Using event detection latency to evaluate the coverage
of a wireless sensor network
You-Chiun Wang, Kai-Yang Cheng, Yu-Chee Tseng
*Department of Computer Science, National Chiao-Tung University, Hsin-Chu 30010, Taiwan, ROC Available online 15 June 2007
Abstract
A wireless sensor network (WSN) consists of many tiny and low-power devices deployed in a sensing field. One of the major tasks of a WSN is to monitor the surrounding environment and to detect events occurring in the sensing field. Given an event appearing in a WSN, the event detection latency is to model the time that it takes for the WSN to be aware of the event. In this work, we analyze the latency using a probabilistic approach under an any-sensor-detection and a k-sensor-detection models, where k > 1 is an integer. Such an analysis can be used as an index to evaluate a WSN’s coverage and thus can help guide the deployment of a WSN. We also develop simulations to verify our analytical results.
2007 Elsevier B.V. All rights reserved.
Keywords: Ad hoc network; Network coverage; Pervasive computing; Ubiquitous computing; Wireless sensor network
1. Introduction and problem statement
Wireless sensor networks (WSN) have been intensively
studied recently [1]. A WSN consists of many tiny and
lower-power sensor nodes, each of which can collect sur-rounding environmental data and communicate with neighboring nodes. Communications in a WSN typically
takes place in an ad hoc manner[2]. Applications of WSNs
include surveillance and agriculture, habitat, traffic, and
civil infrastructure monitoring[3–7].
One of the major tasks of a WSN is to detect events occurring in the sensing field. Given an event appearing in a WSN, the event detection latency is to model the time that it takes for the WSN to be aware of the event. Such latency is an important metric to measure the monitoring capability of a WSN’s deployment for real-time
applica-tions such as surveillance[8–10]or object tracking[11–13].
We propose our model to analyze the event detection latency. Specifically, we are given a sensing field, on which
there are n homogeneous sensors. Each sensor has a sensing distance of r. Without loss of generality, we assume that these n sensors form a connected network. To simplify the analysis, we assume that the time axis is divided into fixed-length slots and the working schedule of each sensor is modeled by a sequence of working cycles, each of length T slots. Each working cycle is led by an active phase fol-lowed by an idle phase. The former consists of the first D
slots, and the latter the rest of the T D slots. Sensors only
conduct detection jobs in their active phases, and go to sleep in idle phases. However, sensors do not synchronize their clocks, so their working cycles are not necessarily
aligned. Fig. 1 shows an example. Note that this model
can be applied to most of the MAC/network protocols that are proposed for WSN recently. For example, for energy
conservation, the Zigbee/IEEE 802.15.4 standard [14]
allows a sensor node to wake up and sleep very similarly
to our working cycles inFig. 1. In fact, several other
pro-tocols (such as Bluetooth [15]and S-MAC[16]) also have
such an awake-sleep behavior.
Our objective is to evaluate the detection latency when an event occurs in the sensing field. Note that we make no assumption on the locations of events. To take errors into account, we also assume that in an active slot, a sensor has 0140-3664/$ - see front matter 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.comcom.2007.05.009
*
Corresponding author. Tel.: +886 3 5131366; fax: +886 3 5724176. E-mail addresses: wangyc@csie.nctu.edu.tw (Y.-C. Wang),
kycheng@csie.nctu.edu.tw (K.-Y. Cheng), yctseng@csie.nctu.edu.tw
(Y.-C. Tseng).
a probability of p to successfully detect the occurrence of an event if the event is within this sensor’s sensing range. To sim-plify the analysis, we assume that p is a constant but not a function of the distance between a sensor and the event [17,18]. We consider two detection models in this work:
• Any-sensor-detection model: To capture the event, the network needs at least one sensor to successfully detect the event.
• k-sensor-detection model: To capture the event, the net-work needs at least k sensors to successfully detect the event, where k > 1. (The value of k is application-depen-dent. For example, positioning protocols using
triangu-lation[19–21]require at least three sensors.)
2. Analysis of event detection latency
To facilitate the calculation of the event detection latency, we establish a system clock, which starts at the instant when the event appears. The system time is also slotted and each set
of continuous T slots forms a system cycle, as shown inFig. 1.
Suppose that an event appears at location (x, y) in the sensing field. Let M(x, y) be the number of sensors whose sensing ranges cover location (x, y). Consider the time slots that these M(x, y) sensors start their new working cycles in a system cycle. We can classify them into T groups such that the ith
group contains the misensors that start their working cycles
at the ith slot in a system cycle, i = 1, . . . , T. For example, in Fig. 1, sensor 1 belongs to group 2, sensor 2 belongs to group
1, and sensor n belongs to group T. Clearly, PTi¼1mi¼
Mðx; yÞ. Taking all combinations of mi’s into consideration,
the event detection latency under this particular M(x, y) can be written as LatencyðMðx;yÞÞ ¼X Mðx;yÞ m1¼0 X Mðx;yÞm1 m2¼0 X Mðx;yÞðm1þþmT2Þ mT1¼0 Mðx; yÞ! m1! . . . mT! 1 T Mðx;yÞ! dðm1; . . . ; mTÞ;
where the first term is the probability to observe a
particu-lar combination (m1, . . . , mT), and the second term
d(m1, . . . , mT) is the expected latency for this particular
combination.
