Experimental Applications of Mapping Services inWireless Sensor Networks

James Shuttleworth, Mohammad Hammoudeh, Elena Gaura, Robert NewmanCogent Computing Applied Research Centre

Department of Creative ComputingCoventry University, Coventry, UK. CV1 5FB

{j.shuttleworth, aa2792, e.gaura, r.m.newman}@coventry.ac.uk

Fig. 1. Height map from USGS data covering the Grand Canyon

Abstract— Wireless sensor networks typically gather data at anumber of locations. However, it is desirable to be able to designapplications and reason about the data in more abstract formsthan points of data. This paper examines one way in which thiscan be done.

By bestowing the ability to predict inter-node values uponthe network, it is proposed that it will become possible to buildapplications that are unaware of the concrete reality of sparsedata. This interpolation capability is realised as a service ofthe network. We present an implementation of this service anddiscuss its merits and shortcomings.

Additionally, we present an initial application of the servicein the form of isopleth generation. That is, the delineation ofcontours of constant parameter value.

Finally, we discuss the improvements required to create moresophisticated applications and services and examine the benefitsthese improvements would bring.

I. INTRODUCTION

The purpose of any wireless sensor network is to gatherdata. In the most simple systems, the collected data mightbe instantly reported. In more complex systems, the data will

be processed, aggregated, distilled and acted upon within thenetwork, possibly with only selected data being reported backto a monitoring user. The wireless sensor nodes are usuallymore than capable of performing these tasks, as long as the ap-plication developer creates the required software, constructingspecific functions from what has become increasingly generichardware.

There has been a consistent effort to change the mechanismof use of certain capabilities in wireless sensor networks,to simplify and abstract them, turning them into services ofthe network rather than being the result of coordinating theservices of individual nodes. TinyDB’s retrieval of data [1],for example, uses the abstraction of SQL to effectively hidethe details of data collection, buffering, and transmission.

Following on from such work and moving to a slightlyhigher level of abstraction, we propose a new network service:map generation. The primary purpose of wireless sensornetworks is to collect and transmit data, but other capabilitieshave arisen to support this goal. Just as clustering, routing andaggregation allow for more sophisticated and efficient use ofthe network resources, a mapping service would support othernetwork services and make many more applications possiblewith little extra effort.

In this paper, we present our first steps towards a generalmapping service. We define a simple service, discuss oursimulation of networks that implement it and present anexample application built on the service.

II. BENEFITS OF AN IN-NETWORK MAPPING SERVICE

We predict many benefits to result from the development ofan efficient and flexible in-network mapping service.

Networks of nodes are built to solve problems and manyproblems are essentially problems of interpolation betweenpoints and it is this set of problems that we intend to addresswith a mapping service. Defining lines of constant height orpressure, contours and isobars, for example, requires knowl-edge not just of measurements at a few scattered locations,but also the likely values between them. Contours and isobarsare, in fact, specific types of isopleth and a very simple earlyimplementation of isopleth determination is presented later inthis paper.

Another problem that is at heart simply a problem of inter-polation is surface reconstruction. While mesh-based modelsare feasible and useful for specific applications int he con-text of WSNs [2], an interpolation-based model of surfacereconstruction offers benefits such as being locally applicable,of arbitrary detail and easily subjected to image processingtechniques.

Other aspects of wireless sensor networks that would benefitfrom such a service include clustering, in which geographicalcontext might be useful; deployment, for which the mappingservice could provide information on network density relatedto terrain or phenomenon complexity, and so on.

As well as these easy to identify benefits, it is likely thathaving such a service would make new applications obvious,just as there has been as recent increase in new Internetapplications combining existing services to produce new anduntil now unthought-of “mashups”.

Of course, the efficacy and efficiency of applications basedon such a service are tied to how well the service is imple-mented.

