SIMULATING TWO-WHEELERS IN CENTRAL LONDON
Assessment of the impact of cyclists on heterogeneous traffic David Carrignon, Colin Buchanan Heterogeneous traffic conditions are normally associated with some developing countries but they are becoming more common in Central London
because of the rise in the number of two-wheelers. This paper describes a first attempt to model twowheelers in London using VISSIM. While more
MODELLING TWO-WHEELERS IN VISSIM Central London traffic conditions are changing! In recent years, pedal cycles and powered two-wheelers have been encouraged as a convenient alternative to overcrowded public transport and to traffic congestion, and it has worked. Today, engineers doing traffic models in the square mile and the surrounding areas are faced with a new challenge because the success of this alternative form of transport has generated heterogeneous traffic conditions. Traditional traffic engineering theory looking into the impact of cyclists and motorcyclists usually assumes a low or a very low proportion of two-wheelers and homogeneous traffic conditions. In the United Kingdom, the Transport Research Laboratory (TRL) published in 1986 a study titled ‘the prediction of saturation flows for road junctions controlled by traffic signals’. This study, more commonly known as RR67, sets the standard in terms of saturation flow calculation in the UK. PCU values were calculated by the method of headway ratios on a lanespecific basis. It is therefore estimated by dividing the average headway associated with a given vehicle type by that for light vehicles. In this key document, motorcycles were assumed to respect lane discipline and were given a PCU value of 0.4, www.tecmagazine.com
calibration is needed, it shows that cyclists and powered two-wheelers could become two new stakeholders in the traffic modelling process.
while pedal cycles would be expected to use lateral gaps in traffic and were given a PCU value of 0.2. In a context where powered two-wheelers would not respect lane discipline, their PCU value should therefore be expected to be closer to 0.2.
HETEROGENEOUS TRAFFIC Heterogeneous traffic conditions combine significant vehicle performance discrepancies and a lack of lane discipline. In most places in London, the volume of cyclists and motorcyclists is not sufficient to impact traffic conditions and motorists tend to respect lane discipline quite well. In some places of Central London however, this is no longer the case. It is not that cars, buses and lorries have changed their driving habits, but the rise in twowheelers is changing traffic conditions. Heterogeneous traffic conditions are usually perceived as frustrating for motorists, unpleasant for vulnerable users and generally unsafe. The discrepancy in vehicle performance leads drivers to be frustrated by vehicles with lower performance and the usage of lateral gaps significantly increases the amount of information a driver has to deal with. Motorcycles are usually at the high end of vehicle performance whilst pedal cycles are at the low
Author’s details: David Carrignon is Principal traffic engineer at Colin Buchanan and can be contacted by email at [email protected]
TEC JULY 2009
SIMULATING TWO-WHEELERS IN CENTRAL LONDON
end. Both vehicle types have been monitored and observed not to be respecting lane discipline. In many countries with large volumes of cyclists, the solution of mandatory cycle lanes on the carriageway was adopted. If you compare two segregated lanes, one for cyclists only and one for motorised vehicles only these two roads would have a perceived safer environment and a higher capacity than the same two lanes with heterogeneous traffic conditions. This example however assumes for fast moving motorised traffic, comparatively slow moving cyclists, and ignores junction capacity issues. In Central London, motorised vehicles are relatively slow, cyclists are surprisingly fast and space for such segregated infrastructure is scarce.
PARLIAMENT SQUARE Colin Buchanan was commissioned by Transport for London (TfL) to undertake traffic models as part of a regeneration project of Parliament Square in Central London. A set of traffic models were developed including one using VISSIM micro-simulation software. Following this commission, the London Cycling Centre of Excellence requested further research using this model to assess the impact of cyclists on the general traffic in Central London. In most cases, the Directorate of Transport Operations (DTO), part of TfL, advises excluding two-wheelers from micro-simulation models. For this project however, pedal cycles and motorcycles had to be represented while remaining fully compliant with DTO guidelines. The DTO modelling guidelines version 2 includes requirements for micro-simulation models. These requirements are mainly on junction turning counts, car journey times and saturation flow. In this case turning counts and car journey time did not require a change in modelling habits, but the saturation flow did. Unaware of where the project would lead, the inception meeting threw up an interesting debate on the issue of two-wheelers. One side was adamant they would have no impact, as it is what the theory tells us. The other side was unconvinced. This initial discussion led to a decision which proved invaluable. Site visits recorded the saturation flow omitting two-wheelers, as if they would be invisible. Classified counts were available, Vehicular Saturation Flow (VSF), so two-wheelers could be assimilated into an environmental constraint! As shown later, this decision was the key to the success in modelling twowheelers.
VEHICULAR SATURATION FLOW (VSF) Survey method
Figure 1: Overtaking issue
The Vehicular Saturation Flow site measurement is a stan-
Other large vehicle
TEC JULY 2009
dard saturation flow measurement at a signalised stopline at a saturated approach. As mentioned above, the main difference is the omission of two-wheelers. This of course reduces the saturation flow recorded when there is large concentration of two-wheelers.