As the event may appear in any location (x, y) inside the sensing field, we have to consider all possible M(x, y). Thus, the overall expected latency of the WSN can be expressed as ET ;D¼
Z
x
Z
y
LatencyðMðx; yÞÞdx dy;
¼X
n
i¼0
Prob½Mðx; yÞ ¼ i LatencyðiÞ: ð1Þ
When the event appears in an i-covered region, it will be sensed by i sensors (i.e., M(x, y) = i). Therefore, Prob[M(x, y) = i] is the ratio of areas that are i-covered
in the sensing field. So Eq.(1)can be simplified as
ET ;D¼
Xn
i¼0
Ai
A LatencyðiÞ;
where A is the area of the sensing field, and Aiis the total
area in A in which each point is covered by exactly i sensors.
In Sections 2.1 and 2.2, we will show how to compute
d(m1, . . . , mT) under our two detection models, respectively.
Table 1summarizes the notations used in this work. 2.1. Any-sensor-detection model
Under this model, the event is considered to be captured by the network if any sensor successfully detects its existence.
Let xi be the number of active sensors at the ith slot,
i = 1, . . . , T. These xi sensors are composed of three types
of sensors: (1) sensors which turn into active at the ith slot, (2) sensors which turn into active between the first and the
(i 1)th slots, and (3) sensors which turns into active before
the first slot. Note that case (2) can be true if D > 1 and i > 1, while case (3) can only occur when i < D. This leads to xi¼ miþ X minðD1;i1Þ j¼1 mijþ X Di1 j¼0 mTj:
We also define xaT+bas the number of active sensors at the
(aT + b)th slot for any a P 1. Since cycles repeat every T slots, we have xaT+b= xb.
Time . . . sensor n sensor 2 sensor 1 T
system cycle i system cycle i +1 slot active phase (D slots) idle phase (T-D slots) . . . . system cycle 1 . . . . Time that an event occurs . . . . . . . . . . . . working cycle (T slots)
The probability that there is at least one sensor success-fully detecting the event in the first slot is ð1 ð1 pÞx1Þ.
For i P 2, the probability that the event is not detected
in the first (i 1) slots but is successfully detected in the
ith slot isð1 ð1 pÞxiÞð1 pÞx1þþxi1. Hence, as the time
goes to infinity, the expected detection latency under the any-sensor-detection model is dðm1; . . . ; mTÞ ¼ X1 a¼0 XT b¼1 ðaT þ bÞ ð1 ð1 pÞxbÞ ð1 pÞaðx1þþxTÞþx1þþxb1: ð2Þ
Eq.(2)contains an infinite number of expressions. The
fol-lowing theorem shows that it will converge.
Theorem 1. The expected delay d(m1, . . . , mT) under the
any-sensor-detection model is bounded by
dðm1; . . . ; mTÞ 6
T2 ð1 aÞ2; where a = (1 p)D·M(x, y).
Proof. Since ð1 ð1 pÞxbÞ 6 1 and (1 p) 6 1, we can
obtain that dðm1; . . . ; mTÞ ¼ X1 a¼0 XT b¼1 ðaT þ bÞ ð1 ð1 pÞxbÞ ð1 pÞaðx1þþxTÞþx1þþxb1; 6X 1 a¼0 XT b¼1 ðaT þ bÞð1 pÞaðx1þþxTÞ; 6X 1 a¼0 ð1 pÞaDMðx;yÞX T b¼1 ða þ 1Þ T ! ; ¼X 1 a¼0 aaða þ 1ÞT2; ¼ T 2 ð1 aÞ2: 2.2. k-Sensor-detection model
Under this model, the event is considered to be captured by the network, once there are at least k sensors success-fully detecting its occurrence. Since the sequence x1, x2, . . .
