Development of a saturation flow rate prediction model at signalized intersections incorporating Geometrics, driver behaviour and motorcycle effects.

Abstract

This research addresses the challenge of predicting saturation flow rates at signalized intersections in mixed traffic conditions, where motorcycles, diverse vehicle types, geometric road elements, and varying driver behaviors significantly influence traffic flow. Existing models often focus on homogeneous traffic streams, making them unsuitable for regions with high motorcycle presence and mixed vehicle types. To bridge this gap, the study developed a saturation flow prediction model that integrates road geometric elements, motorcycle effects, and driver behavior. Data were collected using a Leica Aibot-AX20 unmanned aircraft for video graphic data and a CHC GNSS X900 survey machine for geometric measurements at signalized intersections. Traffic volumes were classified by vehicle type, and the positioning of motorcycles relative to other vehicles during green intervals was assessed, alongside factors such as turning movements for shared lanes, lane width, approach grade, lane changes, and conflicting traffic from other approaches. A linear regression analysis established the relationship between vehicle types and average headway, with passenger car equivalence units calculated as 0.38 for motorcycles, 2.63 for buses, 2.34 for light trucks, 2.91 for medium trucks, and 3.67 for heavy trucks, consistent with existing literature. The model, validated using fivefold cross-validation, showed statistically significant results, with evaluation metrics such as Mean Absolute Percentage Error and R-squared indicating that the model could predict well. Results showed that motorcycles positioned between vehicles explained 71.46% of the variance in vehicle flow rate, while motorcycles at the stop line accounted for only 6.45%. The final model indicated that lane width was the only directly proportional factor affecting saturation flow rate, while other variables had indirect relationships, explaining 62.28% of the variability in saturation flow rates. Recommendations for future research include incorporating green interval duration and pedestrian traffic as independent variables for determination of saturation flow rate. The model can be used to assess capacity and delays at signalized intersections, providing guidance to government transport planners for improving traffic flow.