Basic GCARS System Concept
The central and most basic concept of the GCARS System was the appli-
cation of minimum path analysis techniques to "numerical cost models"
so as to generate a series of ranked alternatives. This concept is shown
in Figure 2, where the various "numerical cost models" are shown as solid
three-dimensional surfaces. In actual practice, they are stored as ma-
trices within the computer. This concept has been described by some
users as "linear programming with maps".
Desirable routes will follow the "valleys" across such "cost models".
The most desirable combines directness and low "elevations" so as to
obtain the lowest "total cost". Less desirable routes follow other
valleys and "passes" over the intervening "high cost" areas. Sometimes
such alternatives are shorter than the first choice, and although having
a higher "cost" per unit length, may be more desirable. Thus the var-
ious choices should be compared in terms of overall length and "total
cost",
Minimum path analysis can be used to locate such valleys and alternate
routes. A grid network is formed from the "cost models" matrix by join-
ing all nodes. Each link is thus assigned the "cost" of traversing it,
thus minimum path analysis will discover the optimal path. Ayad (2)
proposed a method of generating a series of significantly different
ranked alternative choices. If the central links forming the optimal
route are raised in value, their re-use will be inhibited and re-
analyzing the revised network will produce a second minimum - a "second
best" alternative. Repeating the process will allow the generation of
a ranked series of alternatives. Comparison of these paths allows for
sensitivity analyses and leads naturally into impact assessments.
Figure 2 also shows that models for several factors can be superimposed
and summed to produce "cost models" for any desired combination of
factors. Before summation each model can be multiplied by a weighting
factor, allowing it to be enhanced to any desired degree. Repeating
minimum path analysis on networks derived from such combined models
will generate a series of ranked alternatives in terms of combinations
of factors. In the newer GMAPS-GCARS programs this model building con-
cept has been expanded greatly and is contained within the GMAPS por-
tions of the system.
The GCARS I System
In order to satisfy the goal of machine independence (Goal 1), all
GCARS I programs were written a version of FORTRAN IV which closely
followed the basic USASI Fortran standard. The programs were subse-
quently made to run on IBM 360 series or CDC 6000 series computers with
only minor changes.
The goals of economy and interactive man-maching dialogs (Goals 2 & 3)
dictated certain forms of programming and the extensive use of matrices
to represent models. Graphical routines were developed which would