Professional Documents
Culture Documents
Abstract—This paper presents a model for predicting radio present, for example, in a battlefield environment or disastrous
frequency (RF) propagation for Wireless Sensor Network (WSN) areas.
deployment in a dense tree environment. To create the model,
data from a physical deployment are collected and an empirical Wireless sensor nodes may be deployed in a stochastic or
path loss prediction model is derived from the actual deterministic fashion. In the latter deployment approach, the
measurements. Furthermore, the presented measurements and distance separating sensors are carefully planned based on
empirical path loss model are compared with measurements and multiple factors including sensing functionality and
models obtained from WSN deployments in other terrains, such transmission range of sensor nodes, for example, see [1] and
as one characterized by long-grass and another by sparse-tree [2]. On the other hand, in stochastic deployments, the
environments. The results from the comparison of these
different terrains show significant differences in path loss and
deployment process can be modeled as a random process
empirical models’ parameters. In addition, the proposed model driven (in large part) by prediction of the signal propagation
is compared with Free Space Path Loss (FSPL) and Two-Ray [3], [4]. In both cases, accurate RF propagation models are of
models to demonstrate the inaccuracy of these theoretical models great importance to facilitate planning and deployment of
in predicting path loss between wireless sensor nodes deployed in sensor networks.
dense tree environment.
For practical applications, prediction of the WSN
Keywords—Wireless Sensor Networks; Radio Frequency deployment performance critically depends on one’s ability to
Propagation Models; Path Loss Models; Channel Modeling; RF model the propagation of the radio signal between the
Propagation transmitter and receiver nodes, which depends fundamentally
on the type of terrain and the type of objects and foliage on
I. INTRODUCTION that terrain [5]. Therefore, a formal study is necessary for
Since the beginning of the third millennium, Wireless understanding this behavior and for accurate characterization
Sensor Network (WSN) technology has attracted a growing of the terrain-dependent behavior of signal propagation. This
attention from the research and commercial communities. It paper presents RF measurements and a model for prediction of
became practical as a result of advancements in hardware and path loss in wireless sensor nodes that are deployed in a dense
wireless communication technologies. WSNs facilitate tree environment.
monitoring of physical and environmental phenomena in areas
where human presence may be undesired or impossible. Such II. RELATED WORK
networks typically consist of miniature, low-power, low-cost, Despite its importance, the subject of propagation
autonomous sensor nodes that communicate wirelessly over modeling in WSN research literature has been largely
short distances and that are commonly deployed across large neglected. Most of published work adopts one of the two
geographical areas. simplistic propagation models: free space propagation model
or two-ray propagation model, for example, see [6], [7], [8],
Wireless sensor nodes disseminate packets and
and [9]. Both of these models have been proven to be
communicate between one another in an ad hoc fashion. In ad
inaccurate in WSN propagation environments [10], [11]. The
hoc networks, data packets are relayed between nodes to
fundamental assumption behind the free space model is that
extend the reach of the network. This allows the network to
the transmitter and receiver have a line-of-sight (LOS)
cover large geographical areas, which otherwise would be
communication with no obstructions or reflections of any
impossible with direct communication between the sender and
kind. In the case of WSN, the dominant presence of the
the receiver. Ad hoc networks are more likely to be used in
ground affects the propagation of the radio waves in a very
situations where preset communication infrastructure is not
significant manner [12]. On the other hand, the two-ray model
III. EXPERIMENTAL SETUP: The RSS values collected in the dense tree environment
were converted into path loss values. The path loss is a
RF propagation measurements were made at the Botanical positive quantity that represents signal attenuation on the path
Garden at Florida Institute of Technology, see Fig. 1. The between the transmitter and the receiver antennas. In other
Garden has a forest-like environment. The field is densely words, the path loss is a property of the environment and is
filled with tall trees having thick trunks. In almost all RSS largely independent from the system parameters. For this
reason, path loss is usually preferred over RSS. Therefore, for
the rest of this paper, and for the sake of simplicity, the Path Loss Readings - Dense Tree
discussion and the focus are placed on the path loss and path
140
loss exponent rather than RSS. Generally, the relationship
between path loss, transmitted power, and RSS is given as: 130
100
7 135 96 91 99 102 104 111 115 138
95
8 157.5 79 99 95 115 118 127 128 119
90 Average Path Loss Values
Linear Regression Line
9 180 90 91 97 100 114 117 113 116 85
y = 40.15*x + 52.14
10 202.5 81 97 109 117 118 115 110 115 80
0.8 1 1.2 1.4 1.6 1.8 2
Log ( Distance [m] )
11 225 82 108 111 102 103 109 126 118
12 247.5 85 86 94 100 104 123 116 117 Fig. 3. Average path loss and linear regression line and equation.
13 270 99 99 108 114 121 131 114 120 It is clearly seen that the variations among path loss
14 292.5 83 105 113 100 102 122 135 136
readings that share same distances is relatively large. In dense
tree environments, such variations mainly occur as a result of
15 315 101 90 108 117 117 111 132 135 the absence of LOS communication between the transmitter
and receiver, resulting in great signal fluctuation. In some
16 337.5 79 85 97 100 100 110 113 118
instances, multiple waves of the transmitted signal may arrive
at the receiver, causing multipath fading. In other instances,
TABLE II. Standard deviation and average path loss values for the dense only a small scattered portion of the transmitted signal may be
tree environment
received.
Distance [m] 5 10 15 20 25 30 35 40 From Fig. 3, however, one may clearly see that the average
path loss values closely follow the linear regression line,
Average Path Loss [dB] 83 90 99 103 106 112 115 119 which indicates a dependency between the path loss values
and the log of distance of the data collected in this experiment.
Standard Dev. [dB] 7.2 6.9 7.3 7.3 7.5 8.0 8.8 9.0
In fact, the linear regression provided in Fig. 3 produces a
reliable coefficient of determination (i.e., R² = 0.9796), which
Fig. 2 illustrates the various path loss values that are indicates how well the data points fit the linear regression line.
presented in TABLE I. Meanwhile, Fig. 3 illustrates the
From the tables and figures presented so far, it is clear that
average path loss values presented in TABLE II versus the log
path loss increases as a function of distance and that variation
of distance. In addition, it illustrated the linear regression line
in path loss exist. To model the variations in path loss, the
and its equation.
log-normal shadowing equation may be used [16], as follows:
𝑑 Fig. 5 compares the average path loss values versus the log
𝐿𝑝 = 𝐿0 + 10𝛼 𝑙𝑜𝑔 ( ) + 𝑋𝜎 (2) of distance for the dense-tree, long-grass, and sparse-tree
𝑑0
environments. In addition, linear regression line and its
where 𝐿0 is the path loss value at the reference distance
equation are provided for each environment.
(𝑑0 ) in dB, 𝛼 is the path loss exponent, which represents the
rate of the path loss increase as a function of log of distance,
and 𝑋𝜎 is the normally distributed random variable with zero
mean and σ standard deviation. The linear regression equation
as it appears in Fig. 3 is:
TABLE IV. Path loss results of the three environments (in dB) Sparse-tree 33.36 log(d) + 60.84 3.34 60.84 7.30
Environment 5m 10m 15m 20m 25m 30m 35m 40m Dense-tree 40.15 log(𝑑) + 52.14 4.02 52.14 7.75