has a period of T, the expected latency can be written as dðm1; . . . ; mTÞ ¼ X1 a¼0 XT b¼1 ðaT þ bÞ Pkðm1; . . . ; mT; aTþ bÞ; ð3Þ where Pk(m1, . . . , mT, aT + b) is the probability that there are
at least k sensors successfully detecting the event at the (aT + b)th slot, but not so before that slot. To find Pk(m1, . . . , mT, aT + b), let Nebe the number of sensors that
have already succeeded in detecting the event before the
(aT + b)th slot, and Nfbe the number of sensors that succeed
in detecting this event at the (aT + b)th slot for the first time. We first categorize sensors according to their behaviors as
shown inFig. 2. There are xaT+b= xbactive sensors at the
(aT + b)th slot, and the rest of M(x, y) xbsensors are
inac-tive. The inactive sensors can be further divided into a set of
N1sensors which have ever succeeded in detecting this event
before the (aT + b)th slot, and a set of M(x, y) xb N1
sensors which have not. Similarly, the active sensors can be
divided into a set of N2sensors which succeed in detecting
this event at this slot, and a set of xb N2sensors which fail
to detect this event at this slot. From the latter set, we further
identify a set of N3 sensors which have ever succeeded in
detecting this event before the (aT + b)th slot, but fail to de-tect this event at the current slot.
Based on the above definitions, once the values of Ne, N1,
N2, and N3are given, the rest of the variables inFig. 2will all
be fixed. Specifically, the number of sensors that successfully detect this event at the (aT + b)th slot and have also
suc-ceeded in doing that before is Ne N1 N3, and the number
of sensors that succeed in detecting this event for the first time at the (aT + b)th slot is Nf= N2 (Ne N1 N3). In
Eq.(3), the latency is considered to be aT + b if Ne< k and
Table 1
Summary of notations used in this work
Notations Definition
n Number of sensors in the sensing field
k Minimum number of sensors required to successfully detect the event
T Number of slots in a working or system cycle
D Number of slots that a sensor continues detecting the event
p Probability that a sensor successfully detects the event in an active slot
M(x, y) Number of sensors that can detect the event when the event occurs at location (x, y) mi Number of sensors in M(x, y) that repeat their working cycles at the ith slot in a system cycle
xi Number of sensors detect the event at the ith slot in a working cycle
Pk(m1, . . . , mT, aT + b) Probability that there are at least k sensors succeeding in detecting the event
Ne Number of sensors that have ever succeeded in detecting the event before the (aT + b)th slot
Nf Number of sensors that first succeed in detecting the event at the (aT + b)th slot
N1 Number of sensors that have ever succeeded in detecting the event before but do not detect at the (aT + b)th slot
N2 Number of sensors that succeed in detecting the event at the (aT + b)th slot
N3 Number of sensors that have ever succeeded in detecting the event before but fail at the (aT + b)th slot
Si Number of sensors in the subset i
Nf= (N1+ N2+ N3) NeP k Ne. By enumerating all
combinations of Ne, N1, N2, and N3, we can derive that
Pkðm1; . . . ; mT; aTþ bÞ ¼ Xk1 Ne¼0 XNe h1¼0 Prob½N1¼ h1 Xxb h2¼0 Prob½N2¼ h2 X Neh1 h3¼0 Prob½N3¼ h3 Prob½NfP k Ne !!! : ð4Þ Depending on the value of b, we can further derive the four
terms Prob[N1= h1], Prob[N2= h2], Prob[N3= h3], and
Prob[NfP k Ne] with three cases.
Case (1): b < D. Consider the set of M(x, y) xbinactive
sensors at the (aT + b)th slot. We divide them into two subsets:
• S1: The set of sensors whose active phases do not cross
the boundaries of system cycles.
• S2: The set of sensors whose active phases cross the
boundaries of system cycles.
Clearly, jS1j = mb+1+ mb+2+ + mT(D1) and
jS2j = mT(D1)+1+ mT(D1)+2+ + MT(Db). For
example, when b = 2,Fig. 3 shows the above two subsets
in case 1. Recall the definition of N1. Among these N1
sen-sors, let R1be the number of sensors belonging to S1, and
R2 the number of sensors belonging to S2. Since
R1+ R2= N1, we can expand Eq.(4)as follows:
Pkðm1; . . . ; mT; aTþ bÞ ¼X k1 Ne¼0 XNe h1¼0 Xh1 r1¼0 Prob½R1¼ r1 Prob½R2¼ h1 r1 X xb h2¼0 Prob½N2¼ h2 X Neh1 h3¼0 Prob½N3¼ h3 Prob½Nf P k Ne !!! : ð5Þ
Given two integers x and y such that x P y and a probabil-ity value z, let us define
Binoðx; y; zÞ ¼ ðxyÞzy ð1 zÞxy:
The probability functions in Eq. (5) are derived as
follows: Prob½R1¼ r1 ¼ BinoðjS1j; r1;1 ð1 pÞ aD Þ; Prob½R2¼ h1 r1 ¼ XD2 i¼1 mTðD1Þþi jS2j BinoðjS2j; h1 r1;1 ð1 pÞ aDþi Þ; Prob½N2¼ h2 ¼ Binoðxb; h2; pÞ; Prob½N3¼ h3 ¼ Binoðxb h2; h3; Xb1 i¼0 mbi xb ð1 ð1 pÞaDþbiÞ þ X Db1 i¼0 mTi xb ð1 ð1 pÞaDþbÞÞ; and
Succeed for the first time? Succeed at this slot?
