When a bay is added and it is dedicated to the right-turn movement, the extra space is used to store vehicles that want to make that manoeuvre. When the right-turners (RTs) are opposed, the extra space normally accommodates some vehicles out of the way of the other traffic on the same approach who can make their manoeuvre unimpeded as a result for at least part of the cycle (ie until the right-turners queue upstream of the start of their bay and stop the other straight-on traffic).
Firstly, consider the case where there is no early cut-off or late-release stage in the cycle for the RTs. Normally, the recommended method of modelling this situation depends on whether the RT traffic blocks back at all or not; even so the suggested methods are not ideal. They are:
a) If the RT traffic does not block back, use two separate links: the disadvantage is that TRANSYT models two separate queues whereas in reality you have one-and-a-bit queues. This means that the saturation flow of the straight-ahead lane has to be reduced in proportion to the number of RTs in the flow upstream of the bay. It also means that the RT ‘OUT’ pattern will be unrealistic in shape, but this probably does not matter too much as the capacity is modelled with reasonable accuracy.
b) If the RT traffic blocks back and restricts the straight-ahead traffic, it is better to model this approach as a single link along the lines described in the first article in TSN No 20. The problem here, however, is that it is necessary to compensate for the extra capacity afforded by the fact that some straight-ahead traffic will be able to make their manoeuvre freely before the RTs block back. To do this the saturation flow needs to be increased so that it is a time-weighted average of the value before and after blocking occurs, averaged over the relevant green time.
It might be thought that this ‘step’ in the saturation flow can be modelled with the flared approach parameters. It can’t! Flared approach data is ignored whenever a link is specified as a give-way (whether or not it is also a signal controlled link).
Introducing a new method…
Whilst writing this article, I tried something that I have not used before which may be of significant value:
If you split the approach with the RT bay into two short links, and then have a single feeding link to model the queue upstream of the start of the bay (see figure 2) it is possible to, at the very least, find out accurately how many vehicles are likely to use the RT bay. The principle of this method is to control the OUT pattern of link 10 by suitable start and end-lag adjustments. Firstly, make all the links belong to the same node (you can still feed flow from link 10 to links 11 and 12). Then consider the following
- Normal start lag (integreen), I
- Number of vehicles (pcus) that can physically store on link 11
- Time for maximum queue on link 11 to discharge, t
- Journey time along link 11, j
Then the start lag for link 10 = I + t – j.
The end lag needs to be adjusted until the queue predicted in the TRANSYT output is just enough to fill the space on link 11 and no more. Note that the end-lag value will remain the same from one run to the next for uniform-flow entry links, but may differ markedly for internal links depending on the OUT patterns from feeding links.
With all the data in place, the maximum expected queue for link 12 is shown in the output and you can determine whether blocking back will occur or not from it. If there is no blocking back, the resulting model will be an accurate representation of real-life. If there is blocking back, experimentation with the model by, for example, using the limit-queue facility and trying early-cut-off and late-release stages, will allow you to see if blocking back can be avoided. It may also be useful for modelling situation where the RT movement is catered for in a different stage.
If blocking back is unavoidable, method b) above would be the traditional method of modelling the situation. However, it might be possible to extend the new method in some way to take account of the change in capacity before and after blocking occurs. Firstly you need to work out the point in the cycle where the blocking occurs. Then you could specify link 10 as a flare with parameters that give a suitably ‘high’ saturation flow before blocking occurs, and a low one after. This isn’t an easy method as working out the low saturation flow value is not trivial, and the point in the cycle where blocking occurs will be different every time new timings are used. Nevertheless, it may offer a useful alternative if applied very carefully.
In conclusion, right-turn bays present particular modelling problems and I have presented the traditional methods of modelling them, plus a new method which does at least give an accurate figure for the size of the RT queue. I must make a small disclaimer for the new method though: myself and a colleague have looked over the method carefully and believe it does as explained, provided it is used in the way described. However, we have not presented the method in any way as a de-facto standard – you need to justify its use to your own and to clients’ satisfaction.
In the next part I will deal with approaches with two or more lanes.