Travel time reliability

Travel time and travel time reliability are important performance measures for assessing traffic condition and extent of congestion on a roadway. Most commonly used methods to obtain travel time data either produce only estimates of travel times or too few travel time data points for meaningful analysis. This study focuses on using a new probe vehicle technique, the Bluetooth technology, to collect two weeks of travel time data on Interstate-69 in Indianapolis. These data are then used to estimate econometric models, which can be used to predict freeway segment travel times. First, an autoregressive model is estimated based on the collected data. Individual vehicle travel times on a freeway segment are expressed as a function of speed, volume, time of day indicators, and previous vehicle travel times. In addition to the autoregressive formulation, a duration model is estimated based on the same travel time data. The duration model enables calculation of the probability of the vehicle exiting the segment of the road at any point in time. The estimated models indicate that the rate of vehicles exiting the segment as a function of their travel time rises initially until the inflection point and then decreases. It is hypothesized that the inflection point occurs at the onset of congestion, when longer travel time may not result in a higher probability of exiting the freeway segment. Lastly, a seemingly unrelated regression equation model to predict travel time and intervehicle variability is proposed. This model predicts 15-minute interval travel time and the standard deviation of travel time based on speed, volume and time of day indicators. The estimated model shows a good fit with the data. Furthermore, the results indicate that it is superior to the model based on point-speed estimates, which is commonly used in practice. Thus, the SURE model can be used to improve real-time travel time prediction