Succeeded before? Succeeded before? Yes No Yes Yes Yes Yes No No No No Conduct detection at this slot? M(x,y) M(x,y)-xb M(x,y)-xb-N1 N1 xb-N2-N3 N3 Ne-N1-N3 N2-(Ne-N1-N3) = Nf N2 xb-N2 xb
Fig. 2. Classification of sensors in the (aT + b)th slot. Numbers in ovals indicate numbers of sensors.
{
{
{
Case (1) Case (2) Case (3) Slot Number a T + a T + a T + a T + a T + a T + a T + a T + 3 6 5 4 1 2 7 8 1 2 7 8 (a+1)T + (a+1)T + (a 1)T + (a 1)T + S2 S1 S3 S2 S1 S2 S1m
2m
1m
3m
4m
5m
6m
7m
8 ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... .... . .
. . .
: active slotProb½Nf P k Ne ¼ Binoðh2; Nf; Xb1 i¼0 mbi xb ð1 pÞaDþbi þ X Db1 i¼0 mTi xb ð1 pÞaDþb1Þ:
Prob½R1¼ r1 is the probability that r1sensors in S1 have
ever succeeded in detecting this event before the (aT +
b)th slot, where 1 (1 p)aD is the probability that such
a sensor has ever successfully detected this event before the (aT + b)th slot. Prob½R2¼ h1 r1 is derived similarly,
except that we are concerned about sensors in S2 and,
among these sensors, there is a ratio ofmTðD1Þþi
jS2j of sensors
which have tried to detect this event for aD + i slots (and we take their average). Prob[N2= h2] is the probability that
there are h2sensors among xbsensors successfully detecting
the event at the (aT + b)th slot. Prob[N3= h3] is the
prob-ability that there are h3sensors among xb h2sensors that
have ever successfully detected the event before the (aT + b)th slot. Note that the third term in Bino(Æ) is to take care of those sensors whose active slots do not (the first expression) and do (the second expression) cross the boundaries of system cycles, and we take their average.
Prob[NfP k Ne] is similar to the previous probability
ex-cept that these sensors succeed for the first time at the (aT + b)th slot.
Case (2): D 6 b 6 T D + 1. In this case, we divide the
set of inactive M(x, y) xbsensors at the (aT + b)th slot
into three subsets according to whether their active slots cross the boundaries of system cycles:
• S1: The set of sensors which have finished their active
slots in the current system cycle and whose active slots do not cross the boundaries of system cycles.
• S2: The set of sensors which have not started their
active slots in the current system cycle and whose active slots do not cross the boundaries of system cycles.
• S3: The set of sensors whose active slots cross the
boundaries of system cycles.
We can obtain that jS1j ¼PbDi¼1mi, jS2j ¼PTðD1Þi¼bþ1 mi,
and jS3j ¼PTi¼TðD1Þþ1mi. For example, when b = 4,
Fig. 3shows these subsets in case 2. Again, let R3be the
number of sensors belonging to S3. Since
R1+ R2+ R3= S1, we can expand Eq.(4) as follows:
Pkðm1; . . . ; mT; aTþ bÞ ¼X k1 Ne¼0 XNe h1¼0 Xh1 r1¼0 X h1r1 r2¼0 Prob½R1¼ r1 Prob½R2¼ r2 Prob½R3¼ h1 r1 r2 ! X xb h2¼0 Prob½N2¼ h2 X Neh1 h3¼0 Prob½N3¼ h3 Prob½Nf k Ne ; where
Prob½R1¼ r1 ¼ BinoðjS1j;r1;1 ð1 pÞðaþ1ÞDÞ;
Prob½R2¼ r2 ¼ BinoðjS2j;r2;1 ð1 pÞ aD Þ; Prob½R3¼ h1 r1 r2 ¼X D2 i¼0 mTðD1Þþ1þi jS3j BinoðjS3j; h1 r1 r2;1 ð1 pÞaDþiþ1Þ; Prob½N2¼ h2 ¼ Binoðxb; h2; pÞ; Prob½N3¼ h3 ¼ Binoðxb h2; h3; XD1 i¼0 mbi xb ð1 ð1 pÞaDþiÞÞ; and Prob½NfP k Ne ¼ Binoðh2; Nf; XD1 i¼0 mbi xb ð1 pÞaDþiÞ:
Again, Prob½R3¼ h1 r1 r2 is the probability that
h1 r1 r2sensors in S3 have ever succeeded in
detect-ing this event before the (aT + b)th slot, where
1 (1 p)aD+i+1 is the probability that such a sensor
have ever successfully detected this event before the
(aT + b)th slot. However, among these sensors in S3,
there is a ratio of mTðD1Þþ1þi
jS3j of sensors which have tried
to detect this event for aD + i slots, and thus we take their average.
