<html xmlns="http://www.w3.org/1999/xhtml">
<head>
<meta charset="utf-8">
<meta name="generator" content="pdf2htmlEX">
<meta http-equiv="X-UA-Compatible" content="IE=edge,chrome=1">
<link rel="stylesheet" href="https://static.pudn.com/base/css/base.min.css">
<link rel="stylesheet" href="https://static.pudn.com/base/css/fancy.min.css">
<link rel="stylesheet" href="https://static.pudn.com/prod/directory_preview_static/62444798337ddb528507330b/raw.css">
<script src="https://static.pudn.com/base/js/compatibility.min.js"></script>
<script src="https://static.pudn.com/base/js/pdf2htmlEX.min.js"></script>
<script>
try{
pdf2htmlEX.defaultViewer = new pdf2htmlEX.Viewer({});
}catch(e){}
</script>
<title></title>
</head>
<body>
<div id="sidebar" style="display: none">
<div id="outline">
</div>
</div>
<div id="pf1" class="pf w0 h0" data-page-no="1"><div class="pc pc1 w0 h0"><img class="bi x0 y0 w1 h1" alt="" src="https://static.pudn.com/prod/directory_preview_static/62444798337ddb528507330b/bg1.jpg"><div class="c x1 y1 w2 h2"><div class="t m0 x2 h3 y2 ff1 fs0 fc0 sc0 ls0 ws0">COMMUNICATIONS<span class="_"> </span>OF THE ACM<span class="_ _0"> </span><span class="ff2 ls1">May 2000/Vol. 43, <span class="_ _1"></span>No. 5<span class="_ _2"> </span></span><span class="fs1 ls2">51</span></div><div class="t m1 x3 h4 y3 ff3 fs2 fc0 sc0 ls3 ws0">For <span class="_ _1"></span>pervasive <span class="_ _1"></span>computing <span class="_ _1"></span>performance, <span class="_ _1"></span>exploit <span class="_ _1"></span>the </div><div class="t m1 x3 h4 y4 ff3 fs2 fc0 sc0 ls4 ws0">physical <span class="_ _1"></span>limits <span class="_ _1"></span>of <span class="_ _1"></span>these <span class="_ _1"></span>densely <span class="_ _1"></span>distributed <span class="_ _1"></span>networks <span class="_ _1"></span>of </div><div class="t m1 x3 h4 y5 ff3 fs2 fc0 sc0 ls3 ws0">embedded <span class="_ _1"></span>sensors, <span class="_ _1"></span>controls, <span class="_ _1"></span>and <span class="_ _1"></span>processors.</div><div class="t m0 x3 h5 y6 ff3 fs3 fc1 sc0 ls3 ws0">W</div><div class="t m0 x4 h6 y7 ff4 fs4 fc0 sc0 ls5 ws1">ireless integrated network sen-</div><div class="t m0 x4 h6 y8 ff4 fs4 fc0 sc0 ls5 ws2">sors (WINS) provide distrib-</div><div class="t m0 x4 h6 y9 ff4 fs4 fc0 sc0 ls6 ws3">uted network and Internet</div><div class="t m0 x4 h6 ya ff4 fs4 fc0 sc0 ls5 ws4">access to sensors, controls, and</div><div class="t m0 x4 h6 yb ff4 fs4 fc0 sc0 ls5 ws5">processors deeply embedded in</div><div class="t m0 x4 h6 yc ff4 fs4 fc0 sc0 ls5 ws6">equipment, facilities, and the</div><div class="t m0 x3 h6 yd ff4 fs4 fc0 sc0 ls5 ws7">environment. The WINS network repr<span class="_ _3"></span>esents a new</div><div class="t m0 x3 h6 ye ff4 fs4 fc0 sc0 ls5 ws8">monitoring and control capability for applications</div><div class="t m0 x3 h6 yf ff4 