Estimation and Analysis of Multilane Multiclass Headway Distribution Model
HOOGENDORN S P and BOVY P H L, Delft University of Technology, The Netherlands
The availability of accurate estimates for the headway distribution for specific lanes and classes, during diverse periods of the day is of dominant importance for several applications in ITS. For instance, these estimates can be used for realistic multic
The availability of accurate estimates for the headway distribution for specific lanes and classes, during diverse periods of the day is of dominant importance for several applications in ITS. For instance, these estimates can be used for realistic multiclass multilane vehicle generation in microscopic traffic simulation tools, on-line lane- specific capacity and person-car-equivalents estimation, and their consequent use in ITS applications, infrastructural design and planning., safety analysis, and gap- acceptance behaviour.
A large nurnber of models for the description of time-headway distributions have been proposed. This paper adopts and extends the Generalised Queuing Model (GQM). Re- cently, a technique for the estimation of parameters of the GQM was developed (see HOOGENDOORN AND BOTMA (1996) and HOOGENOOORN et al. (1997)). This new sta- tistical procedure is characterised by improved estimation results (in a Kolmogorov- Smimov sense), more efficient CPU time, and improved robustness with respect to parameter estimates.
Until now, the parameter estimation procedure has only been used on two-lane rural roads. In this paper, the tehniques is applied to motorway headway observations sub- divided according to lanes. Distinction of lanes, vehicle-types, leader-types, and sam- ple periods provides insight into the plausibility of the headway distributions and pa- rameter values, as well as into the lane-specific car-following behaviour of the distinct vehicle-classes varying across the different periods. By applying the new estimation procedure to motorway traffic data, the estimated headway distributions are compared with real-life data. We show that headway distributions can be accurately described using the proposed model. Significant differences in estimates between different peri- ods, roadway lanes, vehicle-types, and leader-types are found. These differences can be interpreted from different viewpoints, such as vehicle characteristics, travel pur- pose, time of day, and differences in car-following behaviour during congested and non-congested traffic conditions.
Association for European Transport