scispace - formally typeset
Search or ask a question
Proceedings ArticleDOI

Ontology-driven adaptive sensor networks

TL;DR: A novel, two-phase solution to the wireless sensor network adaptivity problem, where nodes in the network, organized as clusters, execute an efficient algorithm to dynamically calibrate sensed data.
Abstract: A wireless sensor network deployed in an area of interest is affected by variations in environmental conditions associated with that area. It must adapt to these variations in order to continue functioning as desired by the user. We present a novel, two-phase solution to the wireless sensor network adaptivity problem. In the first phase, nodes in the network, organized as clusters, execute an efficient algorithm to dynamically calibrate sensed data. Each node provides its current energy level and the state of each on-board sensor to a cluster-head. In the second phase, each cluster-head executes an efficient, ontology-driven algorithm to determine the future state of the network under existing conditions, based on information received from each sensor node. We describe an example application scenario to show how our two-phase solution can be employed to enable a real-world wireless sensor network to adapt itself to variations in environmental conditions.

Summary (3 min read)

1. Introduction

  • The potential applicability of ad-hoc, wireless sensor networks (WSNs) in a variety of domains, ranging from simple distributed monitoring (e.g., habitat and environmental monitoring) to complex surveillance (e.g., battlefield surveillance, homeland security) [4, 6] has fueled research interest in this area.
  • In security-related applications, such as battlefield surveillance and homeland security, users of WSNs require a high level of accuracy (close to 100%) and may specify tight accuracy bounds for all data reported by the network.
  • Thus, there is a need for sensor nodes to adapt to variations in their environment, if possible, and restore data accuracy levels to their original values.
  • Figure 1 shows the various components of this framework, and pinpoints the component discussed in this paper.
  • The authors discuss the sensor node ontology and the WSN state determination algorithm in Section 5.

2. Wireless Sensor Network Model

  • In their model, the WSN consists of a limited number of static or mobile BSs and RUs, but is otherwise unattended [1].
  • Each BS and RU can support hundreds of sensor nodes, which can range from the simplest, least expensive ones (e.g., Berkeley Motes, SmartDust) to medium sized sensor nodes.
  • Each node contains a protocol stack [1], which enables it to communicate with neighbors, establish connections with them and reach the nearest BS or RU.
  • Further, specifications provide information about energy consumption of each sensor type in each of these modes.
  • Capable sensor nodes can compute the accuracy of each sensor in its present mode of operation.

3. Application Scenario

  • The application the authors consider deploys the WSN in a hostile environment, i.e., a battlefield.
  • These energy-accuracy combinations are generated, prioritized and made available to WSN by the user.
  • There is considerable variation in humidity in the region over different seasons.
  • Their input-output characteristics are directly affected by frequency of input pressure and indirectly affected by temperature.
  • These sensors are deployed both on the bridge and on all paths leading to it, to detect sound generated by moving troops and vehicles.

4. Data Calibration Algorithm

  • The authors discuss the algorithm executed by capable sensor nodes and possibly BS/RU nodes to perform data calibration on each type of sensor whenever significant changes in environmental conditions occur.
  • The algorithm uses either the lookup table or function (Section 2) describing the effect of each core environmental variable on the input-output characteristics of each sensor to determine its expected output value.
  • Therefore, if temperature increases by 10°C, then sensitivity (which is the ratio of output to input of a sensor) changes by 10%, which implies that for the same input value, the output of the acoustic sensor either reduces or increases by 10%.
  • Individual sensor nodes cannot make this determination because they do not possess either the user-defined energy accuracy bounds or a global view of the existing state of the WSN.
  • The capable sensor node transmits the quadruple in a message to the BS/RU node functioning as the sensor node’s cluster-head.

5. Sensor Node Ontology

  • The sensor node ontology the authors have designed attempts to capture the most important features of a sensor node that describe its functionality and its current state.
  • The ontology describes the main components of a sensor node: processor, power supply, radio and sensor modules.
  • A common thread that runs through these modules is energy, both in terms of capacity and consumption.
  • Based on this information, the WSN can decide the most appropriate CPU operating mode for each sensor node or a class of sensor nodes, under existing environmental conditions.
  • Remaining energy capacity dictates the ability of the sensor node to continue functioning in a certain mode.

5.1. WSN State Determination Algorithm

  • This algorithm enables a WSN to dynamically determine its most appropriate state of operation under existing environmental conditions.
  • The operating state of a WSN reflects the corresponding states of sensor nodes and each type of sensor in the WSN.
  • In order to obtain a global view that encompasses all nodes in the cluster, each cluster-head obtains similar class-based views of all other clusters in the WSN computed by their corresponding cluster-heads.
  • The algorithm accomplishes this task by providing the class-based energy consumption and accuracy, along with user-specified accuracy and energy bounds, and sensor node locations (if available) as inputs to a statistical model pre- deployed on each BS/RU node.
  • Additional information output by the statistical model includes WSN accuracy and energy consumption, after nodes have adapted to environmental changes.

