You are on page 1of 9

7th IEEE International Workshop on Performance and Management of Wireless and Mobile Networks P2MNET 2011, Bonn, Germany

Wi-Design: A Modelling and Optimization Tool for


Wireless Embedded Systems in Buildings
Alan McGibney, Antony Guinard, Dirk Pesch
Nimbus Centre for Embedded Systems Research,
Cork Institute of Technology
Cork, Ireland
alan.mcgibney@cit.ie, antony.guinard@cit.ie, dirk.pesch@cit.ie

Abstract—As wireless embedded systems become more and more infrastructure for in-building wireless applications. We show in
common and used across many application domains there is a this paper how Wi-Design simplifies the design process of
need for modeling and design tools to support the deployment wireless sensor networks and creates robust deployments
process. Although a significant amount of research has been compared to the traditional ad-hoc approach. The remainder of
carried out in the area of protocol design, middleware and the paper is structured as follows, section II will describe
energy, packaging and embedded systems design, there remains a current approaches and associated problems for indoor
lack of support tools for designers and system integrators when deployments and will also provide an overview of currently
deploying complex indoor wireless infrastructures (Wi) to available planning tools. Section III will describe the Wi-
support site specific applications. In this paper we present a
Design toolset and highlight the key elements for wireless
modeling tool known as Wi-Design that was developed to provide
deployment support for engineers and system integrators when
system design. As automating the design process is essential
planning a wireless sensor infrastructure with particular focus on for non expert users the optimization kernel will be described
in building wireless applications. We show how the tool can in more detail in Section IV. Section V will present an example
simplify the deployment process and provide enhanced design and Section VI will conclude the paper.
confidence in wireless deployments.
II. RELATED WORK
Keywords-Design tools; Wireless Sensor Network Design;
Propagation Modeling; Optimization tools for deployment A. Deployment Approaches
Deploying WSN in indoor environments is a complex task.
I. INTRODUCTION A successful design can be considered as determining the best
Wireless sensor networks (WSN) are becoming more and combination of hardware, network topology, operation modes,
more common as a mechanism for capturing environmental configuration, quantity and position of nodes that meet the
data and interacting with the physical environment both for application objectives. The most significant consideration that
commercial and military applications. Recently industry has impacts the quantity, position and topology of the wireless
been focusing efforts on harnessing this technology for infrastructure is the layout of the building itself. The building
monitoring and control applications within buildings. These structure and material types have a considerable influence on
applications cover aspects such as lighting, safety and security, Radio Frequency (RF) propagation from devices through the
environment monitoring, energy and general building environment. To include this complexity a designer can
management and operations. However the planning, typically take one of the following options, an ad hoc design, a
implementation and management of wireless sensor networks site survey or a propagation modeling tool.
is a complex task requiring a broad spectrum of disciplines A common technique employed by system integrators when
such as wireless networking, information management and deploying an indoor WSN is a “try and see” approach. The
embedded systems engineering to be successful [1]. A network is built by placing wireless nodes into what seems to
significant amount of research has been carried out in the area be good positions. Tests are carried out to evaluate signal
of protocol design, middleware and energy, packaging and coverage and nodes are moved or added to improve network
embedded systems design but there is still a significant lack of performance. A few iterations are often necessary to get a
tools that make it easy for designers and system integrators to workable design. This approach relies heavily on the designer’s
develop and deploy reliable yet complex indoor wireless experience, is time consuming and provides an optimal solution
infrastructures (Wi) to support site specific applications. only by chance. Moreover this approach rapidly becomes
This paper presents a modeling tool we call Wi-Design that unmanageable as the network size grows. Some hardware
was developed to provide deployment support for engineers venders such as EnOcean provide range planning guidelines
and system integrators when planning a wireless sensor that can be used when deploying their products [2]. These
guidelines can be used to estimate the number and position of
The authors wish to acknowledge the support of Science Foundation devices but they are based on an idealized (circular) coverage
Ireland under the 06-SRC-I1091 ITOBO project in funding work reported in
this paper.
range which is not realistic in typical buildings with walls and