fs4 fc0 sc0 ls5 ws9">in such industries as transportation, manufacturing,</div><div class="t m0 x3 h6 y10 ff4 fs4 fc0 sc0 ls5 wsa">health care, environmental o<span class="_ _3"></span>versight, and safety and</div><div class="t m0 x3 h6 y11 ff4 fs4 fc0 sc0 ls5 wsb">security<span class="_ _4"></span>. WINS combine microsensor technology</div><div class="t m0 x3 h6 y12 ff4 fs4 fc0 sc0 ls5 wsc">and low-po<span class="_ _3"></span>wer signal processing, computation, and</div><div class="t m0 x3 h6 y13 ff4 fs4 fc0 sc0 ls5 wsd">low-cost wireless networ<span class="_ _3"></span>king in a compact system. </div><div class="t m0 x5 h6 y14 ff4 fs4 fc0 sc0 ls7 wse">Recent advances in integrated cir<span class="_ _3"></span>cuit technology</div><div class="t m0 x3 h6 y15 ff4 fs4 fc0 sc0 ls7 wsf">have enabled construction of far more capable yet</div><div class="t m0 x6 h6 y16 ff4 fs4 fc0 sc0 ls7 ws10">inexpensive sensors, radios, and</div><div class="t m0 x7 h6 y17 ff4 fs4 fc0 sc0 ls7 ws11">processors, allowing mass</div><div class="t m0 x8 h6 y18 ff4 fs4 fc0 sc0 ls7 ws12">production of sophisti-</div><div class="t m0 x9 h6 y19 ff4 fs4 fc0 sc0 ls7 ws13">cated systems linking</div><div class="t m0 x9 h6 y1a ff4 fs4 fc0 sc0 ls7 ws14">the physical world to</div><div class="t m0 x9 h6 y1b ff4 fs4 fc0 sc0 ls7 ws15">digital data networks</div><div class="t m0 xa h6 y1c ff4 fs4 fc0 sc0 ls7 ws16">[2–5]. Scales range from local to global for applica-</div><div class="t m0 xa h6 y1d ff4 fs4 fc0 sc0 ls7 ws17">tions in medicine, security<span class="_ _4"></span>, factory automation, envi-</div><div class="t m0 xa h6 y1e ff4 fs4 fc0 sc0 ls8 ws18">ronmental monitoring, and condition-based</div><div class="t m0 xa h6 y1f ff4 fs4 fc0 sc0 ls7 ws19">maintenance. Compact geometry and low cost allow</div><div class="t m0 xa h6 y20 ff4 fs4 fc0 sc0 ls7 ws1a">WINS to be embedded and distributed at a fraction</div><div class="t m0 xa h6 y21 ff4 fs4 fc0 sc0 ls7 ws1b">of the cost of conventional wireline sensor and actu-</div><div class="t m0 xa h6 y22 ff4 fs4 fc0 sc0 ls7 wsd">ator systems.</div><div class="t m0 xb h6 y23 ff4 fs4 fc0 sc0 ls7 ws1c">WINS opportunities depend on development of a</div><div class="t m0 xa h6 y24 ff4 fs4 fc0 sc0 ls7 ws1d">scalable, low-cost, sensor-network ar<span class="_ _3"></span>chitecture. S<span class="_ _3"></span>uch</div><div class="t m0 xa h6 y25 ff4 fs4 fc0 sc0 ls7 ws1e">applications require deliv<span class="_ _3"></span>er<span class="_ _1"></span>y of sensor information to</div><div class="t m0 xa h6 y26 ff4 fs4 fc0 sc0 ls7 ws1f">the user at a low bit rate through lo<span class="_ _3"></span>w-power trans-</div><div class="t m0 xa h6 y27 ff4 fs4 fc0 sc0 ls7 ws20">ceivers. Continuous sensor signal processing enables</div><div class="t m0 xa h6 y28 ff4 fs4 fc0 sc0 ls7 ws21">the constant monitoring of events in an