Elsevier

Computer Communications

Volume 32, Issue 1, 23 January 2009, Pages 1-13
Computer Communications

Dual wakeup design for wireless sensor networks

https://doi.org/10.1016/j.comcom.2008.06.022Get rights and content

Abstract

Energy is often identified as the single most important resource in wireless battery-powered sensor networks. While current wakeup schemes in the literature promise to conserve energy in such networks, they apply several assumptions that may not be always true. First, all applications are assumed to require sensing coverage and network connectivity continuously; Second, a random dense deployment of sensors is always assumed possible; And third, the sensing ranges can be easily modeled by some sensing circles. In this paper, we show that these assumptions are not always valid, and propose sensor node wakeup schemes based on combinatorics block design to address energy-related issues when common assumptions fail. Another distinguishing feature of our work is also the proposal of a dual wakeup design for sensing and communications as these are two very different tasks. Finally, we verified our proposed schemes with simulations and experiments.

Introduction

One of the key resources in wireless sensor networks is battery power. Conserving battery power, while achieving sensing and communications objectives in a network of sensor nodes is of principal importance to many wireless sensor network applications. Energy savings schemes (or wakeup schemes) that allow nodes to operate in low-power or power-off mode when not required have become increasingly popular as a primary means to extend the usable lifetime of a network.

Most propositions of wakeup schemes in the literature [1], [2], [3], [4], [15], [16], [17], [18], [19] consider only a single strategy for both sensing and communications. In fact, a large proportion of these works only describe either the sensing aspect or communication aspect in detail, but not both. Noticeably, the random asynchronous wakeup (RAW) [1] scheme for sensor networks imposes that nodes wake up randomly for a chosen fixed interval, once in every time cycle, TRAW. To ensure high data delivery probability to cope with the random wakeup, a high deployment density of nodes is required. Communications issues are therefore the main considerations of this scheme, where probability of data delivery and data delays are studied. RAW did not, however, consider a separate wakeup scheme for sensing, and it is generally assumed that RAW is of a coupled sensing communication (CSC) methodology, which we illustrate in Fig. 1a. In CSC, nodes wake up to perform both sensing and communications tasks at the same time, i.e. the set of nodes that wakeup at any point in time to perform sensing (sensing plane), are the same set of nodes that are also performing communications (communications plane) such as routing or data dissemination functions.

In the probing environment and adaptive sleeping (PEAS) [2] wakeup algorithm, the scheme maintains a necessary set of active sensors while putting other redundant nodes to sleep. Nodes in the sleep mode transit into the active mode and probe the environment to replace nodes that may have failed. Nodes are selected to be in the active mode based on a probing range Rp. A node remains in active mode when there are no other working nodes within the probing range Rp, until it depletes itself of energy. According to [2], the probing range Rp, usually much less than the transmission range so as to avoid disconnecting the network, can be tuned based on robustness requirements of the application. PEAS is more concerned with the control of network density of active nodes, which does not directly address sensing and communication problems. With a single wakeup strategy, PEAS also belongs to the CSC methodology.

In contrast to RAW and PEAS, the coverage configuration protocol (CCP) [3] considers both sensing and communication issues at the same time. However, CCP does not propose separate strategies for sensing and communications. Instead, the major focus is on sensing coverage. In CCP, different degrees of sensing coverage can be configured based on application requirement. By further enforcing that the communication ranges of sensors, RC, be at least twice their sensing ranges, RS, connectivity is guaranteed. In fact, [3] establishes a sufficient condition (double range property) that if RC  2RS, a 1-covered network implies a 1-connected network. If RC < 2RS, CCP can still cooperate with another ad hoc network algorithm SPAN [4] to provide both sensing coverage and connectivity at the same time. As a result, CCP focused on satisfying the sensing coverage requirements of the application, while exploiting either the double range property or SPAN to ensure that the network is connected. Whether it is CCP or CCP + SPAN, both versions integrate coverage and connectivity in a single framework with only a single wakeup strategy for each node. As such, they also belong to the CSC methodology in general, as illustrated in Fig. 1a.