978-1-61284-928-7/11/$26.00 ©2011 IEEE 640


other obstacles and therefore the actual range can vary devices within building environments. An approach presented
significantly between sites. Often the installers of wireless in [15] includes obstacles from within the deployment
systems in buildings, who are building system integrators or environment however fails to consider the complexity of radio
often even only electricians, have little or no experience with transmissions as it only uses a simple geometrical approach to
wireless technologies and hence this approach can lead to the estimate connectivity and coverage area. A number of
following problems: proposals suggest methodologies to automate the selection of
• Under estimation of the number of devices (sensors, node position for indoor environments [15]-[17]. Although
repeaters or gateways) required to support the these proposals provide support on some aspects of the
application. This results in the supplier/distributer difficult task of WSN design, there still remain a number of
deploying a system with poor performance and as a drawbacks. As is often the case, many approaches use an
result they may not only lose the confidence of their idealized propagation model [15] or at most an empirical
client but also incur additional cost to repair the poorly propagation model such as the Multi-Wall model [16][17].
performing deployment. Despite being computationally efficient, the Multi Wall model
only considers the direct line between transmitter and receiver
• Over estimation of infrastructure requirements. To
ensure good performance the supplier may over and often needs to be tuned for a specific environment in order
estimate the number of devices required to install the to get an accurate prediction of signal propagation in the
system. This can result in an uncompetitive price for building. This leads to an increase in design time. A
the customer in relation to alternative, perhaps wired methodology and software framework proposed in [16] aims
systems available. to address this issue. The proposed solution remains inflexible
as a fixed position is assumed for the sink node. In a large
• To ensure reliability in large deployments labor building environment, many locations are appropriate for a
intensive site surveys, adjustments and re-designs may sink, in fact many applications support more than one sink.
be required. This is expensive and increases lead time Determining the best location for sink nodes is an optimization
to deployment and can make the solution problem that needs to be addressed for WSN design [18]. In
uncompetitive in comparison to alternative solutions. addition, the methodology proposed in [16] only considers the
A number of WSN deployments within industrial plant design of WSN using single-hop communication. A
environments have been carried out [3]-[5]. In these research framework for rapid design and evaluation of WSN has been
works, accurate wireless communication characteristics developed in [19]. The framework incorporates a detailed
prediction is regarded as being a key factor to a successful simulation environment which estimates the performance of
WSN design. In each scenario, this is achieved by proceeding communication protocols and can be used to generate the
to a site survey, which is a suitable solution considering the implementation code for the physical deployment. The tools
limited size of the WSN deployment (varying from 5 to 45 also include a mechanism to capture live measurement data
nodes). While these experiments took place in extremely harsh from nodes to tune simulations and adjust the design. The
environments for WSN, the results presented in [6] show that model used to predict radio propagation characteristics does
the prediction of wireless signal characteristics is also required not consider obstacles within the environment. Instead, the
within more traditional building environments. A site survey
signal attenuation is retrieved from analyzing live data from a
consists of undertaking a measurement campaign within the
deployed WSN. The approach in [19] provides a
target environment. A significant amount of work is required to
carry out several measurement runs, collect and organize the comprehensive simulation environment to test applications. It
captured data and to perform a final evaluation. Hardware does not focus on optimizing the position of nodes in indoor
providers such as Spinwave Systems [7] and Monnit [8] environments and although the authors agree the use of a
provide a solution to enable a designer to assess the feedback loop from live data is an important aspect of the
connectivity with deployed sensors at different locations. A design lifecycle to validate and verify a design post
drawback of proceeding to a site survey using these devices is deployment, making it an integral part of the initial
that several point-to-point measurements are needed at deployment plan as a link quality estimation is similar to doing
different locations, which can become unmanageable as the an extensive site survey.
size of the network increases. It is our opinion that incorporation of a design
methodology where radio propagation characteristics and
B. Existing Planning Algorithms and Tools other requirements such as delays, interference and network
Previous research attempting to address the node lifetime can be modeled with software tools offers a number of
placement problem has largely focused on developing advantages. Firstly it removes or reduces the need for costly
techniques to ensure coverage and connectivity but has largely and time consuming site survey throughout the environment.
focused on obstacle-free environments [9]-[14]. The studied Planning tools ensure that a user, even with little experience
problem was simplified by assuming that both sensing and considers the impact of the deployment environment,
communication ranges are circular. The idealized application requirements and configuration on network
communication characteristics of the wireless nodes, performance. The design tool can also be used to evaluate
combined with the fact that obstacles are not considered, network expansion or the viability of new wireless
render these techniques inappropriate when positioning applications. Additionally the tools can be used to estimate