environment</div><div class="t m0 xa h6 y29 ff4 fs4 fc0 sc0 ls7 ws22">in which short message packets would suffice. Futur<span class="_ _3"></span>e</div><div class="t m0 xa h6 y2a ff4 fs4 fc0 sc0 ls7 ws23">applications of distributed embedded processors and</div><div class="t m0 xa h6 y2b ff4 fs4 fc0 sc0 ls7 ws24">sensors will require v<span class="_ _3"></span>ast numbers of devices. Con-</div><div class="t m0 xa h6 y2c ff4 fs4 fc0 sc0 ls7 ws25">ventional methods of sensor networking repr<span class="_ _3"></span>esent an</div><div class="t m0 xa h6 y2d ff4 fs4 fc0 sc0 ls7 ws26">impractical demand<span class="_ _5"> </span>on cable installation and net-</div><div class="t m0 xa h6 y2e ff4 fs4 fc0 sc0 ls7 ws27">work bandwidth. P<span class="_ _3"></span>rocessing at the source would</div><div class="t m0 xa h6 y2f ff4 fs4 fc0 sc0 ls7 ws28">drastically reduce the financial, computational, and</div><div class="t m0 xa h6 y30 ff4 fs4 fc0 sc0 ls9 ws29">management burden on communication system</div><div class="t m0 x9 h7 y31 ff5 fs5 fc2 sc0 ls3 ws0"></div><div class="t m0 xc h8 y32 ff3 fs6 fc0 sc0 lsa ws2a">G.J<span class="_ _3"></span>. P<span class="_ _3"></span>ottie and <span class="_ _1"></span>W<span class="_ _6"></span>.J<span class="_ _3"></span>. <span class="_ _1"></span>Kaiser</div><div class="t m0 x3 h9 y33 ff3 fs7 fc0 sc0 lsb ws0">WIRELESS</div><div class="t m0 x3 h9 y34 ff3 fs7 fc0 sc0 ls3 ws0">INTEGR<span class="_ _7"></span>A<span class="_ _8"></span>TED </div><div class="t m0 x3 h9 y35 ff3 fs7 fc0 sc0 ls3 ws0">NET<span class="_ _9"></span>W<span class="_ _9"></span>ORK</div><div class="t m0 x3 h9 y36 ff3 fs7 fc0 sc0 ls3 ws0">SENSORS</div></div><div class="t m0 xd ha y37 ff6 fs8 fc0 sc0 ls3 ws0"> </div></div><div class="pi" data-data='{"ctm":[1.461187,0.000000,0.000000,1.461187,0.000000,0.000000]}'></div></div>
</body>
</html>
<div id="pf2" class="pf w0 h0" data-page-no="2"><div class="pc pc2 w0 h0"><img class="bi x0 y0 w1 h1" alt="" src="https://static.pudn.com/prod/directory_preview_static/62444798337ddb528507330b/bg2.jpg"><div class="c x1 y1 w2 h2"><div class="t m0 x3 h6 y38 ff4 fs4 fc0 sc0 ls7 ws2b">components, networks, and human resources.</div><div class="t m0 x5 h6 y39 ff4 fs4 fc0 sc0 lsc ws2c">Her<span class="_ _3"></span>e, we limit ourselves to a security application</div><div class="t m0 x3 h6 y3a ff4 fs4 fc0 sc0 lsc ws2d">designed to detect and identify threats within some</div><div class="t m0 x3 h6 y3b ff4 fs4 fc0 sc0 lsc ws2e">geographic region and report the decisions concerning</div><div class="t m0 x3 h6 y3c ff4 fs4 fc0 sc0 lsc ws2f">the presence and nature of such thr<span class="_ _3"></span>eats to a remote</div><div class="t m0 x3 h6 y3d ff4 fs4 fc0 sc0 lsc ws30">observer via the Internet. In the context of this appli-</div><div class="t m0 x3 h6 y3e ff4 fs4 fc0 sc0 lsc ws31">cation, we describe