The implementation of the mapping service presented inthis paper, and the example application built upon it, are notoptimal. In short, they are inefficient and naıve. However,while more sophisticated implementations of the algorithmare being devised and tested, this simple approach gives usan opportunity to investigate the usefulness and validity of itsapplication.

III. ALGORITHMS FOR MAP GENERATION

Map generation is essentially a problem of interpolationfrom sparse and irregular points. Given a set of known datapoints representing the nodes’ perception of a given measur-able parameter of the phenomenon, what is the most likelycomplete and continuous map of that parameter?

In the field of computer graphics, this problem is known asan unorganised points problem, or a cloud of points problem.That is, since we assume that the position of the points in xyis known, the third parameter can be thought of as height andsurface reconstruction algorithms can be applied.

Simple algorithms use the point cloud as vertices in thereconstructed surface. These are not difficult to calculate, butcan be inefficient if the point cloud is not evenly distributed,or is dense in areas of little geometric variation.

Approximation, or iterative fitting algorithms define a newsurface that is iteratively shaped to fit the point cloud. Al-though approximation algorithms can be more complex, thepositions of vertices are not bound to the positions of pointsfrom the cloud. For applications in WSNs, this means that wecan define a mesh density different to the number of sensornodes, and produce a mesh that makes more efficient use ofthe vertices. Self organising maps are one of the algorithmsthat can be used for surface reconstruction [3]. This methoduses a fixed number of vertices that move towards the knowndata.

Note that surface reconstruction on typical non-overlappingterrains is equivalent to sparse-data interpolation. This kind

of geometric parameter interpolation has been shown to workwell for reconstructing underlying geography when the entirenetwork has been queried [2]. However, It does not extendwell to variable surfaces or overlapping local mapping, sinceit requires a complete data set to define the surface.

A more general method is interpolation by inverse distanceand, specifically, Shepard [4] interpolation which improves onit.

The simple inverse distance algorithm, used in WSN appli-cations before [5], is defined as:

f(P ) =

N∑i=1

d−ui zi

N∑i=1

d−ui

if d 6= 0 for all Di

zi if di = 0 for some Di

Where, P is the point at which the interpolated value isrequired, di is the distance from P to the point numbered iin the N known points and zi is the known value at pointi. The exponent, u, is used to control the smoothness of theinterpolation. High values lead to sharp edges between regionswhile low values lead to soft edges.

Shepard [4] devised a number of improvements to this basicalgorithm to limit the effect of distant points, make use of thedirection of the relationship between known points and thepoint to be interpolated and to incorporate information on theslope between known points. The algorithm, including the firsttwo of these refinements, has been implemented to interpolatebetween sensor readings and is discussed further here.

IV. SIMULATION WITH SENSOR

In order to implement our mapping approach and studyits properties, we have extended an in-house sensor networksimulator called “SenSor” [6]. SenSor is a realistic andscalable Python based simulator, in which each sensor noderuns in its own thread and communicates using the sameprotocols as its physical counterpart. Sensors have a fixed API,with customisable internals. This enables us to experimentwith different algorithms for managing the network topology,simulating fault management strategies and so on, within thesame simulation. Sensors are modelled as a pool of concurrent,communicating threads. Individual sensors are able to:

1) Gather and process data from a model environment2) Locate and communicate with their (geographically or

otherwise) nearest neighbours3) Determine whether they are operating ”correctly” and

act accordingly to alter the network topologySeparate interfaces gather information from the network

and display it. This partitioning allows us to experiment withdifferent ways of processing individual node data into infor-mation. Using this simulation framework, we implement ourmapping services. For our experiments, we provided SenSorwith interpolation capabilities and developed a plug-in to dodraw the isopleths.

