الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 211 |  |  |
| | | from | 211 | | |
Abstract
Most methods of transportation planning studies require origin-destination (OD) matrices, which describe the trip demands between the origin and destination nodes in the network. The
estimation of OD matrices has traditionally been done using a direct technique for gathering OD
data from house and roadside interviews. However, these methods are not only very expensive,labor-intensive, and disruptive to trip makers; they also frequently fail to record traffic peaking behavior, which is important for traffic operational needs. Indirect procedures have become more
efficient and affordable as a result. This work is about the OD matrix estimation problem, which
means estimating OD matrices from observed link flows.
This research aims to provide an easy-to-use and cost-effective alternative software that
researchers and decision-makers in developing countries can access without restriction. The
Fmincon function in Matlab software was used to create specialized code for estimating the OD
matrix in congested networks using three different optimization approaches at the upper level:Generalized least square (GLS), Ordinary least square (OLS), and Maximum Entropy (ME) aggregated with user equilibrium assignment in the lower level.
The target OD matrix, the number of observed traffic counts, and the trip proportion map all
have a significant impact on the accuracy of the OD matrix estimation process. For this reason,
reliability factors (𝛾1, and 𝛾2), and trip proportion maps were added to the proposed software to
account for all possible scenarios. Two hypothetical networks and a part of The Taif Network
were used to test the performance of the provided software. The improvement rate in error between observed and estimated traffic counts in the three
networks was (99.8%, 75.39%, and 32.77%) according to the OLS approach, and (99.78%,
75.56%, and 19.04%) according to the GLS approach. The ME approach improves the error rate
of the two hypothetical networks only by (97.52%, and 52.11%) respectively. Also, to guarantee
performance, the results were compared to PTV Visum’s TFlow fuzzy (TFF) and least square
(LS) results. The results show that both the GLS and OLS methods can produce extremely
accurate results, with the mean absolute error value approximating the results of the LS method by at least 98%, and the error rate is reduced by at least 10% over the TFF method.