the physical principles leading to</div><div class="t m0 x3 h6 y3f ff4 fs4 fc0 sc0 lsc ws32">consideration of dense sensor networks, outline how</div><div class="t m0 x3 h6 y40 ff4 fs4 fc0 sc0 lsc ws33">energy and bandwidth constraints compel a distrib-</div><div class="t m0 x3 h6 y41 ff4 fs4 fc0 sc0 lsc ws34">uted and layered signal processing ar<span class="_ _3"></span>chitecture, outline</div><div class="t m0 x3 h6 y42 ff4 fs4 fc0 sc0 lsc ws35">why network self-organization and reconfiguration ar<span class="_ _3"></span>e</div><div class="t m0 x3 h6 y43 ff4 fs4 fc0 sc0 lsc ws36">essential, discuss how to embed WINS nodes in the</div><div class="t m0 x3 h6 y44 ff4 fs4 fc0 sc0 lsc ws37">Internet, and describe a prototype platform enabling</div><div class="t m0 x3 h6 y45 ff4 fs4 fc0 sc0 lsc ws38">these functions, including remote Internet contr<span class="_ _3"></span>ol and</div><div class="t m0 x3 h6 y46 ff4 fs4 fc0 sc0 lsc ws2b">analysis of sensor-network operation.</div><div class="t m0 x3 hb y47 ff7 fs9 fc0 sc0 lsd ws39">Ph<span class="_ _9"></span>ysical Principles</div><div class="t m0 x3 h6 y48 ff4 fs4 fc0 sc0 lse ws3a">When are distributed sensors better than a single</div><div class="t m0 x3 h6 y49 ff4 fs4 fc0 sc0 lse ws3b">large device, given the high cost of design implicit in</div><div class="t m0 x3 h6 y4a ff4 fs4 fc0 sc0 lse ws3c">having to create a self-organizing cooperative net-</div><div class="t m0 x3 h6 y4b ff4 fs4 fc0 sc0 lse ws3d">work? What are the fundamental limits in sensing,</div><div class="t m0 x3 h6 y4c ff4 fs4 fc0 sc0 lse ws3e">detection theory<span class="_ _4"></span>, communications, and signal pro-</div><div class="t m0 x3 h6 y4d ff4 fs4 fc0 sc0 lse ws3f">cessing driving the design of a network of distrib-</div><div class="t m0 x3 h6 y4e ff4 fs4 fc0 sc0 lse ws2b">uted sensors?</div><div class="t m0 x5 h6 y4f ff8 fs4 fc0 sc0 lsf ws40">P<span class="_ _3"></span>ropagation laws for sensing.<span class="_"> </span><span class="ff4 ls7 ws41">All signals decay with</span></div><div class="t m0 x3 h6 y50 ff4 fs4 fc0 sc0 ls7 ws42">distance as a wavefront expands. F<span class="_ _9"></span>or example, in free</div><div class="t m0 x3 h6 y51 ff4 fs4 fc0 sc0 ls7 ws43">space, electromagnetic waves decay in intensity as the</div><div class="t m0 x3 h6 y52 ff4 fs4 fc0 sc0 ls7 ws44">square of the distance; in other media, they are sub-</div><div class="t m0 x3 h6 y53 ff4 fs4 fc0 sc0 ls7 ws45">ject to absorption and scattering effects that can</div><div class="t m0 x3 h6 y54 ff4 fs4 fc0 sc0 ls7 ws46">induce even steeper declines in intensity with dis-</div><div class="t m0 x3 h6 y55 ff4 fs4 fc0 sc0 ls7 ws47">tance. Many media are also dispersiv<span class="_ _3"></span>e (such as via</div><div class="t m0 x3 h6 y56 ff4 fs4 fc0 sc0 ls7 ws48">multipath or low-pass filtering effects), so a distant</div><div