For the purposes of the experiment reported here, nodesare designated into logical groups called clusters where eachcluster is managed by a master node called cluster-head,as in LEACH [7]. The use of clusters improves networkperformance and reliability by localising network traffic. Also,cluster-head nodes run a time aggregation function to reduceredundancy and minimise energy consumption by decreasingthe number of transmissions. In this algorithm, robustness isachieved by storing multipath and electing backup node(s) thatcan substitute for the cluster-head when the current clusterround ends or in some failures. During the set-up phase nodeswill select to belong to the nearest cluster-head using thenumber-of-hops metric. The use of this metric will result ina more fair distribution of nodes among clusters and establishshorter paths, hence reducing the energy expenditure sincebridging the distance between the node and its cluster-headwill be less expensive. Because of the random transmissiondelays and synchronisation errors, we use a back-off waitingscheme, similar to one used in [8], to lessen these effects.

When a node receives an advertisement message it waits fora constant time to gather more set-up messages; when the timerexpired it rebroadcasts the message with the best metric andresets the back-off timer. The number-of-hops metric togetherwith the back-off time will help to reduce the set-up messagesand create more uniform clusters. During the transmissionphase, sensing nodes transmit data to their cluster-head whichretransmit messages to the querying node. All nodes weregiven a virtually infinite supply of energy and the protocol wasallowed to run until it converged. Since the energy is unlimitedwe have used other metrics to measure energy consumptionlike the number of messages sent. For our experiments, wecreated an 600-node network in such a way that nodes arescattered randomly in area of 256×256 m2. Figure 2 shows arandom topology of 600 nodes scattered over a terrain wherethe edges represent communication neighbours.

The power of the sensor radio transmitter is set to cover allnodes within a 50m radius. The processing delay for transmit-ting a message is randomly chosen between 0 and 5 seconds,simulating real-world characteristics of low-power radio trans-mission. Using this network configuration we gathered the datato interpolate the landscape and draw the isopleths from theinterpolated output.

The landscape image is fed to the simulator and sensorswill “perceive” the colour intensity at the xy position ofthe landscape image corresponding to their coordinates asgathered sense data. After cluster-heads are selected and pathsare established, an external observer can choose to query anynode to collect the information necessary to interpolate thesurface from the network. Every node responds to the querymessage by sending its sense value and position through thecluster head. This collected information could then be used tobuild the network surface using the interpolation algorithm.

With this framework in place, we were able to experimentwith isopleth generation.

Fig. 2. SenSor, the WSN simulator, showing information transmissionbetween nodes and a visualisation of the interpolated surface

Fig. 3. Interpolated field generated from 600 simulated nodes randomlydistributed on the surface from Figure 1

V. EXPERIMENTAL APPLICATION

The determination of isopleths or contours is useful in anumber of applications and the ability to find the “edges”of a phenomenon has been put forward before as a usefuloperation [9]. Such an ability would enable the systematicfinding and delineation of the borders of phenomena suchas gaseous emissions or freezing conditions, creating heightcontour maps, calculating lines of sight, and so on.

To highlight this potential use of an in-network mappingservice, we present an implementation of isopleth generation.This is a simple implementation, with little regard for effi-ciency, but nevertheless provides a genuinely useful output forthe purposes of visualisation or phenomenon edge delineation.

With no optimisation or refinement to reduce communica-

Fig. 4. A contour at 116 units drawn over Figure 1

tion costs, the algorithm is very simple. A request is made toany node, giving the value of the required isopleth. A height of120 metres, for example, if contours of height were needed.This node is then responsible for collecting the informationneeded to produce the result. A threshold figure is also given,t, so that values ±t are included.

The simple algorithm presented here causes this node tothen query every other node in the network for their valueof the parameter in question. Once all data is collected, theinterpolation is performed as described in Section III, andevery value matching the required isopleth is recorded as beingpart of that isopleth.

This algorithm has been implemented using the simulationsoftware SenSor and the simple API described in Section IV.Figure 1 shows a height map of a section of the Grand Canyon,taken from data recorded by the US Geological Survey [10]. Ifwe apply the isopleth generation algorithm to this data directly,we are left with a result such as that shown in Figure 4. Thiscontour was generated for a height of 116 units and a thresholdof 1.

