Advances on stochastic traffic assignment model for driver information system applications

Abstract

Most metropolitan areas around the world face a variety of problems, including serious traffic congestion and high accident rates during peak hours. Upgrading the existing expressway networks by the application of electronic information and communication technology is one way to curtail these problems. Recently, increased attention has been paid to the provision of the driver information system (DIS) such as the estimated journey times on major routes for drivers. The DIS should help drivers to select better routes and guide them to utilise existing expressway networks efficiently. This can be regarded as one possible strategy for effective traffic management. For a traffic surveillance and information system using the DIS, the number of speed detectors (or probe vehicles) required is very important because it affects the precision of the measurement of the vehicular travel times. In this study, a bilevel programming model is proposed to determine the number of speed detectors in a network with travel time information provided by the DIS. The lower-level problem is a stochastic probit assignment model to take account of the effects of various measurement on and perceived errors in travel times. The upper-level problem is to determine the optimal speed detector density that minimizes the measured link travel time errors, as well as the social cost of the speed detectors. Numerical examples are used to illustrate the application of the proposed model and a solution algorithm for determining the optimal speed detector density. Apart from the cost of the DIS (in terms of the social cost of the speed detectors), it is also important to examine the benefits of the DIS for evaluation purposes. In this study, a time-dependent stochastic probit assignment model is proposed for assessing the benefits of providing the DIS. The simulation technique, the shortest path algorithm and the method of successive average (MSA) are adopted for solving the proposed stochastic probit assignment model. By applying the proposed model and solution algorithm to an example network with and without the DIS, the total network travel times can be computed and compared. The total network travel times are used to evaluate the effectiveness of the DIS. The previous related research work on assessing the DIS show that the accuracy of the provided travel time information affects the benefits arising from the DIS provision. Therefore, a new method is proposed to estimate link travel times based on speed detector data in comparison with the other two existing methods. In order to evaluate the estimated link travel times, a manual license plate survey is conducted for collecting the observed link travel times at the Lantau Link Corridor connecting to the Hong Kong Chek Lap Kok (CLK) international airport. The observed link travel times are then used for the validation of link travel times estimated by different methods. Although the above various models have been proposed for assessing the effects of the DIS, it is still necessary to validate the drivers' responses to the DIS in practice. Thus, a stated preference (SP) survey is conducted in the Hong Kong CLK international airport for collecting the drivers' route choice responses. This study area is chosen because the DIS (via variable message signs) provided along the Lantau Link is new to Hong Kong drivers. The collected drivers' route choice responses are used to calibrate the drivers' route choice model (i.e. the SP logit model). A case study is carried out to evaluate the proposed time-dependent stochastic probit assignment model by comparing its results against the SP logit model. In short, there are three principal findings of the study. The first principal finding is that the proposed time-dependent stochastic probit assignment model can be used to assess the effects of the detector density and location of VMS separately under the DIS environment. The second principal finding is that the numerical results of the bilevel programming model show that detector density required for DIS increases when O-D demands increase. It is because larger O-D demand results in larger link travel time and larger measured travel time error variances. The third principal finding is that the calibrated driver perception errors of the stochastic probit assignment model increase as the volume/capacity ratios increase.