Destination choice: Model estimation using destination end surveys

This dissertation documents the development and use of survey data collected at the production and destination ends of travel in the 1993 Beaumont-Port Arthur travel survey for the calibration of vehicle trip multinomial logit destination choice models (DCM). The models are applied and singly constrained (to trip productions) trip tables developed for three trip purposes, home based work, home based non-work, and non-home based. A doubly constrained (total trip productions and trip length frequency distribution) trip table is developed based on a random number generator. These trip tables and the calibrated trip table based on the current four step practice model are compared, The trip tables are assigned to the transportation network using an equilibrium assignment model and the results compared between the different models and to the vehicle counts in the study area.; The objectives of this research were to examine the results of household and work place surveys for the area under study, examine the DCM calibrated from the surveys, and examine the results from applying the models in conjunction with the current practice model and a random number model for both trip tables and traffic assignments.; It was concluded that data from destination end surveys may be used to calibrate DCM. The DCM models for the non-work trip purposes did not produce trip tables with average trip lengths or trip length frequency distributions (TLFD) that matched those observed in. the surveys. Constraining the DCM models to match the TLFD produced assignments that essentially matched those for the current model. The random number model assignments also essentially matched the current model. The trip tables for all of the models were found to be different but the aggregative nature of the assignment process produced results that were similar except for the unconstrained DCM models. The constrained DCM models examined in this research produced acceptable results at the assignment level when compared to the current model for the study area.