6.1. Location-based classification

  • This scheme requires either prior knowledge of sensor node locations (e.g., nodes deployed on the bridge) or mechanisms to determine node locations after deployment (e.g., GPS).
  • Using this classification scheme, BS/RU functioning as cluster-heads of all clusters in regions farthest from the bridge instruct their cluster to transition to the ACTIVE state.
  • From this point onward, sensors on-board nodes in these clusters actively sense and attempt to detect enemy presence and movement in the region.
  • It uses the statistical model to compute the most appropriate operating mode for each sensor on each node in the cluster.
  • This becomes necessary if the accuracy of the previous ring of clusters falls below user-defined bounds; otherwise the next ring is not activated.

6.2. Classification Based on Sensor Type

  • As discussed earlier, the main sensor types on the bridge are pressure pad sensors, vibration sensors and radioactive sensors supported by optical sensors on either side of the bridge.
  • Thus, initially only pressure pad and vibration sensors on the bridge are activated along with the optical sensors.
  • Each BS/RU computes cumulative accuracies of these sensors separately and determines their most appropriate mode.
  • If they detect heavy vehicles with high accuracy then radioactive sensors are activated immediately and their energy consumption is monitored to ensure that the energy-accuracy bounds are continuously met.
  • This process repeats until all vehicles capable of carrying nuclear weapons are identified.

Did you find this useful? Give us your feedback

Content maybe subject to copyright    Report

Ontology-driven Adaptive Sensor Networks
Sasikanth Avancha Chintan Patel
Anupam Joshi
Department of Computer Science and Electrical Engineering,
1000 Hilltop Circle, University of Maryland Baltimore County, Baltimore, MD 21250
savanc1, cpatel2, joshi
@cs.umbc.edu
Abstract
A wireless sensor network deployed in an area of inter-
est is affected by variations in environmental conditions as-
sociated with that area. It must adapt to these variations
in order to continue functioning as desired by the user. We
present a novel, two-phase solution to the wireless sensor
network adaptivity problem. In the first phase, nodes in the
network, organized as clusters, execute an efficient algo-
rithm to dynamically calibrate sensed data. Each node pro-
vides its current energy level and the state of each on-board
sensor to a cluster-head. In the second phase, each cluster-
head executes an efficient, ontology-driven algorithm to de-
termine the future state of the network under existing condi-
tions, based on information received from each sensor node.
We describe an example application scenario to show how
our two-phase solution can be employed to enable a real-
world wireless sensor network to adapt itself to variations
in environmental conditions.
1. Introduction
The potential applicability of ad-hoc, wireless sensor
networks (WSNs) in a variety of domains, ranging from
simple distributed monitoring (e.g., habitat and environ-
mental monitoring) to complex surveillance (e.g., battle-
field surveillance, homeland security) [4, 6] has fueled
research interest in this area. In the near future WSNs
are expected to consist of sensor nodes containing on-
board micro-electromechanical systems (MEMS) as sen-
sors, which we expect to also be remotely programmable.
In security-related applications, such as battlefield
surveillance and homeland security, users of WSNs re-
quire a high level of accuracy (close to 100%) and may
specify tight accuracy bounds for all data reported by the
network. For example, military personnel obtaining in-
formation from a WSN deployed in a battlefield must
be assured that the network accurately reports the pres-
ence or absence of the enemy assets in the area. If the
WSN cannot report data at user-specified accuracy lev-
els or within the bounds, then the data would be considered
useless.
WSNs consist of sensor nodes of various capabilities, in-
cluding base stations (BS) and powerful, “rich uncle” (RU)
nodes, which in turn consist of one or more types of sen-
sors which perform the sensing function. The functioning
of each type of on-board sensor is affected by external en-
vironmental conditions, such as temperature, pressure and
humidity. Additionally, each sensor node is affected by its
own internal environment consisting of variables such as en-
ergy level and available memory. In both these cases, accu-
racy of data reported by the sensor node is adversely af-
fected. Thus, there is a need for sensor nodes to adapt to
variations in their environment, if possible, and restore data
accuracy levels to their original values.
An additional issue in security-related applications and
hostile situations is that the entire WSN is affected by the
prevailing security conditions in the area of deployment.
Thus, the WSN must adapt itself as a whole to variations
in security conditions in addition to environmental condi-
tions so that it can continue to perform its job as desired by
the user. This issue will be addressed in future work.
We address the wireless sensor network adaptivity
problem associated with an established WSN and propose
a two-phase approach as a solution.
In the first phase, sensor nodes possessing moderate
computational and storage capabilities adapt themselves to
variations in environmental variables by dynamically cal-
ibrating data from on-board sensors (Section 4). (We re-
fer to these nodes as capable sensor nodes in the rest of
this paper.) Based on the magnitude of the error between
expected and observed output values of on-board sensors,
these nodes determine the most appropriate state of opera-
tion of each sensor (Section 2) which they communicate to
BS/RU along with the data.
In the second phase, BS/RU employ a pre-deployed sen-