To present the result of generating isopleths on interpolatedfields as described above, Figure 5 shows the result of gener-ating an isopleth with the same parameters as in Figure 4, onthe interpolated field shown in Figure 3.

Clearly the interpolated terrain is similar to the real surface.Determining exactly how close the similarity is, and what thealgorithmic limits to the accuracy of the representation, is aproblem we are currently investigating. Consider, though, thatthe information used to reconstruct the surface in Figures 3and 5 is just 600 points, while the original is recorded as65,536. Taking the position of the nodes into account as extrainformation, the reconstruction is built using less than 3% ofthe original data.

Fig. 5. A contour at 116 units drawn on the interpolated field in Figure 3

In a more advanced implementation, this algorithm wouldbe replaced by one that begins with an efficient search for thefirst matching value and then a process of extending the searchalong the isopleth as it is discovered.

VI. NEXT STEPS

We have presented, above, our very early research intomapping and the applications of mapping in wireless sensornetworks. Leading directly from the work in this paper, thereare a number of avenues that need to be followed.

The naıve approach above collects all of the data in a centrallocation and processes it there. In networks of a few nodes,this might even be the most sensible solution. However, innetworks of nodes large enough to be useful, the number oftransmissions needed to accomplish the task would become solarge as to require a more intelligent solution.

The cost of communication in this initial system is veryhigh because the interpolation step, the mapping service, isfairly simple and requires that all data is collected to a point.For the most simple case, in which the node collecting data isin direct communication with all other nodes in the network,the number of hops is simply N − 1, for N nodes. However,this is an unlikely situation in any real world application. Thelikelihood is exponential growth with N .

Although the cost, in terms of power of transmission of datain wireless sensor networks far outweighs the usual processingload, the nodes are most often power-limited devices with,at best, a fixed schedule of battery replacement, and socomputational drain on this resource cannot be completelyoverlooked. For small collections of data of around 50 nodesinterpolating 65,536 points (a square field 256 × 256) takesvery little time, just a few seconds, on a desktop computerand would take just a little more on modern sensor nodes.

In a real deployment, the number of sensor nodes mighteasily exceed 1,000, depending upon the required density,drastically increasing computational expense. The acceptabil-ity of such a computational load in a wireless sensor networkis debatable, but in this case would probably be unacceptablefor most deployments.

The simplest improvement is to develop a complete imple-mentation of Shepard’s refinements and assess how this affectsthe accuracy of the interpolated field when compared to aknown real surface or phenomenon. Although the algorithmpresented by Shepard [4] is decades old, it is often onlyimplemented in its most simple form and it would be ofgreat benefit to have empirical evidence of the performance ofthe various improvements to this suggested by Shepard, whenapplied to data likely to be perceived by sensor networks.

A particularly interesting feature of the refinements is thatthey deal with limiting the data included in the weightedaverage, or adjusting weights based on context. These im-provements might be instantly applicable as ways to reducecommunication costs as well as improve the quality of theinterpolation. That is, if local information is required, thenonly local nodes need be queried.

Taking this idea further, we arrive at the idea of a mappingservice that allows for local querying. Our proposed solution,currently being developed, is an application of the moreadvanced mapping service outlined above.

An isopleth for a given parameter at a given value can befound by a step-by-step search for the next point in the map atthe same level. That is, a single point is used as the start of theisopleth and the next is found by searching the neighbourhoodin progressively more detail. The process is repeated until theend of the isopleth moves out of the mapped region or meetsthe beginning. The highest level of detail that is searched willdetermine how fine the isopleth is.

The same improvements intended to decrease the cost oftransmission also reduce the cost of computation. The changefrom global to local area interpolation reduces the number oftransmissions and makes the task of interpolation significantlysmaller.