641
infrastructure cost for wireless applications even prior to the A large number of hardware platforms are available to choose
construction of a building offering a mechanism to support from for indoor wireless sensor network [20]. To assist the
feasibility studies for new applications at the building design designer with the wide choice of available hardware, a library
stage. of components that shows differences between hardware
platforms has been integrated into the toolset. This provides the
III. THE WI-DESIGN TOOLSET user the ability to observe changes in a WSN design resulting
The Wi-Design toolset aims to improve on current from a change of hardware platform. The device characteristics
deployment practices for wireless networks with particular currently considered with the hardware library includes device
focus on embedded applications within buildings. The design type (sensor, actuator), manufacturer, cost, sensitivity
procedure is supported with a CAD tool incorporating an threshold, transmission power, operating frequency, maximum
automatic design and optimization kernel to suggest the ideal bit rate, available channels and power consumption
number, type, topology and more importantly positions of characteristics. Application specific requirements need to be
network nodes based on user and application specific considered for a WSN deployment. The approach proposed in
requirements. The development of such a kernel is a non trivial Wi-Design considers demand zones where devices should be
task as it must consider a multitude of design requirements placed, the device classification (generator, repeater, sink),
including signal coverage, connectivity, quality of service sensing interval and transmission delays. These requirements
requirements and network infrastructure. The current toolset are used as design constraints within the optimization kernel
includes propagation modeling, channel characteristics and for the automated WSN design, the goal of the optimization is
optimization algorithms to aid a designer when planning a to satisfy the user defined requirements.
wireless deployment. 2) Infrastructure Design
The main components of the toolset are associated with the
A. Wi-Design Components modeling and predictive tools for wireless communication with
The Wi-Design toolset is composed of a number of a focus on building environments. The following describe the
elements and tools that can be used for wireless infrastructure sub-elements of the infrastructure design.
design. These tool elements can be classified under three Propagation Modeling
categories of the design lifecycle, Requirements, Infrastructure
Design and Verification. Figure 1 shows the components of the As can be seen from the literature, the RF propagation
Wi-Design architecture, the following section will describe the model is a fundamental part of deployment planning, therefore
most important tool components but will focus on the the choice of propagation model is important. Based on the
automated design algorithm in the optimization kernel. reviewed literature and close investigation into existing indoor
propagation models it is clear that there is a trade-off between
computational complexity and model accuracy. Propagation
models used include simple path loss models which will greatly
reduce the quality of a solution particularly in a complex indoor
environment. For most indoor scenarios empirical models such
as the Empirical Motley-Keenan Model [21] are typically
chosen as they are easy to implement and can be evaluated
quickly. To improve the quality of the prediction more
complex ray tracing models can be used, however the higher
accuracy of ray tracing models comes at a price; the complexity
of the problem is increased and computation time can become
unacceptable. Wi-Design tools include the One Slope Model,
Multi-wall Model [22] and a Ray launching approach known as
Figure 1. The Wi-Design Toolset Architecture the Motif Model [23]. The Motif model has been shown to be is
the most accurate of the three models while remaining
1) Requirements computationally acceptable for large scenarios [24] . The Motif
Before any design can be undertaken system requirements Model is our model of choice.
must be captured. An integral part of these requirements is a Topology Modeling
description of the environment where the network should be
deployed in order to accurately predict signal propagation, and In building environments, sensors typically report their
as a result, link quality between nodes. The design tool offers readings to a centralized controller which is accessible through
drawing capabilities that enables a user to define the indoor one or many sink nodes. In such a one-to-many communication
environment with basic features such as walls, doors and paradigm each sensor has several alternative paths that can be
windows. Material types that can influence the propagation can used to reach a sink node. A significant amount of research has
also be defined. As drawing the environment can be very time been done on routing protocols to identify the optimum paths.
consuming Wi-Design also provides import tools to speed up In Wi-Design a fixed topology using primary routes only is
the process. To be in line with construction industry tools floor assumed, it is planned to extend this further by considering
plan described in AutoCAD format or Industry Foundation disjoint paths as an optimization criteria in the future. The
Classes (IFC) can automatically be imported to the design tool. current implementation of the Wi-Design tool models the
network topology that reflects a balanced network and primary

