Weaving areas are critical elements in the operation of the freeway network, involving complex vehicle interactions. Many of the existing procedures for the design and analysis of weaving sections are based on data collected in the late 1960s and mid-1970s and may not reflect current driver/vehicle characteristics, especially in California. This paper describes what the analysis of more recently collected data has revealed. As part of a 2-year project to develop improved weaving analysis procedures, a large amount of data were collected using a video camera at eight major freeway weaving locations throughout California. Information on geometrics, volumes, and speeds of weaving and nonweaving vehicles were then extracted from the tapes, thoroughly checked, and verified with field observations. Six existing methods for the design and analysis of freeway weaving sections were then applied to all data sets. Results indicate that significant discrepancies exist between the predicted and measured average speeds of weaving and nonweaving vehicles. Several statistical analyses were performed using regression analysis and classification and regression trees, to identify basic relationships between weaving section design and traffic characteristics. Regression techniques were also employed in an effort to improve existing analysis and design procedures and to develop new performance prediction models. One important finding is that speed was insensitive to changes in geometric and traffic factors over the range of values in the data set. Overall, analyses suggest that average travel speed may not be an ideal measure of effectiveness. More research needs to be carried out to develop a more accurate procedure for designing and analysing freeway weaving sections. Such procedures could consider measures of effectiveness other than speed. This paper also discusses the direction of possible future research.
Samenvatting