Apart from the commonly adopted CSC methodology in current existing wakeup schemes, they also make three assumptions about applications that need not always be true. First, existing wakeup schemes assume that coverage and connectivity are to be provided to the application in continuous time; Second, a high density random deployment of nodes is possible; And third, coverage can be defined easily with a sensing circle of some fixed radius RS. However, these assumptions are not necessarily true for some applications, as illustrated below.

  • There exist applications that are time-tolerant with respect to both connectivity and coverage. By time-tolerance, we mean that connectivity and coverage are only required within some bounded time frame (and not continuously at all times). The infinitesimally small granularity of continuous coverage and continuous connectivity guarantees are excessive and not required by the applications. For instance, in underwater seismic detection of oilfields, there is hardly any difference whether sensing datasets are collected once every hour or once per second. In this respect, the application is time-tolerant within a time-frame (or equivalently, the sensing sampling resolution) of 1 h or more. These are especially true for applications whose modality of sensing measurements varies slowly and even predictably with time. Examples include temperature or humidity changes in the environment, growth rate of cracks [20] in supporting struts and beams, changes in chemical or mineral compositions in soil, etc.

  • Yet, full connectivity and full coverage1 is often preferred (but not continuously in time) so that commercial companies need not over-deploy an unnecessary number of sensor nodes just to put them to sleep as mere backups when some working sensor depletes itself of battery power. A completely random scattering of a large number of sensors without regard to cost is prohibitively unattractive. In a review article for structural health monitoring [21], it has been specifically pointed out that “There is an increased requirement to both provide cost savings with regard to maintenance and a safer environment by prevention of structural failure”. Cost savings can only be maximised if sensor deployment is planned and placements optimized.

  • There are also applications where the range of sensing cannot be easily defined by the use of some constant sensing radius RS. Examples include measuring the concentration near a localised point in an agricultural landscape, measuring strain energies at a particular point on a rigid body structure, etc. The usual definition of 1-coverage may not be very useful for such purposes. Even for commonly cited sensing modes such as temperature, acoustic and seismic sensing, their detection ranges may not be perfectly circular or spherical, which most coverage preservation algorithms often take for granted. In fact, to the best of our knowledge, we are unaware of any real implementations of sensor networks based on practical applications that are deployed to specifically ensure continuous 1-coverage. The deployment scenario with kc-coverage for any integer kc > 1 is, in our view, not likely to happen in the foreseeable future.

We are, therefore, motivated to propose a framework based on combinatorics block design [5], [6] to address the described limitations of current wakeup schemes. Significantly, we abandon the traditional view that sensing requirement can always be described in terms of sensing kc-coverage. As we also favor a decoupled sensing communications (DSC) methodology, we shall propose solutions in both sensing and communications planes, but based on the same block design framework. In our design, there is guaranteed opportunity for communications within bounded time and meeting certain sensing requirements. Another differentiating feature of our work is the concept of bounded time. Inspired by the work in [6], also based on block design originally proposed for mobile ad hoc networks, we extend their work to address specific issues important to sensor networks.

To summarize, we attempt to solve the sensor network wakeup problem, by proposing separate sensing and communication strategies, but based upon a common block design framework. Our solution is targeting:

  • Applications that have time-tolerance in terms of connectivity and coverage so that deployment need not be prohibitively dense.

  • Scenarios where defining the sensing problem in terms of continuous sensing coverage is difficult, impractical, or meaningless.

It is important to realize that our solution is not one of optimal sensor placement, but rather an energy-savings wakeup scheme that takes advantage of some known optimal sensor placement [22], [23], [24]. Our solutions are also targeted for applications that require long-term monitoring of the environment, a subject, or an event without the luxury of having a dense sensor deployment. Last but not least, our solutions apply only to applications that are time-tolerant in nature, and not of a fast real-time nature.

Section snippets

Problem formulation

We consider a set of points/nodes QS in a space Sp that requires monitoring by sensors (See Fig. 2).

Definition 1

Full sensing coverage (or full coverage) is defined as the maximum achievable coverage when all nodes in the set QS are active.

Definition 2

Full network connectivity (or full connectivity) is defined as the maximum achievable connectivity in the communications graph when all nodes in the set QS are active.

The monitoring of space Sp needs to satisfy the following requirements.

