THE ESTIMATION OF COMBINED MODEL TRIP DISTRIBUTION AND MODE CHOICES PARAMETER BASED ON TRAFFIC COUNT UNDER EQUILIBRIUM ASSIGNMENT

• Main Author: SULISTYORINI (NIM : 35005005); Tim Pembimbing : Prof. Ir. Ofyar Z.Tamin, M.Sc.(Eng), Ph.D; , RAHAYU
• Format: Dissertations
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/16896
• Institution: Institut Teknologi Bandung

Travel is an activity that has become part of our daily life in order to fulfill their basic requirements that are not available at their original place. In the particular <br /> <br /> <br /> level, this necessity to travel will create problems such as congestion, air pollution and environment. The effort to solve the problems is to understand the pattern of people and good’s movements in certain periods of time in area by using available information in Origin-Destination Matrix. <br /> <br /> <br /> Methods for estimating an O-D matrix can be classified into two main group, conventional and unconventional methods. Conventional method rely heavily on extensive surveys, making them very expensive in terms of manpower and time, <br /> <br /> <br /> disruption to trip makers and most importantly the end products are sometimes short-lived and unreliable. As a result of dissatisfaction expressed by transport planners with conventional methods, other techniques for estimating O-D matrix which based on traffic counts have evolved over the years, these are generally called unconventional methods. This method use traffic counts as main input to <br /> <br /> <br /> estimate unknown parameter in trip distribution stage. Traffic counts have many advantages, because of some reasons like: there are routinely collected, easily <br /> <br /> <br /> available, relatively inexpensive in terms of time and manpower and also without disrupting the trip makers. <br /> <br /> <br /> However, the accuracy of the estimated O-D matrices are strongly influenced by several factors which include: (1). The choice of transport demand model itself; (2). The estimation method used to calibrate the model from traffic counts; (3). Trip assignment techniques used in determining the route choices taken through the network; (4) Location and number of traffic counts; (5). The level of errors in <br /> <br /> <br /> traffic count; (6). The level of resolution of the zoning system and the network definition; (7). finally, other factors such as combined model of trip distribution <br /> <br /> <br /> and mode choice. <br /> <br /> <br /> The stressing of the research development is to estimate unknown parameter of transport demand model by combining trip distribution, mode choice and route choice in single process, which have a main objective that estimate this parameter using traffic count information. The model examined was the Gravity (GR) model with exponential negative as deterrence function, combined with the Multi- <br /> <br /> <br /> Nomial-Logit (MNL) model. Non-Linear-Least- Squares (NLLS) estimation methods were used to calibrate the parameter of the combined model. Once, the parameters have been calibrated, they may be used not only for the estimation of <br /> <br /> <br /> the current O-D matrix, also for predictive purposes. This research is developed from the previous research, which accommodate transit in realistic condition in urban network. <br /> <br /> <br /> Furthermore, several factors affecting the accuracy of the estimated O-D matrices will be studied are as follows: (1). The effect of incorporating errors in traffic count information; (2). To obtain the optimum traffic flow with observed location and number of traffic flow data; (3). The effect of applying two methods (random and sorted method); (4). and the effect of trip assignment techniques (All or <br /> <br /> <br /> Nothing and Equilibrium). The model approach has been tested using artificial network system (simple and complex) and Bandung network as real network. The finding of research is presented in terms of comparing estimated O-D Matrix and Traffic Volume with Origin-Destination Matrices and traffic volume observed. <br /> <br /> <br /> Several important findings in artificial and real data can be concluded as: (1). The convergence of (beta) and (gamma) value depends on starting value of (beta) and (gamma). If the starting value of (beta) and (gamma) more away from the solution, it needs iteration more to achieve convergence level; (2). The more errors in traffic count, the less <br /> <br /> <br /> accurate the estimated O-D matrix and also the less traffic counts you have, the less accurate levels; (3). It can be seen that the better accuracy of the estimated OD matrices are obtained under sorted condition rather than under random condition; (4). It can be finally proven that the use of equilibrium assignment in the trip assignment stage also gave significant impact compared to the use of allor- <br /> <br /> <br /> nothing assignment; (5). The result of O-D Matrix and traffic flow estimation compare with observed, expressed by statistical value of estimation with R2 less than usual. <br /> <br /> <br /> Further research is underway in developing algorithm to achieve simplified process and to analyse the impact of others factors of the O-D matrix estimation from traffic count information.