Additionally, since the local interpolation could be per-formed by local nodes, the computational cost is not justreduced, but also distributed. That is, rather than individualnodes taking big losses in battery life and becoming unusablequickly, the entire network’s batteries degrade at a moreconstant rate.

VII. CONCLUSIONS

In this paper we have described a mapping service forwireless sensor networks which turns the network to an entityrather than a collection of separated nodes. This mappingservice uses only existing data and network capabilities ofa classical sensor network at different levels depending on theapplication.

Future work may apply the mapping services to othernetwork exposures such as barriers that might cause discon-tinuities in the interpolated map, contours, or many othersophisticated applications.

We feel that this area of research is pertinent to modernwireless sensor networks, and in this paper we have takeninitial steps towards exploring it. We have identified appli-cations, such as isopleth generation, that the service wouldmake simple to develop, and challenges that need to be metbefore such a service is feasible. Challenges such as localinterpolation, required to make the service efficient enough tobe deployed.

This paper is not the result of a completed project, but theexposition of the start of one. The problems and limitationsdescribed above are the opportunities we intend to take andthe lines we intend to follow.

The work presented above has been useful in a numberof ways. Firstly, it allowed us to extend SenSor so that itis capable of simulating networks with complex predefinedvariations of parameters over space. This is a useful advance,since it allows for precise control over the parameters of thevirtual deployment.

Secondly, it gave us an opportunity to examine the appli-cability of Shepard interpolation in the reconstruction of aparameter map from sparsely sampled data. The experiencegained here is proving to be invaluable in developing moresophisticated implementations of the service and its applica-tions.

Finally, we have been able to identify limitations andproblems as well as potential extensions and solutions thatwould otherwise have been missed until later iterations.

REFERENCES

[1] J. M. Hellerstein, W. Hong, S. Madden, and K. Stanek, “Beyond average:Towards sophisticated sensing with queries,” in Proceedings of the 2ndInternational Workshop on Information Processing in Sensor Networks(IPSN ’03), 2003.

[2] J. K. Shuttleworth, E. I. Gaura, and R. M. Newman, “Surface re-construction: Hardware requirements of a som implementation,” inProceedings of the ACM Workshop on Real-World Wireless SensorNetworks (REALWSN’06), June 2006, pp. 95–96, aCM ISBN: 1-59593-431-6.

[3] Y. Yu, “Surface reconstruction from unorganized points using self-organizing neural networks,” in Proceedings of IEEE Visualization,1999, pp. 61–64.

[4] D. Shepard, “A two-dimensional interpolation function for irregularly-spaced data,” in Proceedings of the 1968 23rd ACM national conference.New York, NY, USA: ACM Press, 1968, pp. 517–524.

[5] R. Tynan, G. OHare, D. Marsh, and D. OKane, “Interpolation forwireless sensor network power management,” in Proceedings of theInternational Workshop on Wireless and Sensor Networks (WSNET-05).IEEE Press, June 2005.

[6] S. Mount, R. Newman, and E. Gaura, “A simulation tool for systemservices in ad-hoc wireless sensor networks,” in Proceedings of NSTINanotechnology Conference and Trade Show (Nanotech’05), vol. 3, May2005, pp. 423–426.

[7] W. Heinzelman, A. Chandrakasan, and H. Balakrishnan, “Energy-efficient communication protocols for wireless microsensor networks,”in International Conference on System Sciences, January 2000.

[8] L. Xia, X. Chen, and X. Guan, “A new gradient-based routing protocolin wireless sensor networks,” in Proceedings of ICESS 2004, 2004, pp.318–325.

[9] M. Elmusrati, R. Jantti, and H. Koivo, “Distributed sensor network datafusion using image processing,” in Proceedings of Systems Communi-cations, August 2005, pp. 383–388.

[10] USGS and C. McCabe, “Grand canyon terrain,” Georgia Insti-tute of Technology Large Geometric Models Archive. http://www-static.cc.gatech.edu/ projects/large models/ index.html, 1998.