Combine Protocols
Evaluate & Score Protocol Combinations
Design Calibration Algorithms
Create Environmental Descriptions
Create Event Descriptions
Create Security Policy
Self-organization Protocols
Data Aggregation Protocols
Power Management Techniques
Security Protocols
External Environmental Variables
Internal Environmental Variables
Event Variables
Desired Security Level
Energy Constraints
Protocol Library
Calibration Algorithm Library
Sensor Node Ontology
Inference
Engine
Secure Adaptive Wireless
Sensor Network
Figure 1. Framework for Secure Adaptive
Wireless Sensor Networks
sor node ontology (Section 5) written in OWL-Lite [9] to
reason over data received from sensor nodes and data that
the BS/RU nodes have themselves sensed. Using data ob-
tained from their sensors, BS/RU nodes calibrate data ob-
tained from resource-poor nodes which did not perform
data calibration before transmission. By reasoning over cal-
ibrated data BS/RU nodes collectively determine the most
appropriate operating state of the WSN under existing en-
vironmental conditions and instruct sensor nodes to operate
at that state.
Our work on adaptive wireless sensor networks, dis-
cussed in this paper, is part of a larger effort to build a
framework to enable the design of secure wireless sensor
networks that can adapt to changing environmental, topo-
logical and security conditions. Figure 1 shows the vari-
ous components of this framework, and pinpoints the com-
ponent discussed in this paper. Our goal is to build, sim-
ulate and evaluate the entire framework as a whole. This
precludes us from simulating the component discussed in
this paper in isolation and providing empirical results. In-
stead, we present a complete application scenario, and dis-
cuss how our two-phase approach is a solution to the wire-
less sensor network adaptivity problem.
The rest of the paper is organized as follows. In Sec-
tion 2, we discuss our WSN model including capabilities of
BS/RU and sensor nodes. In Section 3, we lay out the ap-
plication scenario to which we can apply our solution. In
Section 4, we describe the data calibration algorithm in de-
tail. We discuss the sensor node ontology and the WSN state
determination algorithm in Section 5. In Section 6, we dis-
cuss examples of classification schemes that can be imple-
mented in the application scenario. In Section 7 we present
related work in the area of adaptive sensor networks. We
conclude and present directions for future work in Section
8.
2. Wireless Sensor Network Model
In our model, the WSN consists of a limited number of
static or mobile BSs and RUs, but is otherwise unattended
[1]. Each BS and RU can support hundreds of sensor nodes,
which can range from the simplest, least expensive ones
(e.g., Berkeley Motes, SmartDust) to medium sized sen-
sor nodes. Each BS and RU node possesses significantly
higher computational, storage and communication capacity
compared to sensor nodes. The sensor node ontology, which
completely describes a sensor node in terms of its individ-
ual components and their characteristics, is pre-deployed on
each BS and RU. Each BS and RU possesses sensor inter-
faces to sense a set of core environmental variables, such as
temperature, pressure, humidity and wind velocity.
Each sensor node possesses limited computational, stor-
age and communication capabilities. Each node contains a
protocol stack [1], which enables it to communicate with
neighbors, establish connections with them and reach the
nearest BS or RU. Thus, in our model, the WSN consists of
clusters of sensor nodes, in which either a BS or RU plays
the role of a cluster-head [2, 3].
As discussed in Section 1, each sensor node consists of
one or more sensor types. The following operating modes
are associated with each sensor type, based on information
provided in specifications (e.g., data sheets) for that type:
NORMAL, HIGH-SAMPLING, VERY-LOW-SAMPLING
and OFF. Further, specifications provide information about
energy consumption of each sensor type in each of these
modes. Capable sensor nodes can compute the accuracy of
each sensor in its present mode of operation. Therefore, the
state of each on-board sensor is described by the follow-
ing quadruple:

, where
,
,
and
stand
for sensor type, operating mode, energy consumed and ac-
curacy, respectively. Each sensor node tracks its overall en-
ergy consumption and knows the latest value of the remain-
ing energy level. Thus, dynamic information about a sensor
node can be summarized by the following tuple:
E,
S

,
where E stands for the node’s present energy level and
S
is a set of one or more quadruples, describing the state of
each on-board sensor.
Capable sensor nodes store a limited amount of infor-
mation about input-output characteristics of each on-board
sensor in the form of lookup tables. Input-output character-
istics are a strictly one-to-one mapping between input and
output values of the sensor. Similarly, they store the range
of expected output values from each sensor type for differ-
ent values of each of the core environmental variables, ei-