Case (3): b > T D + 1. In this case, we divide the set of
inactive M(x, y) xbsensors at the (aT + b)th slot into two
subsets according to whether their active slots cross the boundaries of system cycles:
• S1: The set of sensors whose active slots do not cross the
boundaries of system cycles.
• S2: The set of sensors whose active slots cross the
boundaries of system cycles.
We have S1¼PbDi¼1mi and S2¼PTi¼bþ1mi. Fig. 3
gives an example when b = 7. We derive Eq. (4) as
follows: Pkðm1; . . . ; mT; aTþ bÞ ¼X k1 Ne¼0 XNe h1¼0 Xh1 r1¼0 Prob½R1¼ r1 Prob½R2¼ h1 r1 ! X xb h2¼0 Prob½N2¼ h2 X Neh1 h3¼0 Prob½N3¼ h3 Prob½Nf P k Ne ; where
Prob½R1¼ r1 ¼ BinoðjS1j; r1;1 ð1 pÞðaþ1ÞDÞ; Prob½R2¼ h1 r1 ¼X D2 i¼1 mTðD1Þþ1þi jS2j BinoðjS2j; h1 r1;1 ð1 pÞaDþiþ1Þ; Prob½N2¼ h2 ¼ Binoðxb; h2; pÞ; Prob½N3¼ h3 ¼ Binoðxb h2; h3; XTb i¼0 mTðD1Þi xb ð1 ð1 pÞ aDþiþ1 Þ þ X bTþðD1Þ1 i¼0 mTðD1Þþ1þi xb ð1 ð1 pÞ aDþiþ1ÞÞ; and
Prob½Nf P k Ne ¼ Binoðh2; Nf;X Tb i¼0 mTðD1Þi xb ð1 pÞ aDþiþ1 þ X bTþðD1Þ1 i¼0 mTðD1Þþ1þi xb ð1 pÞ aDþiþ1 Þ:
Finally, by replacing Pk(m1, . . . , mT, aT + b) in Eq. (3)
with one of the above three cases, we can obtain the expected latency d(m1, . . . , mT) under the
k-sensor-detec-tion model.
Table 2 summarizes the four terms Prob[N1= h1],
Prob[N2= h2], Prob[N3= h3], and Prob[Nf P k Ne] in
Eq.(4) under the three cases.
3. Using detection latency to guide deployment
Event detection latency can be used as an index to evaluate a WSN’s coverage and thus can help guide the deployment of a WSN. Below, we briefly discuss how to improve the coverage of a WSN. First, we can partition the sensing field into several subregions. Then, we can evaluate the event detection latency of each sub-region. If the expected latency of a region is larger than a tolerable threshold, it means that there are not enough sensors deployed in the region. For such regions, we can deploy more sensors to improve their expected detection latencies.
Beside, the event detection latency can also be used to measure the latency to detect a node newly joining a wireless personal area network (WPAN). We observe that for a device to join a WPAN, usually a network discovery procedure needs to be taken. To facilitate network discovery, coordinators in a WPAN normally need to send beacons periodically to announce their
presence (for example, Bluetooth, WiMedia [22], and
ZigBee follow this model). If we regard the beacon windows as our active phases, then the event detection latency under our any-sensor-detection model is the latency for a new node to discover the WPAN.
4. Simulation results
We have developed a simulator using C++ language to verify our analytical results. In the simulations, we set up a
sensing field of size 10· 10, on which there are 50 sensors
randomly deployed. Each sensor has a sensing distance of 3 units. Events may appear in any location inside the sens-ing field. Given a network configuration, we evaluate the
event detection latency by both Eq.(1)and the simulations.
For each simulation, at least 1000 experiments are repeated, and we take their average.
Fig. 4shows the event detection latencies under different values of detection probability p. The simulation results coincide well with the analytical results, except when p = 0.1 under the 5-sensor-detection model. This is because the simulator only simulates 1000 possible locations that an
event may occur, while the analysis (Eq. (1)) has to
con-sider all possible locations inside the sensing field. Since the value of p is small, the network requires longer time to capture the event than we expect.
Fig. 5shows the event detection latencies under different values of M(x, y). In the simulation, when an event occurs within i sensors’ sensing ranges, we record the detection latency in the corresponding M(x, y) = i statistics. From Fig. 5, we can observe that the simulation results coincide well with the analytical results, except when p = 0.1 and M(x, y) 6 5 under 3-sensor-detection model. This is because our analysis assumes a larger-scale network. It can be observed that a larger M(x, y), which implies a higher network density, can help reduce the detection latency. A larger detection probability p, which reflects the sensibility of sensors, can also reduce the detection latency. The result can be used to determine how sensors should be arranged at the deployment stage.
