Document Type

Journal Article

Publication Date


Subject Area

infrastructure - bus/tram lane, ridership - commuting, place - rural


Two lane roads, Two lane highways, Speed limits, Rural highways, Neural networks, Kansas, Field studies, Backpropagation, Artificial neural networks, ANNs (Artificial neural networks), 85th percentile speed


Recent federal legislation allowing states to set their own speed limits on highways, as well as increases in the number of requests from citizens and neighborhood groups to implement actions to reduce "excessive" speeding on their streets and highways, has created considerable debate about and scrutiny of the appropriate speed limits that should be posted on state highways. Various speed studies have indicated that sensible and cautious drivers will most likely drive at the speed dictated by roadway and traffic conditions rather than relying on a posted speed limit. To incorporate roadway characteristics and traffic volumes into the selection of the most appropriate (i.e., comfortable, safe, and efficient) speed limit, actual engineering field speed studies are carried out. Generally, the 85th percentile speed at which the drivers surveyed are driving is selected as a primary factor in determining the posted speed limit. Carrying out such field studies for all highway sections is a costly and time-consuming process. Therefore, characterizing the relationship between the 85th percentile speed and the roadway characteristics will assist in selecting the most appropriate posted speed limit on highway sections where field surveying is difficult due to resource limitations. A back-propagation neural network is used to extract the relationship between roadway characteristics and 85th percentile speed. The developed neural-network-based speed model was found to perform satisfactorily for characterization of speed on Kansas two-lane, uninterrupted-flow rural highways and for quantifying the influence of prevailing roadway characteristics on the anticipated 85th percentile speed.