ther in a lookup table or as an executable function. Nomi-
nal values of core environmental variables are pre-deployed
on each sensor node. All sensor nodes periodically receive
current values of core environmental variables broadcast by
BS/RU nodes functioning as cluster-heads. Capable nodes
employ these values during data calibration (Section 4).
Each sensor node is in one the following three states at
all times: ACTIVE, LOW-POWER and INACTIVE. Upon
deployment, all sensor nodes discover each other and estab-
lish secure communication channels with each other. Sub-
sequently, all nodes transition to the LOW-POWER state.
All nodes in our model contain the required security mech-
anisms and protocols required to establish secure commu-
nications [2].
We emphasize that the WSN adaptivity problem is mean-
ingful and therefore considered, only for an established net-
work. Thus, issues of routing, transport, power management
and data aggregation are dealt with in other work [1, 2, 3]
and are beyond the scope of this paper.
3. Application Scenario
In this section we describe a scenario that requires a de-
ployed WSN to adapt to changing environmental condi-
tions. The application we consider deploys the WSN in a
hostile environment, i.e., a battlefield. The task of the net-
work is surveillance of enemy forces. There exists a set of
user-defined energy consumption and accuracy bounds that
can be combined in different ways. These energy-accuracy
combinations are generated, prioritized and made available
to WSN by the user.
In this example, the geographical environment is a desert
region consisting of plains intersected by a river bed that can
be crossed using one or more bridges. The major task is to
detect troop/vehicle movement across the river bed and de-
termine, according to prescribed energy-accuracy bounds,
the types of troops, vehicles, weapons and their numbers.
Given that the network is deployed in a desert, the prin-
cipal environmental variable is temperature. The tempera-
ture in this region is very high in the afternoon and very low
at night. Chances of high wind velocity and frequent sand
storms are very high. There is considerable variation in hu-
midity in the region over different seasons. All BS/RU in
the WSN possesses high-quality sensors to detect these en-
vironmental conditions and periodically transmit their cur-
rent values to sensor nodes as appropriate.
Five types of sensors are deployed in the network. Each
type has unique input-output characteristics, dynamic char-
acteristics and responses to changes in environmental vari-
ables.
1. Vibration Sensors: These are fixed on the bridge to de-
tect sudden high vibrations caused by movement of
heavy vehicles (e.g.,. tanks, armored carriers) over the
bridge. Their input-output characteristics are directly
affected by inherent vibrations of the bridge and fre-
quency of input vibrations. The characteristics are in-
directly affected by temperature.
2. Pressure Pad Sensors: Pressure pads are embedded in
the ground on all paths leading to the bridge to de-
tect movement and estimate the weight the vehicles.
Their input-output characteristics are directly affected
by frequency of input pressure and indirectly affected
by temperature.
3. Acoustic Sensors: These sensors are deployed both on
the bridge and on all paths leading to it, to detect sound
generated by moving troops and vehicles. Their input-
output characteristics are directly affected by noise and
indirectly affected by temperature.
4. Radioactive Sensors: These sensors are used to detect
radioactive emission from weapons systems on-board
the vehicles. They are highly accurate over short spans
of the bridge. Their input-output characteristics are di-
rectly affected by available energy. Thus, these sensors
must be activated selectively to conserve energy.
5. Line of sight optical sensors: These sensors employ a
transmitter and receiver on either side of the bridge.
Detection is accomplished by breakage in the line of
sight beam from the transmitter to the receiver. These
sensors are employed on all paths leading to the bridge
for early detection of enemy movement. The number
of times the beam is broken and the amount of time
it stays broken can be used to give information about
counts and lengths of the vehicles. These sensors are
directly affected by reduction in visibility (e.g., fog,
sandstorms) and humidity.
4. Data Calibration Algorithm
In this section, we discuss the algorithm executed by ca-
pable sensor nodes and possibly BS/RU nodes to perform
data calibration on each type of sensor whenever significant
changes in environmental conditions occur.
There are three inputs to the algorithm as shown in the
flowchart in Figure 4: current and nominal values of each of
the core environmental variables and the current data output
of the sensor type. Based on these inputs, the data calibra-
tion algorithm computes the expected output of each sensor
type at existent values of core environmental variables. For
example, the core environmental variable affecting acoustic
sensors described in Section 3 are temperature and exter-
nal noise. Further, the acoustic sensor’s output is a voltage
corresponding to the input sound intensity. The final out-
put of the algorithm for the acoustic sensor is a calibrated
value of the sensor’s output.

Current values of CEV*
Nominal values of CEV
Tolerance to variation
Data output from sensor
|Nominal - Current| >
Tolerance?
DoneNo
Yes
Execute function OR
Access lookup table
associated with CEV
effect model
Sensor Input-Output Characteristics
Effect Model of CEV
Compute
calibrated sensor
output
Sensor Accuracy
-
+/-
Error induced by CEV
*CEV = Core
Environmental Variable
Figure 2. Data Calibration Algorithm
We now explain the flowchart in detail. Initially, the al-
gorithm determines whether data calibration must be per-
formed. This is accomplished by determining if the current
value of each core environmental variable exceeds its nom-
inal value by a pre-determined threshold value. If so, data
calibration is required.
The algorithm uses either the lookup table or function
(Section 2) describing the effect of each core environmen-
tal variable on the input-output characteristics of each sen-
sor to determine its expected output value. The lookup ta-
ble maps different values of a core environmental variable
to corresponding output values of the sensor. On the other
hand, a function relates sensor input, sensor output and the
core environmental variable and describes how the output
value changes according to the environmental variable.
For example, assume that the average temperature co-
efficient of sensitivity (the function in this case) of above
acoustic sensor is 1%/°C with respect to the nominal value.
Therefore, if temperature increases by 10°C, then sensitiv-
ity (which is the ratio of output to input of a sensor) changes
by 10%, which implies that for the same input value, the
output of the acoustic sensor either reduces or increases by
10%. Thus, having obtained the error induced by the in-
crease in temperature, the algorithm corrects the measured
output accordingly.
Based on the measured and calibrated output values, the
algorithm computes the accuracy of the sensor and stores
this value in the
field of the