In both Figs. 4 and 5, we can observe that the event
detection latency can be greatly reduced when we increase the number of active slots D, especially when the detection probability p is small. Thus, we have interest in observing
Table 2
Summary of the four terms Prob[N1= h1], Prob[N2 = h2], Prob[N3= h3], and Prob[NfP k Ne] in Eq.(4)under different cases, where q = 1 p
Cases Terms
b < D Prob½N1¼ h1 ¼Prh11¼0BinoðjS1j; r1;1 q
aDÞ PD2
i¼1 mTDþiþ1jS2j BinoðjS2j; h1 r1;1 q
aDþiÞ
Prob[N2= h2] = Bino(xb, h2, p)
Prob½N3¼ h3 ¼ Binoðxb h2; h3;Pb1i¼0mxbib ð1 q
aDþbiÞ þPDb1
i¼0 mxTib ð1 q
aDþbÞÞ
Prob½NfP k Ne ¼ Binoðh2; Nf;Pb1i¼0mxbib q
aDþbiþPDb1 i¼0 mxTib q aDþb1Þ D 6 b 6 T D + 1 Prob½N1¼ h1 ¼Phr11¼0 Ph1r1 r2¼0BinoðjS1j; r1;1 q ðaþ1ÞDÞ BinoðjS
2j; r2;1 qaDÞ Pb2i¼0mTDþiþ2jS3j Binoðj S3j; h1 r1 r2;1 q
aDþiþ1Þ
Prob[N2= h2] = Bino(xb, h2, p)
Prob½N3¼ h3 ¼ Binoðxb h2; h3;PD1i¼0 mxbib ð1 q
aDþiÞÞ
Prob½NfP k Ne ¼ Binoðh2; Nf;PD1i¼0 mxbib q
aDþiÞ
b > T D + 1 Prob½N1¼ h1 ¼Prh11¼0BinoðjS1j; r1;1 q
ðaþ1ÞDÞ PD2
i¼1 mTDþiþ2jS2j BinoðjS2j; h1 r1;1 q
aDþiþ1Þ
Prob[N2= h2] = Bino(xb, h2, p)
Prob½N3¼ h3 ¼ Binoðxb h2; h3;PTbi¼0TDiþ1xb ð1 q
aDþiþ1Þ þPbþDT2
i¼0 mTDþiþ2xb ð1 q
aDþiþ1ÞÞ
Prob½NfP k Ne ¼ Binoðh2; Nf;PTbi¼0TDiþ1xb q
aDþiþ1þPbþDT2 i¼0 mTDþiþ2xb q
the effect of D on the event detection latency under different
values of M(x, y) and p, as shown inFig. 6. To show the
effect of D, we set the period T as a constant of 16 slots.
From Fig. 6, we can observe that the latencies drop as
the value of D increases, but this effect becomes less signif-icant when D P 4. Since a sensor will consume more
(a) T = 5 and D = 1 (b) T = 5 and D = 2
(c) T = 5 and D = 1 (d) T = 5 and D = 2 0 5 10 15 20 25 30 35 40 45 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Probability p
Detection latency (slots)
any-sensor model (analysis) any-sensor model (simulation) 5-sensor model (analysis) 5-sensor model (simulation)
0 5 10 15 20 25 30 35 40 45 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Probability p
Detection latency (slots)
0 5 10 15 20 25 30 35 40 45 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Probability p
Detection latency (slots)
0 5 10 15 20 25 30 35 40 45 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Probability p
Detection latency (slots)
any-sensor model (analysis) any-sensor model (simulation) 5-sensor model (analysis) 5-sensor model (simulation)
any-sensor model (analysis) any-sensor model (simulation) 5-sensor model (analysis) 5-sensor model (simulation)
any-sensor model (analysis) any-sensor model (simulation) 5-sensor model (analysis) 5-sensor model (simulation)
Fig. 4. The event detection latencies under different values of probability p.