class="t m0 x3 h6 y57 ff4 fs4 fc0 sc0 ls7 ws49">sensor requires such costly operations as deconv<span class="_ _3"></span>olu-</div><div class="t m0 x3 h6 y58 ff4 fs4 fc0 sc0 ls7 ws4a">tion (channel estimation and inversion) to partially</div><div class="t m0 xe h6 y59 ff4 fs4 fc0 sc0 ls7 ws4b">undo the dispersion [12]. F<span class="_ _9"></span>inally<span class="_ _4"></span>, many obstr<span class="_ _1"></span>uctions</div><div class="t m0 xe h6 y5a ff4 fs4 fc0 sc0 ls7 ws4c">can render electromagnetic sensors useless. R<span class="_ _3"></span>egardless</div><div class="t m0 xe h6 y5b ff4 fs4 fc0 sc0 ls7 ws4d">of the size of the sensor array<span class="_ _6"></span>, objects behind walls or</div><div class="t m0 xe h6 y5c ff4 fs4 fc0 sc0 ls7 ws2b">under dense foliage cannot be detected.</div><div class="t m0 xf h6 y5d ff4 fs4 fc0 sc0 ls7 ws4e">As a simple example, consider the number of pix-</div><div class="t m0 xe h6 y5e ff4 fs4 fc0 sc0 ls7 ws4f">els needed to cover a particular ar<span class="_ _3"></span>ea at a specified res-</div><div class="t m0 xe h6 y5f ff4 fs4 fc0 sc0 ls7 ws50">olution. The geometry of similar triangles reveals that</div><div class="t m0 xe h6 y60 ff4 fs4 fc0 sc0 ls7 ws51">the same number of pixels is needed whether the pix-</div><div class="t m0 xe h6 y61 ff4 fs4 fc0 sc0 ls7 ws52">els are concentrated in one large array or distributed</div><div class="t m0 xe h6 y62 ff4 fs4 fc0 sc0 ls7 ws53">among many devices. F<span class="_ _9"></span>or free space with no obstruc-</div><div class="t m0 xe h6 y63 ff4 fs4 fc0 sc0 ls7 ws54">tions, we would typically favor the large array<span class="_ _6"></span>, since</div><div class="t m0 xe h6 y64 ff4 fs4 fc0 sc0 ls7 ws55">there are no communications costs for mo<span class="_ _9"></span>ving infor-</div><div class="t m0 xe h6 y65 ff4 fs4 fc0 sc0 ls7 ws56">mation from the pixels to the pr<span class="_ _3"></span>ocessor<span class="_ _9"></span>. Ho<span class="_ _3"></span>wever<span class="_ _4"></span>,</div><div class="t m0 xe h6 y66 ff4 fs4 fc0 sc0 ls7 ws57">coverage of a large ar<span class="_ _3"></span>ea implies the need to track mul-</div><div class="t m0 xe h6 y67 ff4 fs4 fc0 sc0 ls7 ws58">tiple targets (a very difficult problem), and almost</div><div class="t m0 xe h6 y68 ff4 fs4 fc0 sc0 ls7 ws59">every security scenario of interest involves heavily</div><div class="t m0 xe h6 y69 ff4 fs4 fc0 sc0 ls7 ws13">cluttered environments complicated by obstructed</div><div class="t m0 xe h6 y6a ff4 fs4 fc0 sc0 ls7 ws5a">lines of sight. Thus, if the system is to detect objects</div><div class="t m0 xe h6 y6b ff4 fs4 fc0 sc0 ls7 ws5b">reliably<span class="_ _6"></span>, it has to be distributed, whatever the net-</div><div class="t m0 xe h6 y6c ff4 fs4 fc0 sc0 ls7 ws2b">working cost.