quadruple.
The current accuracy value of a sensor type determines its
most appropriate operating state (
) and corresponding en-
See next figure
MemoryModule:class
ProcessorModule:class
hasCPU:object
hasMemoryComponent:object
hasType:string
hasSize:object
ValueAndUnits:class
hasValue:float
hasUnits:string
ValueAndUnits:class
hasValue:float
hasUnits:string
ValueAndUnits:class
hasValue:float
hasUnits:string
ValueAndUnits:class
hasValue:float
hasUnits:string
ValueAndUnits:class
hasValue:float
hasUnits:string
ValueAndUnits:class
hasValue:float
hasUnits:string
ValueAndUnits:class
hasValue:float
hasUnits:string
ValueAndUnits:class
hasValue:float
hasUnits:string
ValueAndUnits:class
hasValue:float
hasUnits:string
ValueAndUnits:class
hasValue:float
hasUnits:string
ValueAndUnits:class
hasValue:float
hasUnits:string
ValueAndUnits:class
hasValue:float
hasUnits:string
ValueAndUnits:class
hasValue:float
hasUnits:string
ValueAndUnits:class
hasValue:float
hasUnits:string
ValueAndUnits:class
hasValue:float
hasUnits:string
ValueAndUnits:class
hasValue:float
hasUnits:string
PowerSupplyModule:class
hasInitialCapacity:object
hasRemainingCapacity:object
hasMinInputVoltage:object
hasMaxInputVoltage:object
hasMinOutputVoltage:object
hasMaxInputVoltage:object
hasPowerSupplyType:string
CPUModule:class
hasName:string
hasSpeed:object
hasWordLength:object
drawsCurrent:object
RadioModule:class
hasCenterFrequency:object
hasNumberOfChannels:int
hasMinRFPower:object
hasDataRate:object
hasMaxRFPower:object
hasOutdoorRange:object
hasReceiveSensitivity:object
GeoLocation:class
Altitude:string
Latitude:string
Longitude:string
SensorNode:class
hasID: string
hasProcessorModule:object
hasPowerSupplyModule:object
hasRadioModule:object
hasSensorModule:object
hasGeographicLocation:object
hasState: string
Figure 3. Sensor Node Ontology Processor,
Power Supply and Radio Modules
ergy consumption (
) in that state. Individual sensor nodes
cannot make this determination because they do not possess
either the user-defined energy accuracy bounds or a global
view of the existing state of the WSN. Only BS/RU nodes,
which possess both the capability and the information to ob-
tain the global state, can determine the final operating state
of each sensor type in the WSN, under existing environmen-
tal conditions (Section 5.1). Thus, each capable node places
the current operating state and corresponding energy con-
sumption value in the
and
fields of the quadruple, re-
spectively. The capable sensor node transmits the quadru-
ple in a message to the BS/RU node functioning as the sen-
sor node’s cluster-head.
Sensor nodes that are incapable of executing this al-
gorithm do not transmit the quadruple to their respective
cluster-heads. The nodes merely transmit uncalibrated data.
Cluster-heads calibrate data received from these nodes and
use prior state information associated with sensor types on-
board the nodes to determine the final operating state of
each sensor type.