0 3 6 9 12 15 4 9 10 11 12 M (x,y)
Event detection latency (slots)
Event detection latency (slots)
p = 0.1 (simulation) p = 0.1 (analysis) p = 0.5 (simulation) p = 0.5 (analysis) p = 0.9 (simulation) p = 0.9 (analysis) 0 3 6 9 12 15 p = 0.1 (simulation) p = 0.1 (analysis) p = 0.5 (simulation) p = 0.5 (analysis) p = 0.9 (simulation) p = 0.9 (analysis) 0 10 20 30 40 50 60 p = 0.1 (simulation) p = 0.1 (analysis) p = 0.5 (simulation) p = 0.5 (analysis) p = 0.9 (simulation) p = 0.9 (analysis) 0 10 20 30 40 50 60 p = 0.1 (simulation) p = 0.1 (analysis) p = 0.5 (simulation) p = 0.5 (analysis) p = 0.9 (analysis) p = 0.9 (simulation)
(a) T = 5 and D = 1 (any-sensor model) (b) T = 5 and D = 2 (any-sensor model)
(c) T = 5 and D = 1 (3-sensor model) (d) T = 5 and D = 2 (3-sensor model)
Event detection latency (slots)
Event detection latency (slots)
5 6 7 8 4 9 10 11 12 M (x,y) 5 6 7 8 4 9 10 11 12 M (x,y) 5 6 7 8 4 9 10 11 12 M (x,y) 5 6 7 8
energy as the length of active slots D increases, this result can be used to decide the length of a sensor’s active phase to reduce both event detection latency and energy con-sumption of a WSN.
5. Conclusions
We have proposed a methodology to analyze the event detection latency of a WSN. Such a latency analysis can be used to measure the network coverage and the time that a new node needs to discover a network. We have adopted a probabilistic approach to analyze the latency under an
any-sensor-detection and a k-sensor-detection models. We have also developed a simulator to verify our analyses. Simulation results not only coincide well with the analyses, but also indicate the potential factors that affect the latency.
Our analysis assumes that the detection probability p is a constant. It deserves to further study the same problem when the value of p is a function of the dis-tance between a sensor and the event. Also, our analysis models time in a discrete manner (by fixed-length slots). It is also interesting to investigate the continuous time case.
Acknowledgements
Y.C. Tseng’s research is co-sponsored by Taiwan’s MOE ATU Program, by NSC under Grant Nos. 93-2752-E-007-001-PAE, 95-2623-7-009-010-ET,
95-2218-E-009-020, 95-2219-E-009-007, 94-2213-E-009-004, and
94-2219-E-007-009, by MOEA under Grant No.
94-EC-17-A-04-S1-044, by ITRI, Taiwan, and by Intel Inc.
References
[1] I.F. Akyildiz, W. Su, Y. Sankarasubramaniam, E. Cayirci, A survey on sensor networks, IEEE Communications Magazine 40 (8) (2002) 102–114.
[2] D. Niculescu, Communication paradigms for sensor networks, IEEE Communications Magazine 43 (3) (2005) 116–122.
[3] J. Burrell, T. Brooke, R. Beckwith, Vineyard computing: sensor networks in agricultural production, IEEE Pervasive Computing 3 (1) (2004) 38–45.
[4] K. Chintalapudi, T. Fu, J. Paek, N. Kothari, S. Rangwala, J. Caffrey, R. Govindan, E. Johnson, S. Masri, Monitoring civil structures with a wireless sensor network, IEEE Internet Computing 10 (2) (2006) 26– 34.
[5] T.T. Hsieh, Using sensor networks for highway and traffic applica-tions, IEEE Potentials 23 (2) (2004) 13–16.
[6] A. Mainwaring, D. Culler, J. Polastre, R. Szewczyk, J. Anderson, Wireless sensor networks for habitat monitoring, in: ACM Interna-tional Workshop on Wireless Sensor Networks and Applications, 2002, pp. 88–97.
[7] R. Szewczyk, E. Osterweil, J. Polastre, M. Hamilton, A. Mainwaring, D. Estrin, Habitat monitoring with sensor networks, Communica-tions of the ACM 47 (6) (2004) 34–40.
[8] P.K. Biswas, S. Phoha, Self-organizing sensor networks for integrated target surveillance, IEEE Transactions on Computers 55 (8) (2006) 1033–1047.
[9] T. He, S. Krishnamurthy, L. Luo, T. Yan, L. Gu, R. Stoleru, G. Zhou, Q. Cao, P. Vicaire, J.A. Stankovic, T.F. Abdelzaher, J. Hui, B. Krogh, VigilNet: an integrated sensor network system for energy-efficient surveillance, ACM Transactions on Sensor Networks 2 (1) (2006) 1–38.
[10] Y.C. Tseng, Y.C. Wang, K.Y. Cheng, An integrated mobile surveillance and wireless sensor (iMouse) system and its detection delay analysis, in: ACM International Symposium on Modeling, Analysis and Simulation of Wireless and Mobile Systems, 2005, pp. 178–181.
[11] C. Gui, P. Mohapatra, Power conservation and quality of surveillance in target tracking sensor networks, in: ACM Inter-national Conference on Mobile Computing and Networking, 2004, pp. 129–143.
[12] H.T. Kung, D. Vlah, Efficient location tracking using sensor networks, in: Wireless Communications and Networking, 2003, pp. 1954–1961.
[13] W. Zhang, G. Cao, DCTC: dynamic convoy tree-based collaboration for target tracking in sensor networks, IEEE Transactions on Wireless Communications 3 (5) (2004) 1689–1701.