642
routes that should be used under normal operation of the WSN. But scalability becomes an issue when the problem becomes
The mechanism used to build the network topology is large. Based on the previous work we did on Evolution
explained in Section IV.C.1). Strategies [34] in the context of IEEE802.11 based wireless
Interference Modeling network planning, we found that taking a global view of the
optimization problem produces poor designs when applied to
Interference can play a major role in the quality of
large design spaces. Evolutionary algorithms represent a
communications in a building. Wi-Design provides a
mechanism to enable a designer to define an existing wireless solution globally and as a result are unable to consider all,
infrastructure that can interfere with wireless nodes. For often conflicting metrics as part of their solution. The
example by defining access point locations of an existing WiFi convergence becomes unpredictable and the quality of the
network and allocated channels, the optimization algorithm can solution deteriorates [34].
consider this and will avoid channel interference and maximize Much attention has been given to game theoretic approaches
signal to noise ratio for the to be deployed infrastructure. This to problems in wireless sensor networks [35] (power control
can also enable the design to optimize channel selection pre- games, self organizing networks). Game theoretical algorithms
deployment. are often implemented in agents playing a game to reach a
Visualization solution to a problem. Our proposed approach reduces the
wireless network design problem to a system of agents that
The visualization tools of Wi-Design are important to
only have local knowledge of the environment playing a game
provide feedback to the user on deployment plans. This enables
the designer to visually assess a plan and make adjustments to that achieves a (near) optimum solution regardless of problem
design constraints or requirements for example define areas size. In our approach the optimization is not driven by one
that should not be covered or with restricted access. Wi-Design global function but rather a global optimum will emerge as a
provides a user friendly interface to input and capture result of simple local decisions and interactions between
requirements and provide 2D and 3D representations of the rational agents within the design environment.
building environment and associated coverage/connectivity 1) Wi- Design Agent
maps. In our wireless infrastructure design approach an agent is
3) Verification representative of a wireless device trying to find its best
The Wi-Design toolset includes site survey tools that can be position within the environment by maximizing its personal
linked directly back to the propagation models and hence verify utility function. An agent has the ability to perceive its
initial predictions. The tools also allow for the visualization of environment, persisting for a long period of time, adapting to
network statistics when available enabling the designer to change and being capable of taking on the goals of other
monitor a deployed network for route changes and overall agents [36]. We consider a simple reflex, utility-based agent
network performance. The tools provide graphical views of (Figure 2). The main components of an agent are environment
system performance that can be used to analyze the network percept, condition-action rules and utility function.
performance.

IV. OPTIMIZATION KERNEL


The real novelty of Wi-Design lies in the optimization
kernel. Existing tools provide a user with the ability to
manually select device positions and evaluate different
configurations of their proposed network. But it is the
integration of an optimization and design algorithm that
enables Wi-Design to automatically suggest an optimal solution
to the user’s site specific requirements with little configuration.
This allows even users with little or no experience of indoor
wireless deployments to create an optimal deployment plan. In
the following we describe our optimization approach and the
problem specific optimization criteria we consider for the
design algorithm. Figure 2. Agent Design for WSN Design

A. Optimization Approach Figure 2 shows the architecture of the agents. The agent
The optimization techniques that have been applied to encapsulates all constraints that have been defined during the
requirements phase including candidate node positions, target
wireless network design range from simple search methods
zones, sensor positions and connectivity requirements. As all
[25][26], Integer Linear Programming [27][28], Quadratic
this information is pre-processed and available to the agent it
Programming [29][30] and Genetic Algorithms [30]-[32]. The
accelerates the optimization process. The optimization is
most popular optimization approaches applied in the area of modeled as an iterative game played by the agents competing
wireless networks are direct search methods, mainly due to the to maximize their utility, for this example the agent is
fact that they are easy to implement, flexible and robust [33].