</div><div class="t m0 xf h6 y6d ff4 fs4 fc0 sc0 ls7 ws5c">There are also example situations (such as radar) in</div><div class="t m0 xe h6 y6e ff4 fs4 fc0 sc0 ls7 ws5d">which it is better to concentrate the elements, typi-</div><div class="t m0 xe h6 y6f ff4 fs4 fc0 sc0 ls7 ws5e">cally where it is not possible to get sensors close to tar-</div><div class="t m0 xe h6 y70 ff4 fs4 fc0 sc0 ls7 ws5f">gets. There are also many situations in which it is</div><div class="t m0 xe h6 y71 ff4 fs4 fc0 sc0 ls7 ws60">possible to place sensors in proximity to targets,</div><div class="t m0 xe h6 y72 ff4 fs4 fc0 sc0 ls7 ws2b">bringing many advantages.</div><div class="t m0 xf h6 y73 ff8 fs4 fc0 sc0 lsf ws61">Detection and estimation theor<span class="_ _1"></span>y fundamentals. <span class="ff4 ls3 ws0">A</span></div><div class="t m0 xe h6 y74 ff4 fs4 fc0 sc0 ls7 ws62">detector is given a set of observables {X</div><div class="t m0 x10 hc y75 ff4 fs1 fc0 sc0 ls3 ws0">j</div><div class="t m0 x11 h6 y76 ff4 fs4 fc0 sc0 ls7 ws62">} to determine</div><div class="t m0 xe h6 y77 ff4 fs4 fc0 sc0 ls7 ws63">which of several hypotheses {h</div><div class="t m0 x12 hc y78 ff4 fs1 fc0 sc0 ls3 ws0">i</div><div class="t m0 x13 h6 y79 ff4 fs4 fc0 sc0 ls10 ws63">} is true. These obser<span class="_ _1"></span>v-</div><div class="t m0 xe h6 y7a ff4 fs4 fc0 sc0 ls10 ws64">ables may<span class="_ _4"></span>, for example, be the sampled output of a</div><div class="t m0 xe h6 y7b ff4 fs4 fc0 sc0 ls10 ws65">seismic sensor<span class="_ _9"></span>. The signal includes not only the</div><div class="t m0 xe h6 y7c ff4 fs4 fc0 sc0 ls10 ws66">response to the desired target (such as a nearb<span class="_ _3"></span>y pedes-</div><div class="t m0 xe h6 y7d ff4 fs4 fc0 sc0 ls10 ws67">trian) but background noise and interference from</div><div class="t m0 xe h6 y7e ff4 fs4 fc0 sc0 ls10 ws68">other seismic sources. A hypothesis might include the</div><div class="t m0 x3 h3 y2 ff1 fs1 fc0 sc0 ls11 ws0">52<span class="_ _2"> </span><span class="ff2 fs0 ls12">May 2000/Vol. 43, No. <span class="_ _1"></span>5<span class="_ _a"> </span><span class="ff1 ls0">COMMUNICATIONS<span class="_"> </span>OF THE ACM</span></span></div><div class="t m0 x7 hd y7f ff7 fsa fc3 sc0 ls13 ws69">Figure 1. Remote access to seismic and image data; </div><div class="t m0 x14 hd y80 ff7 fsa fc3 sc0 ls13 ws69">(a) browser image of photo and seismic response of vehicle; (b) sensor node control panel; and </div><div class="t m0 x15 hd y81 ff7 fsa fc3 sc0 ls14 ws69">(c) sampling rate controls.</div><div class="t m0 xc hd y82 ff7 fsa fc0 sc0 ls15 ws0">(a)</div><div class="t m0 x16 hd y83 ff7 fsa fc0 sc0 ls15 ws0">(b)</div><div class="t m0 x17 hd y84 ff7 fsa fc0 sc0 ls13 ws0">(c)</div></div><div class="t m0 xd ha y37 ff6 fs8 fc0 sc0 ls3 ws0"> </div></div><div class="pi" data-data='{"ctm":[1.461187,0.000000,0.000000,1.461187,0.000000,0.000000]}'></div></div>