![](bg2.png)
generate the alignment grade line, several vertical
intersection points (VIPs) are fixed and
interconnected successively to form a vertical
piecewise linear trajectory. The number of VIPs is
mainly affected by the variations in ground
elevations. Parabolic curves are then fitted at VIPs to
depict the vertical alignment. The vertical grade line
of the alignment is then evaluated based on the
design requirements and the amounts of earthwork
for both cut and fill sections. The process will
continue repeatedly until finding the most suitable
one.
The selection of the final alternative alignment is
accomplished by focusing on the detailed design
elements. HIPs, deflection angles, curve radii, VIPs,
tangents, grade values, and sight distances are
among the design elements of highway alignment in
3D. Most of these design elements are constrained
by standard limits described by such documents as
the Design Manual for Roads and Bridges (DMRB,
1992-2008) and AASHTO design standards
(AASHTO, 1994).
As the two processes are considered apart from
each other, the generated alignment likely represents
a local optimum rather than a global one. This
approach takes into account many design elements
and, at the same time, neglects numerous possible
solutions due to non simultaneous consideration of
both alignments. This process is also very expensive
in terms of time.
Researchers have tried to speed up the process of
highway alignment planning and design and to find
better solutions. Attempts have been done to
optimize either horizontal or vertical or both
simultaneously. Calculus of variations by Shaw and
Howard (1982), numerical analysis by Chew et al
(1989), linear programming by Easa (1988), and
genetic algorithms by Jong (1998), Fwa et al.
(2002), and Tat and Tao (2003) are some of the
techniques that have been used. The work done by
Jong (1998) has also been extended to incorporate
more cost components, GIS integration, and to
formulate the model to handle the problem as a multi
objective problem. All these can be seen in (Jong
and Schonfeld, 1999) (Jha and Schonfeld, 2000)
(Maji and Jha, 2009). It should be noted that all
these studies are based on the conventional design
principles of highway alignment design which
consider HIP, VIP, tangents, and curve fittings.
Since its introduction, despite the extreme
development in computers and highway surveying
field instruments technologies (e.g. total station),
highway engineers and planners are still using the
same convensional design approach. None of the
studies has exploited the technology development to
explore the possibility of changing some ideas
imposed on highway alignment planning and design.
A question arises here, do we still need to keep the
same planning and design approach or do we need to
change to reflect technology development? That is
the question that this study seeks to answer.
1.2 The New Approach
This study introduces a novel technique for
alignment optimization. It suggests optimizing
simultaneously the horizontal and vertical alignment
of a highway through station points. Station points
as points along the centre line of alignment, which
are defiend by their X, Y, and Z coordinates, are
used to define the alignment configuration. This
research study is inspired by the fact that any
generated alignment by whatever method will finally
consist of a series of station points and it will be
implemented on the ground depending on those
station points. Figure 1 shows the difference in
alignment generation and configuration between the
traditional and proposed method.
In this study GA, as an evolutionary adaptive
search technique (Beasley et al, 1993), is used to
perform the search. Some modifications to suit the
nature of the problem have been included (Davis
1991; Mitchell 1996).
A variety of studies have proven that GA is an
efficient tool for planning and optimization
problems. Mathews et al (1999) applied GA to land
use planning, Mawdesley et al (2002) used GA for
construction site layout in project planning, Ford
(2007) used GA for housing location planning, Jong
(1998), Fwa et al (2002), Tat and Tao (2003), and
Kang (2008) used GA for alignment optimization
problems.
2 THE MODEL FORMULATION
2.1 The Study Boundary
The study area is defined and divided into
rectangular grid cells usually produced from a GIS
model of the area under consideration. The size of
the grid cells falls within the user preferences and
depends on the desired accuracy. Each grid cell may
handle one or more than one average value. In this
study two different values are assigned to each cell.
Average land unit cost values are used for the
alignment location dependent cost calculations while
average ground elevations are used to calculate the
ICEC 2010 - International Conference on Evolutionary Computation
130