable of approaching the performance limit as well as suppressing the communication overheads. Our idea is based on the observation that there is synchronization-error correlation between nodes receiving the same sequence of time-synchronization packets. Finally the theoretical analyses have been validated by simulation results.

Introduction

The integration of small-size, low-cost, highly sensitive sensors and low-power inexpensive wireless communication radios brings wireless sensor networks into reality [1], [2], [3]. Wireless sensor networks (WSN) are composed of a set of planned or ad hoc deployed sensors that are sensitive to their surrounding environments and capable of communicating with each other through their wireless channels. WSN can be applied in many application scenarios [4], such as smart environments, inventory management, precision agriculture, battlefield surveillance, animal herds tracking, and civil infrastructure monitoring.

Time synchronization is essential for distributed networks to realize certain functionality such as real-time event management and system monitoring. For example, data fusion [5] in WSN may require time synchronization. Consider cases where sensors are deployed in a dense fashion. When an event happens, multiple sensors that observe the phenomena may report at the same time. With synchronized clocks, redundant messages can be recognized and suppressed to save energy by reducing unnecessary data traffic across the sensor networks. Moreover, time synchronization can be used to realize synchronized sleep periods. It is desirable to put sensor nodes into sleeping mode and wake them up only when necessary to exchange information. To achieve this goal, synchronization is needed to activate the sensor nodes at scheduled time instance. Last but not least, if sensor nodes are time synchronized, the technology of Time Division Multiple Access (TDMA) can be used to improve the overall throughput of WSN.

In general, the time of any hardware clock in a computer or a sensor node does not follow the Coordinated Universal Time (UTC) provided by the National Institute of Standards and Technology (NIST) in an exact way. For a standalone computer system, there are several methods to improve the accuracy of its computer time, such as using precise clock board or a radio clock that receives time references transmitted from radio stations administrated by NIST. However, these approaches are costly with price ranging from hundreds to thousands of dollars a piece, and their sizes are not suitable for general sensor nodes. Another possible approach is to use Global Position System (GPS) devices to synchronize hardware clocks with satellites. Some manufactures are striving themselves to make GPS modules smaller, cheaper and more energy saving. One example is Leadtck GPS 9546 module especially designed for handheld devices, watches, and auto pilot systems; and it has already been integrated into Crossbow’s MTS420CA MICA2 sensor nodes with the cost of hundreds of dollars per node [17]. Considering WSN applications with thousands to millions of sensor nodes, having a GPS device in every node for time synchronization is too expensive.

For networked computers, Network Time Protocol (NTP) [6], [7] is the Internet standard for time synchronization, which synchronizes computer clocks in a hierarchical way by using primary and secondary time servers. NTP’s server–receiver synchronization architecture is widely accepted in designing time synchronization algorithms. However, its implementation is too heavy-weight to be supported by sensor nodes [10], [11].

The resource-availability constraints of sensors and ad hoc-network topologies of WSN impose specific requirements on protocol design of time synchronization for WSN applications. For example:

  • A requirement of lightweight feature. The underlying reason for this requirement is the hardware limitations of sensor nodes. In general, a sensor node has a small amount of memory, limited battery power, weak computation capability, and narrow data bus. For example, one commercial version of Berkeley Mote [8] only has 512-byte memory, 4-MHz microprocessor and 8-bit data bus. These constraints lead to the requirement of lightweight protocol design with small storage footprint, low computation complexity, and little energy consumption. In other words, the protocol should be storage-, computation-, and especially energy-efficient so that a sensor node is capable of supporting time synchronization as a basic function.

  • A requirement of self-configuration feature. Since sensor nodes may run out of their battery power or get destroyed in a harsh environment, the network topology and the resource distribution in WSN may change dynamically. A networked sensor node should not be constrained to synchronize its local clock with some specific or fixed sources of time references. WSN should be able to autonomously maintain synchronized time globally even when a large number of nodes fail or join the network.

