Application of conditional demand analysis in Origin-Destination (OD) matrix estimation using traffic counts and zonal characteristics
Student thesis: Master's Thesis
Related Research Unit(s)
|Award date||4 Oct 2004|
Origin-Destination (OD) matrix is the fundamental information in planning and management of transportation system. With this information, traffic engineers can make appropriate planning decisions in a bid to release the stress on congested roads. Traditionally, the conventional 4-step transportation model has been used to estimate the OD matrix and link choice proportions for more than two decades. However, this approach is based on the results from large-scale household or roadside surveys, and thus, the cost is very high. To avoid the cost of conducting expensive household or roadside surveys, many transport researchers have proposed statistical approaches to estimate the OD matrix using traffic counts that can be observed directly via detectors on traffic links. In this study, based on the concepts of the conventional 4-step transportation model and statistical approaches using traffic counts, a combined trip generation, trip distribution and traffic assignment model is proposed, which makes combined use of the Conditional Demand Analysis (CDA) and the Traffic Assignment (TA) technique. Conditional demand analysis is a technique first used in the estimation and prediction of household’s total electricity demand. This approach aims to disaggregate a household’s total electricity consumption into demand functions of different electric appliances. The electricity consumption rates of different appliances are estimated using the ownership pattern of electric appliances and demographic characteristics of the households. Conditional demand analysis is adapted and modified in this study to disaggregate observed traffic counts into different OD pairs in the transportation network using zonal characteristics and estimated link choice proportions. Traffic assignment is incorporated in the proposed algorithm to assign estimated OD flows into different traffic links by the method of Successive Average (MSA) under the principal of Stochastic User Equilibrium (SUE). Multinomial Logit (MNL) model is used in traffic assignment to determine the chosen routes based on the costs of the routes. The dispersion parameter θ governs the sensitivity towards cost. When θ is zero, path choice is insensitive to cost and when θ tends to infinity, path choice becomes concentrated on the least cost alternative. In many previous transportation studies, it is assumed that there exists only one dispersion parameter, which implies homogeneous traveling behaviours of travelers. However, in reality, different travelers may have heterogeneous traveling behaviours. In order to examine the feasibility of applying the proposed algorithm in both homogeneous population and heterogeneous population, random coefficients approach is considered in this study. In this study, a regression model is formulated by expressing OD flows of different OD pairs in terms of OD trip rates and their corresponding zonal characteristics. Then, based on the framework of conditional demand analysis, observed traffic counts can be expressed in terms of link choice proportions, OD trip rates and zonal characteristics. Bayesian approach is used to estimate OD trip rates and dispersion parameter θ simultaneously by optimizing the posterior probability density function (pdf). Secondly, OD matrix is obtained using the estimates of OD trip rates. Thirdly, the estimated OD flows are assigned to traffic links using the estimates of dispersion parameter θ by means of traffic assignment. In this stage, link choice proportions are updated. These three stages are repeated until the value of the objective function stabilizes. In this dissertation, chapter 1 briefly states the trend of studies in the estimation of OD matrix in the past two decades, from using conventional 4-step transportation model to statistical approaches using traffic counts. It is then followed by literature reviews in chapter 2. A review of the use of conditional demand analysis in the estimation of energy consumption is also included in this chapter. Chapter 3 presents the model specification and the estimation algorithm. In this algorithm, the combined use of conditional demand analysis and traffic assignment is proposed, where the conditional demand analysis model is appropriately adjusted. Also, a comparison of the similarities in using conditional demand analysis in energy consumption estimation and OD matrix estimation is made. Chapter 4 describes the validation of the new algorithm by using a hypothetical example. Three research designs including a method of data simulation are described. The new algorithm is then applied to the simulated data and findings are discussed. The results of the validation clearly demonstrate that the new algorithm of combined models can provide reasonable and reliable estimates and updates of the OD matrix using traffic counts and zonal characteristics. Finally, conclusion and discussions are presented in chapter 5.
- Traffic flow, Traffic estimation, Traffic assignment, Origin and destination traffic surveys