Block designs

Both Time-tolerant requirements can be considered together using a single analysis framework based on block designs related to the field of combinatorics [5].

A functionality z is defined to be in the sleep mode2 when that

The communications plane

Consider a network of 13 nodes (labeled A–M) deployed as in Fig. 4 employing the (13, 4, 1) cyclic design set of Fig. 3 for the communications functionality x. Each sensor randomly selects any wakeup schedule (S1–S13) from the block design set. Lines between two nodes indicate bidirectional communication links. We adopt the neighbour discovery recommendation in [6] to have all nodes exchange their schedules with only their immediate one-hop neighbours after deployment.

Now, consider node J

The sensing plane

Sensing is an equally important task in sensor networks. It is one key distinguishing feature between pure wireless ad hoc networks and sensor networks. In this section, we adopt the TFCV requirement (Section 2) as our key sensing requirement and show later (Section 5.3) that this requirement can be met with our block design.

However, there may exist other sensing requirements very specific to the application itself. We illustrate this point by considering a typical realistic sensing application

The architecture

In this section, we describe several final components that put together the communications plane and sensing plane in a DSC methodology.

Comparisons with other wakeup schemes

Our proposed block design wakeup schemes have been proposed to meet the TFCN requirement for communications, and the MS safety and TFCV requirements for sensing. In this subsection, we briefly describe why other existing wakeup schemes cannot meet these requirements.

In CCP, the assumption is that nodes are deployed in relative abundance so that continuous 1-coverage (or kc-coverage) of the field is possible with a wakeup strategy to conserve energies of those sensors that are not required for

Conclusions

In this paper, a dual sensor node wakeup strategy for wireless sensing and communications has been proposed and justified. We proposed adaptations of the same cyclic symmetric block designs for different sensing and communication tasks, and show that they can provide certain time-bounded properties. In particular, we have shown that our designs satisfy the TFCN, TFCV, and other requirements with configurable parameters depending on the needs of applications. We have verified our work using

Glossary of acronyms

AR
assignment request (message)
AWSF
adaptive wakeup schedule function
BEACONS
Beacon packets (AWSF protocol)
CCP
coverage configuration protocol
CSC
coupled sensing-communication (methodology)
CYCLIC
original cyclic symmetric block design wakeup (scheme)
DSC
decoupled sensing communications (methodology)
ESM
eigenvector sensitivity method
FSA
free slot assignment (AWSF reconstruction)
MS
minimum safety (requirement)
PEAS
probing environment and adaptive sleeping (scheme)
RAW
random asynchronous wakeup (scheme)
RCD

References (25)

  • V. Paruchuri, S. Basavaraju, A. Durresi, R. Kannan, S.S. Iyengar, Random asynchronous wakeup protocol for sensor...
  • F. Ye, G. Zhong, J. Cheng, S. Lu, L. Zhang, Peas: a robust energy conserving protocol for long-lived sensor networks,...
  • L. Xing et al.

    Integrated coverage and connectivity configuration for energy conservation in sensor networks

    ACM Transactions on Sensor Networks

    (2005)
  • B. Chen et al.

    Span: an energy-efficient coordination algorithm for topology maintenance in ad hoc wireless networks

    ACM Wireless Networks Journal

    (2002)
  • I. Anderson

    Combinatorial Designs and Tournaments

    (1998)
  • R. Zheng, J.C. Hou, L. Sha, Asynchronous wakeup for ad hoc networks, Mobihoc’03,...
  • F. Warthman, Delay-tolerant networks (DTNs) a tutorial”, InterPlaNetary Internet Project Special Interest Group...
  • Y.F. Wong, L.H. Ngoh, W.C. Wong, An adaptive wakeup scheme to support fast routing in sensor networks, in: Proceedings...
  • Y. Yu, R. Govindan, D. Estrin, Geographical and energy aware routing: a recursive data dissemination protocol for...
  • Crossbow Technology Inc. [Online], available:...
  • Crossbow Technology Inc. [Online], available:...
  • V. Shnayder, M. Hempstead, B.-R.Chen, G. Werner-Allen, and M. Welsh, Simulating the power consumption of large-scale...
  • Cited by (0)

    View full text