643
comparable to a sink node attempting to maximize the number Where,
of associated sensor nodes. The agent incorporates general
RU……………………Resource Utilisation
rules pertaining to the problem being solved, such as agents
may only connect to a maximum number of end devices with a OS……………………Over Subscription
specified sensitivity threshold or a maximum number of hops GC...…………………Global Coverage
have to be adhered too. The agent operates in a fully
observable environment where it has sufficient facts about it; B……………………..Balance Factor
therefore a logical approach enables the agent to derive a plan wi……………………..Weighting Factor
that is guaranteed to work. Once the agent knows what the
world looks like (agent percept) it can then use condition- The elements of the utility function include resource
action rules to establish a connection between agent percept utilization (RU) which is calculated as the ratio of successfully
and which action should be considered. connected nodes (based on link quality) to the number of
A game iteration begins with a single agent randomly possible connected nodes (i.e. within maximum coverage
positioned within the design environment. From this position range). This element encourages an agent to move to a position
where it can connect to the most nodes within its range. An
the agent is aware of its communication range and the sensors
assumption is made that in a situation where a node can
that are covered by this cell, the agent can evaluate the quality
connect to multiple sink devices it will connect to the one with
of the communication to these sensor nodes (with additional the strongest link. Over subscription (OS) is used to include the
considerations such as energy usage, balanced usage). As the number of devices that an agent has ownership of (strongest
agent is aware of the nodes that it cannot successfully act as a link) but for instance has a limit on the maximum number of
gateway for, it can make a decision to add intermediate nodes sensors it can cater for. By having this element the agent will
to aid in the communication, creating a multi-hop mesh move to a location that reduces the overlap with other agents,
network or it can add another agent to cover the remaining this also has a positive impact on signal to noise ratio for the
nodes. The agent uses its utility function to decide on what overall network. Global Coverage (GC) and Balance Factor (B)
action to take; as the agent is rational the only option is to take are included within the utility function to balance the signal
the action that will increase its own utility. coverage within the environment if all nodes are connected.
The actions that an agent can take include Move, Split and GC is calculated as the ratio of connected nodes to the
Die. The actions are also governed by general game rules such maximum possible nodes to connect too, this element supports
as move has a higher priority than a split action. The agent can the minimum number of agents in the environment also. B is
either Move to a better location, Split, resulting in a new agent calculated as the mean received signal of an agents overall
being created or Die. The last action does not improve an coverage map. The variables RU, OS, GC and B have
agent’s utility but improves the global optimal by eliminating normalized values that are within the range [0, 1]. The last
agents that are no longer necessary. It is only possible for an element of the utility is the weighting factor. The weighting
agent to split when an agent cannot improve its utility with a factor is used to allocate priority to an agent’s objective
through the single utility function. For example a higher weight
move action and there remain some unconnected nodes. In
should be put on maximizing resource utilization and as it is a
some cases a split may occur before an agent (or neighboring
secondary goal a lower weight should be put on the balance
agents) has moved to an appropriate location that can be a factor so it is less influential at the beginning of the
number of iterations ahead. This mechanism allows the optimization process.
optimization algorithm to quickly converge toward an
adequate solution. But as agents are myopic and cannot C. Multi-hop Scenario
foresee the impact of other agents on their decision the Die
action can be taken. At this stage an agent can decide to die if The multi-hop scenario differs from its single-hop
counterpart as sensor nodes can act as forwarders and repeaters
this action results in no sensors being left unconnected. The
can be used instead of sink nodes. In addition of finding the
algorithm converges when all sensors are connected to an
optimal number and positions of agents (sink nodes) the
agent and a Pareto optimal solution has been reached i.e. no network topology must also be defined. Given a set of sensors
agent can improve their own utility without impacting and a sink, several topologies can be created to connect the
neighboring agents. The utility function and optimization sensors to the sink using multi-hop communication. As part of
mechanism are described in the following. the optimization kernel a topology building algorithm has been
B. Single hop Scenario developed. This is a prerequisite for a multi-hop scenario as the
network topology must be defined in order to gauge the quality
For a scenario where a star configuration is required (only of an agent action. In addition the topology impacts the power
single hop) the agent is representative of a sink device and has consumption of the sensor nodes. In order to estimate the
a very basic objective, maximize the number of connected different levels of power consumption within the network, a
nodes within the environment. The utility function (U) used is battery model has been developed. The following will first
shown in (1). describe the algorithm developed to build the multi-hop
topology followed by the battery model description and finally
U = w1 RU − w2 OS − w3GC + w4 B (1)
the utility function used for the multi-hop WSN design
scenario.

