(contributed by Diane Larin, INRO Consultants)
Observed link volumes are essential when one attempts to validate a traffic assignment model. In practice, such volumes are obtained by installing count posts on some of the links of the network. Of course, the utility of this information increases when the location of the count posts is suitably chosen, with the objective of capturing as much auto traffic as possible.
If one is not sure that the current set of count post links is satisfactory, it may be desirable to locate additional count posts on the network or, if feasible, to find a completely new set of count posts. What would then be the best strategy?
An appropriate approach is to attempt to identify those links which carry most of the flow on the paths to which no count posts have yet been assigned. This is exactly the purpose of the macro CNTPOST. It finds the best next potential count post link, given an existing set of count posts. This set may be empty.
The macro uses the Additional Options of the auto assignment to find the paths which do not contain any count post links, and the Network Calculator, to find the link with maximum flow on these paths. For each assignment, one count post is selected.
The data requirements to run this macro are: one link user data which
contains information on the status (observed volumes,
no permission to locate, etc.) of the count posts on each link,
and six available scalar matrices (ms91 to ms96)
to store the intermediate results.
The macro is called with two mandatory arguments and up to four optional ones:
~<cntpost count gpq [optional arguments]
where count is an integer that specifies the number of iterations desired, each iteration of the macro yielding the selection of a new count post, and gpq is the demand matrix. The optional arguments allow the user to perform two more analyses: the computation of the percentage of the not counted demand, and the generation of histogram comparisons of counted demand versus not counted demand.
The procedure was tested on our demonstration data base and proved to be quite efficient. With the original set of 70 count post links, the percentage of the not counted demand was around 60%. By adding just 12 count post links to this set, this percentage dropped by half.