VISSIM measurement options Once the VSF values were recorded, a method had to be developed to replicate this measurement in VISSIM. Headway values between vehicles at a stopline can be obtained from the software using a special evaluation file. This file is usually the basis of the saturation flow calibration in VISSIM in homogeneous traffic conditions. Empirical results shows that a headway of 2.5 seconds is a good indicator for the end of saturation and that a measurement should only be recorded from the third vehicle passing the stopline. Because two-wheelers are using lateral gaps, the headway measurement output file became unusable as we could not distinguish cyclists and motorcyclists from other vehicles. The development of an Excel VBA programme enabled us to extract the following information: • The headway of each vehicle crossing a stopline during the green time • The PCU value of the vehicle attached to a particular headway With the PCU value and the headway, it becomes possible to derive the VSF. The difficulty consists in identifying the end of saturation, as 2.5 seconds was not suitable anymore as it would lead to too few records. Nine cases have therefore been detailed to identify the end of saturation. • Case 1 – A Car or an LGV preceded by a Car or an LGV The headway has to be less than 2.5s • Case 2 – A Car or an LGV preceded by an HGV or a Bus The headway has to be less than 5s • Case 3 – A Car or an LGV preceded by a Bicycle or a Motorbike The headway has to be less than 2s • Case 4 – An HGV or a Bus preceded by a Car or an LGV The headway has to be less than 5s • Case 5 – An HGV or a Bus preceded by a HGV or a Bus The headway has to be less than 5s • Case 6 – An HGV or a Bus preceded by a Bicycle or a Motorbike The headway has to be less than 2s • Case 7 – A Bicycle or a Motorbike car preceded by a Car or an LGV The headway has to be less than 2.5s • Case 8 – A Bicycle or a Motorbike car preceded by a HGV or a Bus The headway has to be less than 5s • Case 9 – A Bicycle or a Motorbike car preceded by a Bicycle or a Motorbike The headway has to be less than 1s In the case of more than five consecutive two-wheelers, the recording of data will suspend until the beginning of the following green. The ending of the recording process ensures that the recording remains in relatively standard conditions. This VSF measurement tool enabled us to calibrate the driving behaviour.
DRIVING BEHAVIOUR TfL DTO very usefully keeps an up-to-date VISSIM template which includes the most suitable parameters for www.LocalGov.co.uk
SIMULATING TWO-WHEELERS IN CENTRAL LONDON
London. This template is the usual starting point and this model was no exception. In order to allow two-wheelers to overtake and be overtaken within the same lane, two new link types were created; one for single lane roads and one for multiple lane roads. These link types helped reproduce the usage of lateral gaps and produced acceptable driving behaviour at this stage. The difficulty with this setup is that the lane width is becoming an important factor which meant that the geometry of the links in VISSIM are having an influence on driving behaviour. This was new, and had to be calibrated. Through trial and error, the lateral distance parameter was eventually calibrated which determines if a twowheeler decides to use a gap or not. Although very good results were being achieved, this new method affected the stability of the results.
Position on street (cyclists tend to go closer to the kerb to overtake other vehicles on the left).
Figure 2: Cycle position in relation to the kerb
Position in VISSIM (the edge of the vehicle plus the appropriate lateral distance)
STABILITY OF THE RESULTS The difficulty with the stability of the saturation flow measurement with the proposed method is that it depends on the traffic composition. The PCU factor is here to compensate the difference, but it is best to rely on a large sample. To calculate the VSF with an acceptable level of accuracy, the sample recorded to undertake the calculation should reflect the traffic composition as much as possible. Overall, it was found that at least 30 headways have to be recorded in saturated conditions for the measurement to become reliable.
CURRENT LIMITATIONS OF THE DRIVING BEHAVIOUR Although the outputs of the model we developed were good, the visual aspect and driving behaviour flexibility of two-wheelers still needs to be improved. When a vehicle wants to do a ‘necessary lane change ‘, a lane change due to a change in direction, this vehicle will have a limited number of lanes to choose from. Once in the appropriate lane, it remains in it. The type of issue faced then corresponds to a situation where left turning vehicles cannot use the space available on the inside lane to overtake the two-wheeler. In reality, it is very likely that the motorised vehicle would overtake. Another limitation is the lateral distance of two-wheelers against the kerb. The same parameter is used to identify gaps between vehicles and between a vehicle and a kerb. Many two-wheelers, cyclists overtaking a queue of traffic at a stopline in particular, would have their wheels very close to the kerb. VISSIM is also currently unable to let a proportion of vehicles jump a light. This can be an issue in an environment where a significant proportion of the cyclists do not respect traffic lights. A special vehicle type could be setup for this purpose but the light jumping behaviour is expected to be site specific and this solution is therefore not appropriate. The last major limitation of the software we encountered is linked to the calculation power available. For the moment, a vehicle can only look up to ten vehicles ahead to anticipate conflicts. With large number of two-wheelers on the network, it happens that some vehicles collide with others because they cannot anticipate enough. These limitations appeared as the model was developed and they are very likely to be overcome in the future.
STRATIFIED SATURATION HYPOTHESIS One of the objectives of the further study commissioned by the London Cycling Centre of Excellence was to assess the impact of lane width on the PCU value of cyclists using VISSIM. While doing the testing, it was discovered that there were issues with the way saturation was included. So far, the VSF saturation flow was calculated without looking at two-wheelers. When attempting to calculate the PCU value of cyclists, the stratified saturation hypothesis had to be formulated. This hypothesis states that vehicles respecting lane discipline do not necessarily achieve the same level of saturation as vehicles not respecting lane discipline. For example, in traffic flow mainly composed of cars with a wide lane and a couple of two-wheelers, cars might be at saturation, but two-wheelers might not be. They are filling the gaps in the traffic and they still have some space for more cyclists. The calculation showed that the two-wheelers’ traffic reaches saturation at 10% of traffic proportion. This hypothesis enabled the reproduction of the statement present in the existing literature that in heterogeneous traffic conditions, the PCU value of a vehicle type is variable. The table below is only a preliminary result from the model and should be re-assessed against measured values on street, but it shows the average cyclist PCU value per lane width.
CONCLUSION This work was a very successful first attempt to model two-wheelers in Central London. It shows that more calibration is needed, but it also shows that cyclists and powered two-wheelers could become two new stakeholders being included in the traffic modelling process.
Table 3: Cyclists average PCU analysis TEC JULY 2009