  • A requirement of tunable feature. Clearly, different applications have different requirements of timing precision. For example, detecting the speed of a truck moving across a bridge requires the observing time in an accuracy of millisecond, while detecting the temperature changes during an explosion needs time accuracy of microseconds. Generally, the better accuracy a synchronization protocol achieves, the more synchronization packets it requires and the more energy consumption and computation complexity it needs. Thus, tunable synchronization algorithms are desirable to make a trade-off between the synchronization accuracy and energy consumption for various applications.


With these requirements in mind, we have studied the performance limitation of time synchronization for wireless sensor networks in terms of synchronization accuracy. In this paper, we first identify the sources of random synchronization errors and present a mathematical model to analyze the performance of synchronization algorithms. Using this model, the error distribution and bounds of synchronization accuracy are discussed. Note that each sensor node in WSN can transmit packets by broadcasting. It means that all its neighbors would receive the same sequence of packets broadcasted. Based on the observation of there exists correlation of the synchronization delays among the nodes in a neighborhood, we propose a lightweight synchronization protocol, LESSAR (Level Synchronization by Sender, Adjuster and Receiver), which is capable of approaching the accuracy limitation. LESSAR is a sender–receiver time synchronization mechanism just like NTP. But it extends the basic idea of NTP and makes the implementation more simple and lightweight for WSN. Details about the performance of LESSAR are given in [18].

The rest of this paper is organized as follows. Section 2 summaries existing time synchronization algorithms proposed for WSN. Section 3 discusses the sources of errors, describes the mathematical model, and analyzes accuracy limitation in details. Section 4 proposes the new time synchronization approach of LESSAR. Simulation results and associated discussions are presented in Section 5. Finally, Section 6 concludes this paper.

Section snippets

Related works

Over the last decade, a number of time synchronization protocols have been proposed to maintain the clocks in distributed systems [6]. NTP [7] has been the standard Internet time protocol because of its robustness, scalability, and flexibility. For mobile ad hoc networking environments (MANET) [9], some specific synchronization protocols, such as NAME [19], have been designed. A theoretical model for keeping clocks in ad hoc networks tick at the same speed is presented in [20]. However, there

Time errors of a single node

In general there is an internal hardware oscillator in any node, say X, which provides an almost stable vibration frequency to keep X’s local clock continuously running. The vibration frequency depends on the size, thickness and cutting edge of a quartz plate inside the oscillator and the conditions of X’s surrounding physical environment. In practice, the relationships between X’s clock tx and the Coordinated Universal Time (UTC) t matches the following function:tx=axt+t0x+Driftx(t),where ax

LESSAR protocol

Based on above-mentioned analysis model, we propose a NTP-like lightweight time synchronization protocol, namely LESSAR (LEvel Synchronization by Sender, Adjuster and Receiver), which is able to achieve the accuracy limitation while keep characteristics of low power consumption, affordable storage requirement, and small computation complexity by dramatically reducing the packets transmissions overheads.

Simulation results

We have implemented and tested LESSAR in both one hop and multi-hops wireless sensor network scenarios. OMNeT++ [15] is chosen as the simulator because it would not introduces any unclear delay into the simulation process and supports total control of system time. To mimic the behaviors of practical wireless sensor network, we take the hardware specifications of Berkeley Mote to model the sensor nodes in our simulation. The random PADRr is assumed to follow either a uniform distribution as an

Conclusion and future work

In this paper, we have studied the major sources of time synchronization inaccuracy that include the clock drift and the PADR delay. The time synchronization error distributions and the accuracy limitations are formulated according to different PADR distributions. We have also introduced a segmented linear model to analyze time synchronization schemes. The simulation results prove that the performance of time synchronization schemes can be predicted with high accuracy using the proposed model.

Acknowledgment

This research has been partly financed by grants from the National Science Foundation and the Commonwealth of Pennsylvania, Department of Community and Economic Development, through the Pennsylvania Infrastructure Technology Alliance (PITA).

Qing Ye is a PhD student of Computer Science and Engineering Department at Lehigh University. He received both of his M.S.E.E degree and B.S.E.E. degree from Tsinghua University, Beijing, PR China. His current research includes protocol design and management in wireless sensor network, multicast communication over wireless networks and autonomic networking.

References (21)

  • I.F. Akyildiz et al.

