Sunday 9 March 2014

Traffic Signal Pie: Second Slice - SCOOT Shenanigans

ok, perhaps a small slice this week, but I was intrigued to learn that Transport for London is about to conduct live trials of a new kit to detect pedestrians at junctions which will work with scoot.

My idea for a series of (irregular) posts on traffic signals has been disrupted somewhat by the announcement, but what they hey. Here is a bit on how signals at junctions work and what SCOOT can do to help or indeed hinder.


Imagine a cross roads such as the one to the left which is a relatively simple layout. The job of the signals engineer is to give every movement some time with the skill being how to do so with the actual flows through the junction. I am not a signals engineer and I do not profess to be an expert by any means!

My little example would probably be set up with the north-south arms running together (each road approaching the junction is called an arm), the east-west arms running together and then pedestrians cross all four arms at the same time while traffic is held.

This is the basic "Method of Control" for a junction and all signal controlled arrangements can be unpicked to show who goes when. This is all very good when traffic flows are balanced and pedestrians are happy to wait for a green man. 

My simple example can be shown on a "Staging Diagram" which graphically represents the stages the junction runs through. I have omitted some of the detail from this diagram, but you can see 4 stages. 1 and 2 are the two pairs of arms running together, 3 is the green man for all four arms and 4 is what is known as "all red". In other words, pedestrians and drivers will all be seeing a red signal and this like a safety buffer. 

In general terms, we would be looking at a 90 second cycle time, perhaps up to 120 seconds. Green given to drivers could be simply fixed, or perhaps fixed, but different at different times of the day. Pedestrians press a button to initiate "demand" and so when it is the turn of stage 3, they would get a green man. When setting up signals, actual traffic flows are needed to create appropriate timings for each stage. To get a bit cleverer, we can have vehicle detection ranging from simple microwave vehicle detectors (MVDs) which are the little camera-type units on top of the signal heads, to loops cut into the road surface.

This detection can simply "see" if a vehicle is approaching and initiate a demand. For example, with my example, if stage there are no vehicles detected on stage 2 and no pedestrians are pushing the button, stage 1 can just be left to run on green (sometimes known as "rest on green") - a good example of this will be on a trunk road which stays green until and unless a vehicle is detected in a side road or a pedestrian pushes a button. Loops can do the same job, but also build in cleverer speed detection which can tweak timings to deal with people who may be going a little fast. They are routinely used where speeds are a little high (35mph+ in a 30 limit for example) or on a 40mph or higher speed limit.

SCOOT, or Split Cycle Offset Optimisation Technique takes detection to a new level. A traditional set up may become inefficient over time if traffic patterns change and so will create a need for costly traffic surveys to recalibrate the signal controller. SCOOT uses detection loops on the approaches and exits of the junction and can count vehicles. It can also detect the speed of traffic which can be used to predict queues. The system can link multiple junctions to a central computer which can then look at the network as a whole. While a driver may be stuck in traffic at one location at one stage, the network as a whole will be operating efficiently.

As SCOOT is linked up to a computer system, we can be even cleverer. By using selective detection, extra green time can be given to an arm to help buses get through, but only if the bus has the correct transponder. It could even be linked to a bus GPS. This can be useful if a bus is running late and needs to catch up. There is much more on the SCOOT website.

If the traffic demand exceeds the junction or network capacity, the smooth flows will become disrupted and congestion created - SCOOT cannot create new capacity, it can only squeeze it out of an existing and fixed physical layout. The trouble is, if pedestrians don't have a green man, then SCOOT can run so well, that opportunities to cross can become more difficult to come by. I have a junction on my journey to work which doesn't have green men/ bikes on the side roads (it's a shared-use cycle track) and sometimes, there is no decent gap between traffic stages which makes crossing more difficult.

The other interesting innovation which is certainly coming to the roads of London (if not already here) is the use of wireless magnetometers. These are buried in the road surface and replace the detection loops mentioned above. Magnetometers are more reliable than loops and don't need all of the ducting and draw pits needed for loops (which cost a fortune in digging the place up to install). They are pretty sophisticated and can count axles which means that it is possible to measure the types of vehicle going through a junction.

Oh look, a bicycle-detecting magnetometer!
Image from Clearview Traffic.
As they are so much less costly to install than loops and don't come with all the digging, it makes retrofitting them to existing junctions a doddle. They do have a drawback in that they have on-board batteries which need replacing every few years, but this is a minimal problem compared to wider signal maintenance anyway. So, using the feedback of what is actually happening on the street, SCOOT does away with the need for manual traffic counts. It is linked to a hub computer system which makes changes relatively easy to model as the data is real time.

So, what has this to do with pedestrians and indeed cyclists? Well, where push button demand is concerned, whether you press the button once or ten times, or there is one person or 10 people waiting to cross, the signal computer only registers one demand when the button is pushed and it has no idea how many people want to cross. The Mayor is obsessed by "smoothing traffic flow" which does often seem to concentrated on motorised traffic and indeed, the speed at which SCOOT is being rolled out across London does seem very motor-oriented.

The Traffic Management Act 2004 "Network Management Duty" does also implies the same at face value, but this does mean users and that includes pedestrians and cyclists. The new pedestrian detection is exciting as it will be able to estimate the number of people wanting to cross. It will be using cameras with image recognition to pick out pedestrians - the technology already being used in airports and shops to track people and so is well established. Of course, there is no reason why the software couldn't be used to track cyclists through junctions as well.

The two trial sites will be outside Balham and Tooting Bec tube stations which are both on TfL's network and both have diagonal crossing elements. In short, very busy for pedestrians. TfL is also going to test "call cancel" where the demand to cross is deleted if nobody cross or they cross before the green man - although this was a feature of Puffin crossings already. I don't know, but I am assuming that demand will still be called by a push of the button rather than being an automatic demand (which traffic generates). Of course, vehicles follow clear paths and so it is an awful lot easier to detect them than pedestrians milling about.

It is possible to give a green man pretty quickly after the button is pushed, but unless it is just before the pedestrian stage, people still need to wait until their turn in the cycle. It is possible to juggle stages on the fly and have extra pedestrian (or traffic stages). So, in my example, we could have north-south, pedestrians, east-west, pedestrians and so on. This will be at the expense of other arms' green time if we are to be within our 90 to 120 second range.

Pedestrian SCOOT certainly enables a good argument to be put for much higher pedestrian priority and where we get people walking outstripping people driving through a junction, surely pedestrians must be given the dominant green time; plus with the ability for cycle detection, we certainly have the kit available. So, let's see what happens and remember that the technology doesn't decide who gets what from the traffic signal pie; that is the political consideration.

No comments:

Post a Comment