Wireless Network Design: Optimization Models and Solution ...

6 Optimization Based WLAN Modeling and Design 135
signed to quickly create a good design (but not necessarily optimal) and comprises
several modules: i) a database module to create and store a digitized version of
the building, ii) a propagation prediction module that implements a 3D ray-tracing
algorithm to determine propagation loss between two points, iii) an access point
optimization module for transceiver location, and iv) a graphical user-interface for
coverage display.
For the case of n base stations, where (x j,yj,zj) is the location of base station j,
the optimization problem is to maximize f (x1,y1,z1,...,xn,yn,zn). A grid is superimposed
on the floor plan and f is defined as the ratio of the number of grid points
with received power sufficient for communication to the total number of grid points.
The authors use a direct heuristic search method that exploits the characteristics of
this design problem and report that there are many local optima.
Panjawni et al. [14] developed a software system called Site Modeling Tool
(SMT) to determine good locations for radio frequency transceivers. Their system
is an extension of AutoCAD and runs on either a PC or workstation. Under the assumption
that the coverage regions are connected sets without holes, the coverage
regions are modeled as polygons. Their algorithm involves a search for boundary
points of these polygons. The first boundary point is located by searching outward in
a straight line until communication is not possible. Each additional boundary point
is found in a similar manner by a sweep in a counterclockwise direction around the
transceiver. The authors use 36 boundary points to define the coverage region polygon.
The system supports both single and multifloor buildings and provides a visual
display of the solution which allows a designer to determine the locations of the
transceivers. Once transceiver locations are placed, the SMT calculates and displays
the expected coverage.
Hills [8] describes the design process used to deploy the wireless infrastructure
(Wireless Andrew) for Carnegie Mellon University. Wireless Andrew covers 65
buildings with a total floor area of approximately 3 million ft 2 . A five step iterative
approach is used to deploy the WLAN: i) initial selection of AP locations, ii) use
signal strength measurements to adjust AP locations, iii) create a coverage map, iv)
assign frequencies to APs, and v) document the system. The report recommends using
a coverage-oriented design in low density areas and a capacity- oriented design
in high density areas.
Two very sophisticated systems, one academic (S 4 W) and one commercial
(SitePlanner R○) are discussed in Skidmore et al. [19]. Both systems use state-of-theart
channel modeling, global optimization procedures for transmitter placement, and
a simulation system for performance evaluation. The optimization modules attempt
to minimize average shortfall of penalty functions of coverage and bit error rates.
These systems are defined as Problem Solving Environments (PSEs) for wireless
network design and analysis.
Longo [11] and colleagues at OPNET Technologies, Inc. developed the WLAN
Design Solution for the Department of the Navy as part of the Small Business Innovation
Research (SBIR) program. Their system is designed to optimize AP placement
for military ships or other military structures. The system is combined with a
simulation module that can be used to estimate throughput for a proposed design.