[14] LAN/MAN Standards Committee of the IEEE Computer Society, IEEE Std 802.15.4-2003, Wireless medium access control (MAC) and physical layer (PHY) specifications for low-rate wireless personal area networks (LR-WPANs), IEEE.
[15] LAN/MAN Standards Committee of the IEEE Computer Society, IEEE Std 802.15.1-2005, Wireless medium access control (MAC) and physical layer (PHY) specifications for wireless personal area networks (WPANs), IEEE.
[16] W. Ye, J. Heidemann, D. Estrin, Medium access control with coordinated adaptive sleeping for wireless sensor networks, IEEE/ ACM Transactions on Networking 12 (3) (2004) 493–506.
[17] S.S. Dhillon, K. Chakrabarty, Sensor placement for effective coverage and surveillance in distributed sensor networks, in: IEEE Wireless Communications and Networking, 2003, pp. 1609–1614.
[18] Y. Zou, K. Chakrabarty, A distributed coverage- and connectivity-centric technique for selecting active nodes in wireless sensor networks, IEEE Transactions on Computers 54 (2005) 978–991. [19] A. Nasipuri, K. Li, A directionality based location discovery scheme
for wireless sensor networks, in: ACM International Workshop on Wireless Sensor Networks and Applications, 2002, pp. 105–111. [20] D. Nicules, B. Nath, Ad-hoc positioning system (APS) using AoA, in:
IEEE INFOCOM, 2003, pp. 1734–1743.
[21] A. Savvides, C.C. Han, M.B. Strivastava, Dynamic fine-grained localization in ad-hoc networks of sensors, in: International Confer-ence on Mobile Computing and Networking, 2001, pp. 166–179. [22] LAN/MAN Standards Committee of the IEEE Computer Society,
IEEE Std 802.15.3-2003, Wireless medium access control (MAC) and physical layer (PHY) specifications for high rate wireless personal area networks (WPANs), IEEE.
You-Chiun Wang received his B.S. and M.S. degrees in Computer Science and Information Engineering from the National Chung-Cheng University and the National Chiao-Tung Univer-sity, Taiwan, in 2001 and 2003, respectively. He is currently a Ph.D. candidate in the Department of Computer Science, National Chiao-Tung Univer-sity, Taiwan. His research interests include wireless communication and mobile computing, QoS management and wireless fair scheduling, mobile ad hoc network, and wireless sensor networks.
Kai-Yang Cheng received his B.S. and M.S. degrees in Computer Science from the National Tsing-Hua University and the Chiao-Tung Uni-versity in 2003 and 2005, Taiwan, respectively. His research interests include wireless communi-cation and sensor networks.
Yu-Chee Tseng received his B.S. and M.S. degrees in Computer Science from the National Taiwan University and the National Tsing-Hua University in 1985 and 1987, respectively. He obtained his Ph.D. in Computer and Information Science from the Ohio State University in January of 1994. He was an Associate Professor at the Chung-Hua University (1994–1996) and at the National Cen-tral University (1996–1999), and a Professor at the National Central University (1999–2000). Since 2000, he has been a Professor at the Department of Computer Science, National Chiao-Tung University, Taiwan, where he is currently the Chairman.
Dr. Tseng served as a Program Chair in the Wireless Networks and Mobile Computing Workshop, 2000 and 2001, as a Vice Program Chair in the International Conference on Distributed Computing Systems (ICDCS), 2004, as a Vice Program Chair in the IEEE International Conference on Mobile Ad-hoc and Sensor Systems (MASS), 2004, as an Associate Editor for The Computer Journal, as a Guest Editor for ACM Wireless Networks special issue on ‘‘Advances in Mobile and Wireless Systems’’, as a Guest Editor for IEEE Transactions on Computers special on ‘‘Wireless Internet’’, as a Guest Editor for Journal of Internet Technology special issue on ‘‘Wireless Internet: Applications and Systems’’, as a Guest Editor for Wireless Communications and Mobile Computing special issue on ‘‘Research in Ad Hoc Networking, Smart Sensing, and Pervasive Com-puting’’, as an Editor for Journal of Information Science and Engineering, as a Guest Editor for Telecommunication Systems special issue on ‘‘Wireless Sensor Networks’’, and as a Guest Editor for Journal of Information Science and Engineering special issue on ‘‘Mobile Computing’’. Dr. Tseng received the Outstanding Research Award, by National Science Council, ROC, in both 2001–2002 and 2003–2005, the Best Paper Award, by International Conference on Parallel Processing, in 2003, the Elite I.T. Award in 2004, and the Distinguished Alumnus Award, by the Ohio State University, in 2005. His research interests include mobile computing, wireless communication, network security, and parallel and distributed computing. Dr. Tseng is a member of ACM and a Senior Member of IEEE.