644
1) Topology Building Algorithm been developed. The battery model combines duty cycle
Based on the position of the wireless devices and the parameters with the hardware power consumption
environment layout, the propagation model is used to predict characteristics to provide an estimation of node lifetime. While
RF signal attenuation between the different wireless devices. only providing a coarse estimation of a node lifetime, the
This information is used to predict link quality and is stored as developed battery model is useful to compare energy
an asymmetric graph (connectivity graph) where nodes expenditure between nodes. The electric charge of a sensor
represent wireless devices and an edge cost calculated as a node EC, expressed in mAh, is calculated according to (2).
function of the predicted link quality. The topology building Then the expected lifetime of the wireless device is calculated
algorithm extracts a set of sub-trees from the connectivity using (3).
graph. There is one sub-tree per sink node and each sub-tree
3600
corresponds to a cluster within the network. The algorithm EC = × (t a × I a + t s × I s ) (2)
determines to which cluster sensors should be associated to and ta + ts
the network topology within each cluster.
Multi-hop WSN topologies provide several possibilities CC
which impact network performance in terms of link quality,
NodeLifetime = (3)
EC
power consumption and communication delays. In order to The active and deep sleep duration of one duty cycle are
analyze long term large scale network performance, computer represented by ta and ts and are expressed in seconds. Ia and Is
simulation is typically used. However, using a simulation tool represent the current draw which is required by the wireless
for placement optimization is impractical as it requires long device in active and deep sleep modes and is expressed in mA.
processing time, is implementation specific based on network In active mode the radio chip is turned on and Ia is an average
and MAC layer protocols and requires expert knowledge of the of the current required to transmit and to receive. It is assumed
wireless networking domain and simulation tool itself. While that when a node is active equal time is spent on transmitting
Wi-Design does not implement a suite of routing protocols as and receiving data. The current capacity of the battery CC is
part of its toolset, in order to validate the quality of a design it expressed in mAh. NodeLifetime is expressed in hours.
is necessary to build a topology model. The topology model is
constructed based on the following metrics: 3) Utility Function
The utility function provides a normalized equation that
Link quality: a high quality link between devices is reflects the quality of a network design. The observations about
assumed when the received signal strength is above a particular network performance for building the multi-hop topology are
threshold. Link quality is an important factor as a faulty or the foundation for the utility function development. Equation 4
unreliable link can cause packet retransmissions which affect shows the utility used by an agent to evaluate the success of an
both lifetime and communication delays within the network. It action. It contains GC as described in the previous section as
can also lead to route instability. connectivity remains the primary objective.
Power consumption: in multi-hop networks, reducing the
transmission power and therefore using more hops to reach a U = w1GC + w2 LQ + w3 L + w4 IC (4)
destination reduces the individual power consumption of the Where,
sensor nodes. However it increases the global energy
expenditure of the network due to the increased requirement for LQ………………....…Link Quality
forwarding sensor data [37]. In order to account for this and L..………………….…Lifetime
reduce the global energy expenditure of the network, the
proposed approach builds the topology such that any sensor is IC…………………….Infrastructure Cost
connected with the minimum number of hops to a sink, while Link Quality (LQ) reflects the average link quality within
satisfying the requirement for reliable link quality. the agent’s cluster. LQ is calculated as the average of the edge
Data forwarding traffic: most of wireless sensors energy cost of the agents cluster as predicted during the topology
consumption is related to the radio activity [38]. Therefore a building algorithm. This parameter promotes topologies with
share of a node’s energy usage depends of the traffic a node is strong, reliable communication links. Lifetime (L) is used to
forwarding toward the access point. In order to avoid having represent the average lifetime of the sensors that are part of an
some nodes dying earlier than other which result in potential agent’s cluster. The sensors’ lifetime is affected by the network
bottleneck within the network the topology building algorithm topology and is calculated using the battery model presented
aims to harmonize data forwarding load among the different previously. Infrastructure Cost (IC) represents the cost
sensors of the WSN. effectiveness of a cluster. IC is calculated as the ratio of sensor
devices over total number of devices required for the cluster.
2) Battery Model This metric penalizes the addition of repeaters nodes and
During WSN operation, sensor nodes alternate between instead promotes the use of existing sensors to act as repeaters.
active states when they perform their sensing and This aims to reduce the infrastructure cost of a design.
communication tasks and deep sleep periods in order to save Similarly to the single-hop scenario, the variables GC, LQ, L
energy. A sensor’s duty cycle is based on its application and IC have normalized values in the range [0, 1], which
requirements and the amount of traffic coming from other allows the utility to be tuned using the weighting factors in
sensors that needs to be forwarded to the gateway. In order to order to allocate priority on an agent’s objective.
predict energy consumption, a minimalistic battery model has

645
V. CASE STUDY A sensitivity threshold of -80dBm was selected to
The following describes a sample design that was carried guarantee connectivity as a result the automatic optimization
out using the Wi-Design tools. Figure 3 shows the structure of suggests a total of 6 sink nodes are required to successfully
the NIMBUS building at Cork Institute of Technology that was connect all sensor devices. It can be seen from the figure that
used for the design. A wireless sensor network is required to all nodes are connected to a sink and the influence of the
monitor temperature, light and humidity across the first floor of balance factor has on the position of an agent. All agents are
the building (approx 2800 m2). After consultation with an close to the centre of the cluster of nodes they serve. Figure 5
environmental engineer specific sensing positions have been shows the signal coverage of each sink as predicted by the
defined and input using the Wi-Design interface marked with a propagation model within the environment.
red square on the floor plan shown in Figure 3. A total of 37
sensors are to be deployed operating at 2.4Ghz (IEEE802.15.4)
with a transmit power of 0dBm.

Figure 5. Single hop result showing signal coverage maps

B. Multi-hop Results
For the second part of the design the user wishes to assess if
a multi-hop topology was to be used what could be the saving
Figure 3. Nimbus Building Sensor Layout
in terms of infrastructure cost. In order to maintain the same
connectivity requirements as the single-hop scenario, a
The goal is to find the minimum infrastructure to support sensitivity threshold of -80dBm was selected. The size of a
reliable communication of data from 37 sensors distributed cluster is constrained to 4 hops between any sensor and a sink.
through the floor plan to the data warehouse. Sensors can Figure 6 presents the results of the design utilizing multi-hop
communicate with a sink node which is intern is connected to utility. As sensing nodes can be used as forwarders the
the back end data storage. The optimization must therefore optimization suggests that a minimum of 2 sink devices are
output the number and position of sink nodes and or repeaters required. A saving of 4 sink devices in comparison to a star
to meet these demands. topology. The automatic design allows a user to rapidly
evaluate the impact of different design constraints such as the
A. Single Hop Results limit of hops and topology configuration. The multi-hop
The Wi-Design tool was used to automatically optimize the approach is currently being extended to include the impact of
sink positions based on a star topology, the output of this network configuration on network lifetime estimation to
design is presented in Figure 4. provide the designer with a more detailed evaluation of a
design with respect to functional requirements.

