Delay distribution estimation at a signalized intersection

Previous studies found that delay at signalized intersection accounted for more than one-third of total travel time. Different level of services tends to have different delay probability. An early assumption was made that multimodality occurs because of two traffic states experienced by the traveler; delay and non-delay at the signalized intersection. This study proposed a method to quantify delay distribution at a signalized intersection by analyzing the patterns in the speed time and speed distance profiles when passing two consecutive intersections and redrawing vehicle trajectory generated by 'second by second' GPS data. The searching algorithm was developed to search for the times when the vehicle enters and leaves the delay section by checking the `second by second' speed data. To differentiate between the queue, move up and stop and go traffic, the algorithm searches for the idle time (i.e. when the speed less than 3.5 km/h). Along with Sydney Coordinates, Adaptive Traffic Systems (SCATS) degree of saturation and signal settings (green and red time) generating from SCATS systems at upstream loop detectors, a realistic delay probability model also was developed. This model used the General Additive Model for Location, Shape, and Scale (GAMLSS) which allows location, scale and shape parameter of selected distribution as a function of the explanatory variable. Signal settings and SCATS degrees of saturation were used as an explanatory variable. This model enables us to estimate delay experiencing by traveler depending on traffic parameters including traffic flow, signal settings and degree of saturation that are readily available on the site.