Several traffic agencies and centers now have installed vehicle toll tag readers to collect tolls on some roads in their jurisdictions. They have realized that these may also be used to collect information on the traffic performance, such as expected travel times and reliabilities. One can locate tag readers to obtain travel times on portions of the routes in the transportation network. A natural objective is to maximize the total vehicle miles monitored in the network for a given number of sensors. The problem can be formulated as a 0-1 mathematical program. This paper addresses also the issue that some links have higher travel time variances than others. To monitor the network for predicting travel times, one must now take into account both the mean and variance of links. Another model is presented to determine on which routes to locate sensors to minimize the variance of the predictive distribution of network travel time. The problem is formulated as a 0-1 program. A case study for illustrating the models is given using data from Harris County, Texas.
Abstract