    A survey on sensor networks

    IEEE Communications Magazine

    (2002)
  • J. Hill, R. Szewczyk, A. Woo, S. Hollar, D. Culler, K. Pister, System architecture directions for networked sensors,...
  • D. Estrin, J. Heidemann, R. Govindan, S. Kumar, Next century challenges: scalable coordination in sensor networks, in:...
  • K. Römer et al.

    Middleware challenges for wireless sensor networks

    ACM Mobile Computing and Communication Review

    (2002)
  • W.R. Heinzelman, A.C. Chandrakasan, H. Balakrishnan, Energy efficient communication protocol for wireless microsensor...
  • D.L. Mills

    Internet time synchronization: the Network Time Protocol

    IEEE Transactions on Communications

    (1991)
  • D.L. Mills, Network Time Protocol (Version 3) specification, implementation and analysis, Request for Comments (RFC)...
  • M. Horton et al.

    MICA: the commercialization of microsensor Motes

    Sensors

    (2002)
  • C.E. Perkins

    Ad Hoc Networking

    (2001)
  • J. Elson, D. Estrin, Time synchronization for wireless sensor networks, in: Proceedings of the 2001 International...
There are more references available in the full text version of this article.

Cited by (25)

  • Novel clock synchronization algorithm of parametric difference for parallel and distributed simulations

    2013, Computer Networks
    Citation Excerpt :

    However, the signal of GPS-based clock synchronization is very difficult to drill through the barriers, an attribute that makes such synchronization not suitable for given environments (e.g., underwater, indoors, and underground). GPS receivers also require relatively higher energy and cost [35], which are impossible for large-scale deployment. NTP [10,36], proposed by Professor Mills of Delaware University in 1985, is an excellent mechanism because of its robustness, scalability, flexibility, and ubiquitous deployment.

  • A survey on cross-layer solutions for wireless sensor networks

    2011, Journal of Network and Computer Applications
    Citation Excerpt :

    Cryptography can protect communication from eavesdropping, however the tradeoff between data protection and energy consumption must be studied (Misic, 2008). And finally, in order to enable WSNs to use contentionless medium access and coordinated sleep schedules, they need to be synchronized (Ye et al., 2005). All the solutions for the aforementioned problems conflict directly with energy saving, since the techniques to solve them require resources from one or more nodes.

  • Model based design of a slot based time synchronization protocol for HF TDMA networks

    2020, 5th International Conference on Computer Science and Engineering, UBMK 2020
View all citing articles on Scopus

Qing Ye is a PhD student of Computer Science and Engineering Department at Lehigh University. He received both of his M.S.E.E degree and B.S.E.E. degree from Tsinghua University, Beijing, PR China. His current research includes protocol design and management in wireless sensor network, multicast communication over wireless networks and autonomic networking.

Yuecheng Zhang is a PhD candidate of Electrical and Computer Engineering at Lehigh University. He received both of his M.S.E.E. degree and B.S.E.E. degree from Tsinghua University, Beijing, PR China in 2001 and 1998 respectively. His research interests include routing protocols, cross-layer optimizations, time synchronization and positioning in wireless networks.

Liang Cheng is the Director of LONGLAB (Laboratory Of Networking Group, http://long.cse.lehigh.edu) and an Assistant Professor of Computer Science and Engineering Department at Lehigh University. He received his PhD degree from Rutgers, The State University of New Jersey, the M.S.E.E. degree from Tsinghua University, Beijing, PR China, and the B.S.E.E. degree from Huazhong University of Science and Technology, Wuhan, PR China, as an Honors Graduate. Currently his research interests include: sensor and ad hoc networks, middleware computing/architecture, network management of heterogeneous networks, network processing, and computer supported collaborative works. Cheng has been the PI and a Co-PI of projects supported by the National Science Foundation (NSF), the Defense Advanced Research Projects Agency (DARPA), Pennsylvania Department of Community and Economic Development, and Agere Systems, Inc. He has served 2005 IEEE Sarnoff Symposium on Advances in Wired and Wireless Communications as the Program Chair. Cheng is an awardee of Christian R. and Mary F. Lindback Foundation Minority Junior Faculty Award with Career Enhancement Grant.

View full text