EnvironmentVariableModel:class
hasName:string
hasPresentValue:object
hasAffectDescription:string
EnvironmentVariableModel:class
hasName:string
hasPresentValue:object
hasAffectDescription:string
hasMinInputSignal:object
hasMaxInputSignal:object
SensorInputType:class
hasName:string
SensorModule:class
hasSensorType:object
ValueAndUnits:class
hasValue:float
hasUnits:string
ValueAndUnits:class
hasValue:float
hasUnits:string
ValueAndUnits:class
hasValue:float
hasUnits:string
hasMinOutputSignal:object
hasMaxOutputSignal:object
SensorOutputType:class
hasName:string
ValueAndUnits:class
hasValue:float
hasUnits:string
ValueAndUnits:class
hasValue:float
hasUnits:string
ValueAndUnits:class
hasValue:float
hasUnits:string
ValueAndUnits:class
hasValue:float
hasUnits:string
ValueAndUnits;class
hasValue:float
hasUnits:string
ParameterAndTypicalValue:class
hasTypicalValue:object
hasName:string
ValueAndUnits:class
hasValue:float
hasUnits:string
ValueAndUnits:class
hasValue:float
hasUnits:string
ValueAndUnits:class
hasValue:float
hasUnits:string
ValueAndUnits:class
hasValue:float
hasUnits:string
hasAffectDescription:string
DynamicCharacteristic:class
hasPresentValue:object
isAffectedBy:object
hasName:string
InputOutputModel:class
hasIOModelDescription:string
hasOutputTolerance:float
hasParameterWithTypicalValue:
object
hasDynamicCharacteristic:object
isAffectedBy:object
hasParameterTolerance:float
SensorType:class
hasSensorInputType:object
hasSensorOutputType:object
hasSensorMechanismType:
string
hasMinOperatingTemperature:
object
hasMaxOperatingTemperature:
object
hasInputResolution:object
hasOutputResolution:object
hasInputOutputModel:object
hasAccuracy:float
hasOperatingMode:string
hasNormalEnergyConsumption:
float
Figure 4. Sensor Node Ontology Sensor
Module
5. Sensor Node Ontology
The sensor node ontology we have designed attempts to
capture the most important features of a sensor node that
describe its functionality and its current state. The informa-
tion captured by our ontology is used by BS/RU to deter-
mine the current and future state of the WSN. The ontology
describes the main components of a sensor node: processor,
power supply, radio and sensor modules. A common thread
that runs through these modules is energy, both in terms of
capacity and consumption. Figures 3 and 4 show our sen-
sor node ontology in detail.
The processor module consists of the CPU and memory
components. The most interesting and important facets of a
CPU that influence the functioning of a sensor node are the
CPU’s available operating modes (i.e., active, low-power,
idle) and the amount of power it consumes in each mode.
Based on this information, the WSN can decide the most
appropriate CPU operating mode for each sensor node or a
class of sensor nodes, under existing environmental condi-
tions. The memory component reflects both static and dy-
namic memory on the sensor node.
The power supply module is an essential component of
a sensor node. This module determines how long the sen-
sor node can function usefully. Given that a typical sensor
node is self-contained and is typically unattended after de-
ployment, the power supply module is the final limiting fac-
tor on the sensor node’s capabilities. This is reflected in two
of the most important characteristics of the power supply
module initial energy capacity and remaining energy ca-
pacity. Remaining energy capacity dictates the ability of the
sensor node to continue functioning in a certain mode. In-
formation regarding energy capacity of a node plays a very
important role in WSN adaptivity.
The radio transceiver module helps the sensor node com-
municate with other nodes in the network. It consists of a
transmitter and receiver. Identically to the CPU component,
the most important properties of the radio transceiver that
affect the sensor node’s functioning are its available modes
of operation (i.e., transmit, receive, low-power, sleep) and
power consumption in each mode. Each operating mode of
the transceiver reflects environmental conditions affecting
the node.
The sensor module consists of descriptions of one or
more sensor types. For each sensor type, primary informa-
tion captured in the ontology includes the

quadruple, i.e, sensor type, operating mode, accuracy and
energy consumption along with a description of the effect
of each core environmental variable on the sensor type. Ad-
ditionally, it also captures static information about the sen-
sor type, such as its input and output signal types, input and
output resolutions, and minimum and maximum operating
temperatures.
Additional useful information present in the description
of a sensor node includes its ID and geographical location.
The former is pre-deployed on the node while the latter can
be obtained post-deployment if the node possesses a Global
Positioning System (GPS) unit on-board.
5.1. WSN State Determination Algorithm
This algorithm enables a WSN to dynamically determine
its most appropriate state of operation under existing envi-
ronmental conditions. The operating state of a WSN reflects
the corresponding states of sensor nodes and each type of
sensor in the WSN. Figure 5 shows a flowchart of this al-
gorithm, which is executed by each BS/RU node currently
functioning as a cluster-head (Section 2).
The input to this algorithm is the
E,
S

tuple trans-
mitted by each sensor node to its cluster-head. We recall that
E is the remaining energy on the sensor node and
S
is the
set of

quadruples associated with its sensor
types.
Initially, the algorithm extracts and classifies each

quadruple according to a pre-determined classi-
fication scheme (e.g., sensor type, sensor energy consump-
tion, sensor operating mode, node location). In Section 6,
we discuss two different classification schemes applied to
the WSN described in the application scenario (Section 3)

Citations
More filters
26 Oct 2009
TL;DR: The state of the art for the semantic specification of sensors, one of the fundamental technologies in the semantic sensor network vision, is reviewed.
Abstract: Semantic sensor networks use declarative descriptions of sensors promote reuse and integration, and to help solve the difficulties of installing, querying and maintaining complex, heterogeneous sensor networks. This paper reviews the state of the art for the semantic specification of sensors, one of the fundamental technologies in the semantic sensor network vision. Twelve sensor ontologies are reviewed and analysed for the range and expressive power of their concepts. The reasoning and search technology developed in conjunction with these ontologies is also reviewed, as is technology for annotating OGC standards with links to ontologies. Sensor concepts that cannot be expressed accurately by current sensor ontologies are also discussed.

228 citations


Cites background from "Ontology-driven adaptive sensor net..."

  • ...Avancha, Patel and Joshi [10] describe an ontology for adaptive sensor networks, where nodes react to available power and environmental factors, calibrating for accuracy and determining suitable operating states....

    [...]

