22-01-2013, 11:06 AM
Computing turn-dependent delay times at signalized intersections based on floating car data
ABSTRACT
Current link-based travel times for (urban) road networks play an important role in dynamic navigation and routing. They are basic ingredients when searching for fastest paths between given locations on a digital map. Typical routing applications use at least static travel times for each link which are based on road types and speed limits. However, static daily variation curves with traffic-state-dependent travel times are used more and more. Sometimes even current traffic states are considered if there are suitable traffic data available.
Nevertheless, in most cases no distinction is drawn between different turning directions at (signalized) intersections or there are only some very raw static assumptions realized. The true travel time, however, significantly depends on whether a vehicle is driving straight on or turns right or left at an intersection. Furthermore, the physical structure of the considered road junction influences the turn-dependent travel times, too. When turning left (in case of right-hand driving) vehicles normally have to give way to oncoming traffic first. And, particularly in urban areas, they sometimes have to wait for pedestrians crossing their way when turning left or right. The assumptions indicate that considering turn-dependent delay times will result in a much more accurate navigation and routing performance.
This paper describes a new approach how to automatically compute the required turn-dependent delay times based on common floating car data. Here, a number of probe vehicles provide information about their current positions every few (typically about 30) seconds so that travel times can be obtained for each pair of data points received from the same car. Filtering for the different turning directions at a given (signalized) intersection finally allows computing travel times for each traffic stream separately.
The challenge is that these travel times even in case of the same turning direction are related to very different driving distances as well as they are not link-based in general. However, the decomposition to obtain so-called link travel times or delay times is not trivial as has been shown in the literature. Accordingly, this paper presents a new idea to solve this task by introducing a simple linear model for the measured travel times which arises from the superposition of turn-dependent delay times and a basic travel time depending on driving distance and average flow resistance on the relevant road sections. The unknown parameters are then computed by (weighted) linear regression.
As a main result, the approach yields plausible estimates for turn-dependent delay times as is demonstrated for several signalized intersections in the German city of Hamburg where data of more than 1500 vehicles are available to the authors for several years. The results confirm the nave assumptions mentioned above and quantify the effects. Furthermore, the results indicate that the method may also give hints to the designers of traffic signal plans where in a city and how to improve the signalling.
ABSTRACT
Current link-based travel times for (urban) road networks play an important role in dynamic navigation and routing. They are basic ingredients when searching for fastest paths between given locations on a digital map. Typical routing applications use at least static travel times for each link which are based on road types and speed limits. However, static daily variation curves with traffic-state-dependent travel times are used more and more. Sometimes even current traffic states are considered if there are suitable traffic data available.
Nevertheless, in most cases no distinction is drawn between different turning directions at (signalized) intersections or there are only some very raw static assumptions realized. The true travel time, however, significantly depends on whether a vehicle is driving straight on or turns right or left at an intersection. Furthermore, the physical structure of the considered road junction influences the turn-dependent travel times, too. When turning left (in case of right-hand driving) vehicles normally have to give way to oncoming traffic first. And, particularly in urban areas, they sometimes have to wait for pedestrians crossing their way when turning left or right. The assumptions indicate that considering turn-dependent delay times will result in a much more accurate navigation and routing performance.
This paper describes a new approach how to automatically compute the required turn-dependent delay times based on common floating car data. Here, a number of probe vehicles provide information about their current positions every few (typically about 30) seconds so that travel times can be obtained for each pair of data points received from the same car. Filtering for the different turning directions at a given (signalized) intersection finally allows computing travel times for each traffic stream separately.
The challenge is that these travel times even in case of the same turning direction are related to very different driving distances as well as they are not link-based in general. However, the decomposition to obtain so-called link travel times or delay times is not trivial as has been shown in the literature. Accordingly, this paper presents a new idea to solve this task by introducing a simple linear model for the measured travel times which arises from the superposition of turn-dependent delay times and a basic travel time depending on driving distance and average flow resistance on the relevant road sections. The unknown parameters are then computed by (weighted) linear regression.
As a main result, the approach yields plausible estimates for turn-dependent delay times as is demonstrated for several signalized intersections in the German city of Hamburg where data of more than 1500 vehicles are available to the authors for several years. The results confirm the nave assumptions mentioned above and quantify the effects. Furthermore, the results indicate that the method may also give hints to the designers of traffic signal plans where in a city and how to improve the signalling.