Figure 4. Single Hop Optimization Results


Figure 6. Multi hop Optimization Results

646
VI. CONCLUSION [15] You-chiun Wang, Chun-chi Hu, and Yu-chee Tseng. “Efficient
Deployment Algorithms for Ensuring Coverage and Connectivity of
The paper presents a modeling and optimization tool set Wireless Sensor Networks” pages 114–121. IEEE, 2005.
that has been developed to address the need for deployment [16] Pinto, M. DAngelo, C. Fischione, E. Scholte, and A. Sangiovanni-
planning support for in building wireless embedded Vincentelli. “Synthesis of Embedded Networks for Building Automation
applications. The tool consists of a number of modules that and Control.” In Proceedings of the 2008 Americal Control Conference,
2008.
offer the user an easy to use modeling environment for
evaluating various network configurations. The paper also [17] Jiun-Jian Chang, Pi-Cheng Hsiu, and Tei-Wei Kuo. “Search-Oriented
Deployment Strategies for Wireless Sensor Networks”. pages 164–171.
described the automatic optimization algorithm that was IEEE, May 2007.
developed to enable even non expert users to generate a design [18] M. Younis and K. Akkaya. “Strategies and techniques for node
based on site specific application requirements. A case study of placement in wireless sensor networks: A survey”. Ad Hoc Networks,
a design was presented to show a sample design that was 6(4):621–655, 2008.
undertaken using the toolset. Although components of the [19] M. Kuorilehto, M. Hännikäinen, T.D. Hämäläinen, “Rapid design and
design tool such as the propagation model has been validated evaluation framework for wireless sensor networks” Ad Hoc Networks,
with real measurement, future work will involve validating the 6(6):909-935, 2008.
Wi-Design outputs with a large scale sensor network [20] J. Yick, B. Mukherjee, and D. Ghosal. “Wireless sensor network survey.
Computer Networks”, 52(12):2292–2330, 2008.
deployment within different application requirements. Wi-
Design tools can simplify the deployment process and provide [21] A. J. Motley and J. M. Keenan, “Radio coverage in buildings,” Bell
System Technical Journal (BTSJ), vol. 8, pp. 19– 24, Jan. 1990.
enhanced confidence in wireless deployments particularly for
[22] COST Action 231, “Digital mobile radio towards future generation
indoor applications. systems”, final report, European Commision, Brussels, 1999.
[23] M. Klepal, P. Pechac, and Z. Hradecky. “Prediction of Wide-Band
REFERENCES Parameters of Mobile Propagation Channel” XXVIIth URSI General
Assembly the International Union of Radio Science, Maastricht, page
[1] F. Zhao and L.J. Guibas. Wireless Sensor Networks: an Information 459, 2002.
Processing Approach. Morgan Kaufmann Publication, 2004.
[24] Alan Mc Gibney, Martin Klepal, Dirk Pesch, “A Wireless Local Area
[2] EnOcean, Range Planning Guide. Network Modeling Tool for Scalable Indoor Access Point Placement
http://www.enocean.com/fileadmin/redaktion/pdf/white_paper/WP_RA Optimization”, Annual Simulation Symposium, ANSS 10, April,
NGE_PLANNING_Jun09_en.pdf Orlando, Florida, USA, 2010.
[3] Johnstone, J. Nicholson, B. Shehzad, and J. Slipp. “Experiences from a [25] P. Wertz, M. Sauter, F. M. Landstorfer, G Wölfle and R Hoppe,
wireless sensor network deployment in a petroleum environment”. In “Automatic Optimization Algorithms for the Planning of Wireless
Proceedings of the 2007 international conference on Wireless LAN”, Proceedings of the 60th IEEE Vehicular Technology Conference,
communications and mobile computing, page 387. ACM, 2007. VTC2004 -Fall, September 2004.
[4] Lakshman Krishnamurthy, Robert Adler, Phil Buonadonna, Jasmeet [26] G. Wölfle, R. Hoppe and F. M. Landstorfer, “Radio Network Planning
Chhabra, Mick Flanigan, Nandakishore Kushalnagar, Lama Nachman, with Ray Optical Propagation Models for Urban Indoor and Hybrid
and Mark Yarvis. “Design and Deployment of Industrial Sensor Scenarios”, Proceedings of the 11th IEEE Conference on Wireless, July
Networks: Experiences from a Semiconductor Plant and the North Sea”. 1999.