Proceedings ArticleDOI
27 Jun 2007
TL;DR: The performance analysis demonstrated the ability of the ontology-based search to improve both the precision and recall rates and enhance the interoperability between different sensor networks domains through the use of the universal SUMO ontology.
Abstract: In this paper, we present our work towards the development and evaluation of an ontology for searching distributed and heterogeneous sensor networks data. In particular, we propose a two layer prototype ontology that utilizes the IEEE Suggested Upper Merged Ontology (SUMO) as a root definition of general concepts and associations and two sub- ontologies: the sensor data sub-ontology and the sensor hierarchy sub-ontology. The proposed ontology was implemented using Protege 2000 and eventually evaluated using the RDQL language (RDF Data Query Language). The performance analysis demonstrated the ability of the ontology-based search to improve both the precision and recall rates and enhance the interoperability between different sensor networks domains through the use of the universal SUMO ontology.

117 citations


Cites background or methods from "Ontology-driven adaptive sensor net..."

  • ...The ability to search, task, control, and fuse data collected from heterogeneous sensors can significantly facilitate the discovery of added value knowledge that is unreachable using classical information retrieval techniques [1-2]....

    [...]

  • ...The proposed ontology was implemented using Protégé 2000 and eventually evaluated using the RDQL language (RDF Data Query Language)....

    [...]

Proceedings ArticleDOI
14 May 2008
TL;DR: A service-oriented sensor ontology which enables service- oriented services in future ubiquitous computing is proposed and the results of service query which used the SPARQL query language are indicated.
Abstract: Wireless sensor networks (WSNs) provide various environment data in the real-world, and also WSNss middleware is able to offer field data in real-time by user queries. For materialization of the future ubiquitous computing which enables networking with things at anytime, anywhere and any-devices, WSNs occupy the important position with RFID technologies, and it has evolved and advanced currently. This paper proposes a service-oriented sensor ontology which enables service-oriented services in future ubiquitous computing. Taking reuse of ontology into consideration, ServiceProperty, LocationProperty and PhysicalProperty classes were derived from Geography Markup Language (GML), Sensor Web Enablement(SWE), SensorML and Suggested Upper Merged Ontology(SUMO) and OntoSensor ontology, and its properties and constraints were also defined newly as service-oriented service. We presented the validation and consistency check of the proposed ontology using Protege 3.3.1 and RACER 1.9.0, respectively, and indicated the results of service query which used the SPARQL query language.

62 citations


Cites background from "Ontology-driven adaptive sensor net..."

  • ...This was demonstrated in some recent work on the use of process ontologies [1, 2, 3, 4, 5] that showed an increase in the precision of service discovery queries when semantic representations were used over syntactic representations....

    [...]

  • ...Reference [3] presents a detailed description of the development stages of ontologies....

    [...]

  • ...The work in [3] presents an attempt to capture the most important features of a sensor node that describes its functionality and its current state....

    [...]

Proceedings ArticleDOI
12 Jul 2006
TL;DR: It is argued that the key to enabling scalable and precise sensor information search is to define an ontology that associates sensor information taxonomy for searching and interpreting raw data streams.
Abstract: Sensor networks have seen an exponential growth in the last few years. They involve deploying a large number of small sensing nodes for capturing environmental data. Searching such networks is limited by two major constraints: scalability and precision. We argue that the key to enabling scalable and precise sensor information search is to define an ontology that associates sensor information taxonomy for searching and interpreting raw data streams. We present the motivation and description of the development of the proposed ontology, partial evaluation of the early prototype ontology, a discussion of design and implementation issues, and directions for future research works.

47 citations


Cites background from "Ontology-driven adaptive sensor net..."

  • ...The work in [4] presents an attempt to capture the most important features of a sensor node that describes its functionality and its current state....

    [...]

Journal ArticleDOI
08 Feb 2008-Sensors
TL;DR: In this article, the authors present an approach to automatically transform raw sensor data into a representation that matches a predefined model of the problem context, which can be used to reason with, and draw conclusions about, spatial data.
Abstract: In the context of hazard monitoring, using sensor web technology to monitor and detect hazardous conditions in near-real-time can result in large amounts of spatial data that can be used to drive analysis at an instrumented site. These data can be used for decision making and problem solving, however as with any analysis problem the success of analyzing hazard potential is governed by many factors such as: the quality of the sensor data used as input; the meaning that can be derived from those data; the reliability of the model used to describe the problem; the strength of the analysis methods; and the ability to effectively communicate the end results of the analysis. For decision makers to make use of sensor web data these issues must be dealt with to some degree. The work described in this paper addresses all of these areas by showing how raw sensor data can be automatically transformed into a representation which matches a predefined model of the problem context. This model can be understood by analysis software that leverages rule-based logic and inference techniques to reason with, and draw conclusions about, spatial data. These tools are integrated with a well known Geographic Information System (GIS) and existing geospatial and sensor web infrastructure standards, providing expert users with the tools needed to thoroughly explore a problem site and investigate hazards in any domain.

44 citations

References
More filters
Journal ArticleDOI
TL;DR: The concept of sensor networks which has been made viable by the convergence of micro-electro-mechanical systems technology, wireless communications and digital electronics is described.

17,936 citations


"Ontology-driven adaptive sensor net..." refers background in this paper

  • ...Thus, issues of routing, transport, power management and data aggregation are dealt with in other work [1, 2, 3] and are beyond the scope of this paper....

    [...]

  • ...In our model, the WSN consists of a limited number of static or mobile BSs and RUs, but is otherwise unattended [1]....

