Speed Zoning Why And How PDF Download

The objective of this study was to quantify effects that selected characteristics and adjacent development patterns have on roadway speeds. Based on the results of a literature search and the availability of data for Kansas highways, twenty-seven variables were identified as possibly affecting speeds on rural state highways and thirty-two variables were identified as possibly affecting speeds on urban state highways. Speed data and data for the potential explanatory variables were collected for a total of 539 sections of state highway (186 rural and 353 urban sections). Two approaches were used to develop and test models to predict speeds on rural and urban state highways based on roadway characteristics and adjacent development patterns. The first approach was based on models in the form of multiple linear regression equations. The second approach employed artificial neural networks (ANNs) to predict highway speeds. None of the regression models were entirely satisfactory in terms of their ability to predict the 85th percentile speeds on rural and urban highways within +/-5 mph (+/-8 km/h). A number of regression models are presented, however, they should be used with caution. Two independent databases were used to train two sets of ANN models of rural and urban speeds. The first set of ANN models (Stage 1) was developed using the same data used in the regression analysis. The second set of ANN models (Stage 2) was developed using a database provided by the Kansas Department of Transportation (KDOT) Bureau of Traffic Engineering. The second database contained only those variables that KDOT believed drivers consider in selecting a driving speed. Overall, the Stage 2 ANN models developed in this study were found to perform much better than either the Stage 1 ANN models or the regression models in predicting rural and urban highway speeds.