Estimation and Simulation Gap Acceptance Behaviour at Congested Roundabouts
Noel Kay, Sonal Ahuja, Tan Na Cheng, Tom van Vuren, Mott MacDonald, UK
In congested networks, drivers may choose to behave more aggressively. We present a methodology for developing locally appropriate gap acceptance parameters based on local video observations which have a bearing on roundabout design and capacity.
Gap acceptance parameters form a crucial part of many simulation and operations models. VISSIM, SATURN, SIDRA, PARAMICS and many other models recommend default values, but allow the modeller to change these to more relevant values. When a traffic network is congested, drivers may choose to behave more aggressively, and accept smaller gaps than is modelled by standard parameters. Existing conditions may be hard to reproduce unless more aggressive parameters are adopted.
In this paper, we present a methodology for developing locally appropriate gap acceptance parameters based on local video observations. Using CCTV video recordings of driver behaviour around a congested roundabout in Birmingham, gap acceptance behaviour of vehicles has been analysed, resulting in appropriate input values for representation in VISSIM.
The video data was used to gather lengths (time gaps) for both accepted and rejected gaps. Accepted gaps are where a car on the approach leg deems the gap between vehicles of sufficient length (time) to allow it to safely move into the circulating stream. A rejected gap is where the car chooses not to move into the circulating stream as the gap is insufficient. The driver behaviour for heavy (PSV and HGV) and light (cars and LGV) was classified and analysed separately.
Once the data was collected and collated, curves representing the frequency of rejection and acceptance gaps where developed. Three methods of analysis were investigated. These include
? the accepted Raff?s critical gap,
? An ?equal? overlapping area gap and; and
? 50th Percentile of rejected gap curve.
In the Rafts critical gap method, the point where a driver is equally likely to accept as to reject a gap is considered the critical gap. In practice, this is the point where the rejection and acceptance curves cross.
The ?equal? overlapping area gap is also based on the relationship between the rejection and acceptance curves. In this method, the overlapping regions of the curves are examined. The critical gap is the point where the area under the ?rejects? curve equals the area under the ?accepts?. In practice, all gaps outside the critical gap are rejected. Some gaps are wrongly accepted, and some are wrongly rejected. The equal overlapping area critical gap is the one where the number of wrongly rejected gaps equals the number of wrongly accepted ones. In comparison, the standard Raff?s critical gap lends itself to either wrongly accept more than it rejects or vice versa, unless the Reject and Accept curves happen to be symmetrical.
In the third method, we look at the gap representing the 50th percentile of rejection data. Congested conditions mean there is an abundance of rejected gap data. A vehicle can only have one accepted gap, but any number rejected gaps. Graphing the rejected gaps, a ?negative exponential like? curve is generated. This curve resembles the traditional headway distribution curve used to calculate gap probabilities.
The problem with the first two methods is the relative infrequency of accept data, especially higher acceptance values. The acceptance curve tends to look linear, almost horizontal, when it should appear as an ?exponential like? curve. This makes the reliability of values calculated using the accept curve doubtful.
The study concludes that the 50th percentile reject curve method provides more reliable values, all values which are below the standard gap acceptance values. Further analysis of driver behaviour by disaggregating the observations by vehicle type reveals that gaps accepted by heavy vehicles differ from light vehicles. Accepted and rejected gaps are dependent on the traffic composition of conflicting / opposing streams, with higher time gaps required if the conflicting stream contains heavy vehicles. In congested situations heavy vehicles exhibit rather surprisingly more aggressive behaviour on the minor arm by accepting smaller gaps.
This gap acceptance study reveals behaviour that is contrary to expectation. It is found that heavy vehicles accept a much lower gap (ranging from 1.79 seconds to 2.28 seconds) compared to both the recommended value and the minimum observed value for cars/LGVs.
The observed minimum acceptable time gap for light vehicles ranges from 2.05 seconds to 2.72 seconds for different arms compared to the default recommended value of 2.6 seconds.
Congested traffic conditions also exhibit a state of ?mutual conflict resolution? between the minor and conflicting approaches where the vehicles on the main approach may give way to allow the minor stream vehicles to enter the intersection, where the accepted gap is not just a function of time but of free space available. This needs to be reflected in the gap parameter values.
Sensitivity testing of the simulation model reveals that the gap acceptance behaviour is function of traffic composition affects the capacity and performance of the roundabouts and has an impact on the design of roundabouts. The calculated gap acceptance parameters were validated in a microsimulation model and compared with observed travel times, showing a better fit than the default values. The paper will present the results.
Association for European Transport