    [...]

Proceedings ArticleDOI
01 Aug 1999
TL;DR: It is found that the SPIN protocols can deliver 60% more data for a given amount of energy than conventional approaches, and that, in terms of dissemination rate and energy usage, the SPlN protocols perform close to the theoretical optimum.
Abstract: In this paper, we present a family of adaptive protocols, called SPIN (Sensor Protocols for Information via Negotiation), that efficiently disseminates information among sensors in an energy-constrained wireless sensor network. Nodes running a SPIN communication protocol name their data using high-level data descriptors, called meta-data. They use meta-data negotiations to eliminate the transmission of redundant data throughout the network. In addition, SPIN nodes can base their communication decisions both upon application-specific knowledge of the data and upon knowledge of the resources that are available to them. This allows the sensors to efficiently distribute data given a limited energy supply. We simulate and analyze the performance of two specific SPIN protocols, comparing them to other possible approaches and a theoretically optimal protocol. We find that the SPIN protocols can deliver 60% more data for a given amount of energy than conventional approaches. We also find that, in terms of dissemination rate and energy usage, the SPlN protocols perform close to the theoretical optimum.

2,525 citations

Journal Article
TL;DR: In this article, a two-phase post-deployment calibration technique for large-scale, dense sensor de-ployment is presented, where the first phase derives relative calibration relationships between pairs of co-located sensors, while the second phase maximizes the consistency of the pair-wise calibration func- tions among groups of sensor nodes.
Abstract: Numerous factors contribute to errors in sensor measure- ments. In order to be useful, any sensor device must be calibrated to adjust its accuracy against the expected measurement scale. In large- scale sensor networks, calibration will be an exceptionally dicult task since sensor nodes are often not easily accessible and manual device-by- device calibration is intractable. In this paper, we present a two-phase post-deployment calibration technique for large-scale, dense sensor de- ployments. In its �rst phase, the algorithm derives relative calibration relationships between pairs of co-located sensors, while in the second phase, it maximizes the consistency of the pair-wise calibration func- tions among groups of sensor nodes. The key idea in the �rst phase is to use temporal correlation of signals received at neighboring sensors when the signals are highly correlated (i.e. sensors are observing the same phenomenon) to derive the function relating their bias in amplitude. We formulate the second phase as an optimization problem and present an algorithm suitable for localized implementation. We evaluate the perfor- mance of the �rst phase of the algorithm using empirical and simulated data.

264 citations

Proceedings ArticleDOI
28 Sep 2002
TL;DR: By exploring the criteria for effective infrastructure configurations, this paper opens the door for network optimizations that control the effective topology to better achieve the application requirements.
Abstract: In a sensor network, the infrastructure (in terms of the sensor capabilities, number of sensors, and deployment strategy) plays a significant role in determining the performance of the network. In this paper, we study the effect of infrastructure decisions on the performance of a sensor network. We study the effect of the infrastructure for two types of network delivery models (phenomenon driven and continuous) and different network protocols (DSR, DSDV and AODV). We show the performance both in terms of network efficiency as well as meeting the application accuracy and latency demands. By exploring the criteria for effective infrastructure configurations, we open the door for network optimizations that control the effective topology to better achieve the application requirements.

261 citations

Book ChapterDOI
22 Apr 2003
TL;DR: In this paper, a two-phase post-deployment calibration technique for large-scale, dense sensor deployments is presented, in which the first phase is to use temporal correlation of signals received at neighboring sensors when the signals are highly correlated (i.e. sensors are observing the same phenomenon) to derive the function relating their bias in amplitude.
Abstract: Numerous factors contribute to errors in sensor measurements. In order to be useful, any sensor device must be calibrated to adjust its accuracy against the expected measurement scale. In large-scale sensor networks, calibration will be an exceptionally difficult task since sensor nodes are often not easily accessible and manual device-by-device calibration is intractable. In this paper, we present a two-phase post-deployment calibration technique for large-scale, dense sensor deployments. In its first phase, the algorithm derives relative calibration relationships between pairs of co-located sensors, while in the second phase, it maximizes the consistency of the pair-wise calibration functions among groups of sensor nodes. The key idea in the first phase is to use temporal correlation of signals received at neighboring sensors when the signals are highly correlated (i.e. sensors are observing the same phenomenon) to derive the function relating their bias in amplitude. We formulate the second phase as an optimization problem and present an algorithm suitable for localized implementation. We evaluate the performance of the first phase of the algorithm using empirical and simulated data.

238 citations

Frequently Asked Questions (2)
Q1. What are the contributions mentioned in the paper "Ontology-driven adaptive sensor networks" ?

The authors present a novel, two-phase solution to the wireless sensor network adaptivity problem. The authors describe an example application scenario to show how their two-phase solution can be employed to enable a realworld wireless sensor network to adapt itself to variations in environmental conditions. 

In future work, the authors will describe results of simulations of the proposed solution, in conjunction with the other components of their framework for secure, adaptive sensor networks, on a large scale. Work on incorporating mechanisms to enable a deployed WSN to adapt to changing security conditions in addition to environmental conditions is ongoing and results will be reported in future work.