2005.
[27] Y. Lee, K. Kim and Y. Choi, “Optimisation of AP Placement and
[5] Ankit Tiwari, Prasanna Ballal, and Frank L. Lewis. “Energy-efficient Channel Assignment in WLAN”, Proceedings of the 27th Annual IEEE
wireless sensor network design and implementation for condition-based Conference on Local Computer Networks, LNC’02, 2002
maintenance”. ACM Transactions on Sensor Networks, 3(1):1–es, 2007.
[28] A. Eisenblätter and H-F Geerdes, “Wireless Network Design: Solution-
[6] WS Jang and WM Healy. “Wireless sensor network performance metrics Oriented Modelling and Mathematical Optimization”, IEEE Journal on
for building applications”. Elsevier Journal, Energy and Buildings, 2010. Wireless Communications, Volume 13, Issue 6, MWC 2006, December
[7] Spinwave Systems. “Product Datasheet: Site Survey Tool - SWSST” 2006.
http://www.spinwavesystems.com/ last accessed 10th Febuary 2011 [29] E. Amaldi, A. Capone, M. Cesana, F. Malucelli and F. Palazzo, “WLAN
[8] Monnit Corp. “Wireless Placement Sensor”, http://www.monnit.com/ Coverage Planning: Optimization Modals and Algorithms”, Proceedings
last accessed 10th Febuary 2011. of the 59th IEEE Vehicular Technology Conference, VTC2004-Spring,
[9] Yung-tsung Hou, Tzu-chen Lee, Chia-mei Chen, and Bingchiang Jeng. May 2004.
“Node Placement for Optimal Coverage in Sensor Networks”, pages [30] K. Maksuriwong, V. Varavithya and N. Chaiyaratana, “Wireless LAN
352–357. IEEE, 2006. Access Point Placement using Multi-Objective Genetic Algorithm”,
[10] K. Kar and S. Banerjee. “Node placement for connected coverage in Proceedings of IEEE International Conference on Systems, Man and
sensor networks”. In Proceedings of Symposium on Modeling and Cybernetics, October 2003.
Optimization in Mobile, Ad Hoc, and Wireless Networks. WiOpt. 2003. [31] L. Nagy and L. Farkas, “Indoor Base station locations optimization
[11] S. Meguerdichian, F. Koushanfar, M. Potkonjak, and M.B. Srivastava. using Genetic Algorithms”, The 11th IEEE International Symposium on
“Coverage Problems in Wireless Ad Hoc Sensor Networks”. In Personal, Indoor and Mobile Radio Communications, PRIMC 2000,
Proceedings of Twentieth Annual Joint Conference of the IEEE September 2000.
Computer and Communications Societies (INFOCOM 01), Volume 3, [32] H. Jianjun and E. Goodman, “Wireless Access Point Configuration by
pages 1380–1387, 2001. Genetic Programming”, Congress on Evolutionary Computation, CEC
[12] X Wang, G Xing, Y Zhang, C Lu, R Pless, and C Gill. “Integrated 2004, June 2004.
Coverage and Connectivity Configuration in Wireless Sensor [33] R. M. Lewis, V. Torczon and M. W. Trosset, “Direct Search Methods:
Networks”. ACM Transactions on Sensor Networks (TOSN) Volume 1, Then and Now” Journal of Computational and Applied Mathematics,
Issue 1, August 2005. Volume 124, 2000.
[13] P.L. Chiu and F.Y.S. Lin. “A simulated annealing algorithm to support [34] Alan Mc Gibney, Martin Klepal, Dirk Pesch, “WLAN Design: A
the sensor placement for target location.” In Proceedings of Electrical Distributed Approach”, Proc. IEEE Vehicular Technology Conference
and Computer Engineering, pages 867–870. IEEE, 2004. Spring 2008, Singapore, May 2008.
[14] E. Chan. “Energy Efficient Fractional Coverage Schemes for Low Cost
Wireless Sensor Networks.” pages 79–79. IEEE, 2006.

647
[35] B. MacKenzie, S. B. Wicker ,“Game Theory in Communications: [37] Q. Wang, M. Hempstead, and W. Yang. “A realistic power consumption
Motivation, Explanation, and Application to Power Control” Global model for wireless sensor network devices”. In SECON ’06: Sensor and
Telecommunications Conference, 2001. Ad Hoc Communications and Networks, pages 286–295, 2006.
[36] Stuart Russell, Peter Norvig, “Artificial Intelligence, A Modern [38] G. Anastasi, M. Conti, M. Di Francesco and A. Passarella, “Energy
Approach”, Second Edition, Prentice Hall, 2003. conservation in wireless sensor networks: A survey”, Ad Hoc
Networks 7 (3) (2009).

648

You might also like