Signalized Merge

[Peter Freeman]

^ Yes, I realise that. I do understand the Switch Island history. I didn't literally mean that it should have been built as it is now but with an M58/A59 braid. It should have been built as a grade-separated subset of what the grade-separated final configuration would be. That would not have included the signalised merge.

(Not that I'm at all impressed by the intended final configuration shown in your Maghull Interchange link. But I digress ...)

[Peter Freeman]

The other spot in the area that I believe could make use of either a signalised merge, or simple ramp metering is the exits for Coast Road off Central Motorway. Having even just a ramp meter on the northbound exit would be helpful in reducing the more dangerous queues that build on the southbound carriageway.

I see the spot you're referring to: the approach (2+2=3) to Osborne Road crossroads, with bad congestion at both am and pm peak. There's a lot going on there! A 'ramp meter' on only one of the inputs would not do as good a job as a signalised merge. By signalising both inputs, with 2-phase turn-taking, there could be a degree of control over the relative lengths of nb and sb queues.

BTW, I assume that by 'simple ramp meter' you mean a UK-style platoon-release meter. Such a meter might marginally help here, but this situation is not one that ramp metering can solve.

Ay, if a signalised merge was to be implemented here, it would need to tie into a scheme to upgrade the Coast/Osbourne Junction.

Perhaps not "need" to. Changes to the merge might help, even without your suggested upgrade to the Osborne intersection.

The congestion is worst at PM peak, and (as you say) the worst aspect is queuing back onto Central Motorway (CM) southbound. An important cause is the road marking at the merge. Traffic coming from CM southbound isn't treated equally: its right lane has to cross a broken line to merge. On the other hand, for traffic coming from CM northbound, there is continuous priority up to the Osborne intersection. This is (partly) what causes the southbound queue to be longer than the northbound one. It is possible that simply erasing the broken line, so that neither stream has priority into the Coast Road middle lane, would reduce the southbound queue. Alternatively, and more aggressively, reverse the priority, so that traffic from northbound must cross a broken line.

The second probable problem is that some vehicles arriving at the intersection have not managed, in the congested short space available, to get into their desired lane. A last-minute scramble for position, just before the signals, reduces the efficiency of flow through the intersection (for all concerned, not only the turners). The solution to this is to signalise the merge. Then, each time that vehicles get a green light and flow onto the Coast Road eastbound, they can immediately select their proper lane without competition.

Introducing a signalised merge solved exactly this problem at an Australian location well-known to me (close to my home). Further information is in my post on 30 May 2022 on this page -
viewtopic.php?p=1242371&hilit=ferntree+ ... d#p1242371

Any of these actions would be low cost, compared with major intersection works.

[Peter Freeman]

A414 off M1 J8 (https://maps.app.goo.gl/nXsxEixzGbiucsax6)

... I'm interested in how the signals operate at the following roundabout if anyone local knows? There are only signals on the A414 East and Green Lane South entry arms and not on the circulatory, presumably these two just alternate for green time?

I'm not local, but my guess would be that traffic approaching from the west will be measured, and if it builds up too much (from traffic turning to/from Green Lane northbound), they would both go red to allow that traffic to clear. I'm interested to know what they actually do, though.

These look like 'approach metering' signals. They work as Keiji suggests (there's a few variations on the theme). Walk up the non-signalised arms and you'll see the asphalt cuts where the detector loops are embedded.

There are several (at least) of these in Australia, though not as many now as previously. I can think of only one set left in Melbourne. They are effective, in the narrow range of problems where they're applicable. In AU they've generally been 'band-aids' (lasting a few years!) until the site comes up for roundabout deletion.

[jackal]

So when you're talking about a bridge in the Danish case you mean a hypothetical bridge for braiding?

Yes. In all cases of a signalised merge 'saving' the cost of a bridge, the bridge would have been a 'braid' in the pure sense of eliminating a 'weave'.

That's such a far-fetched suggestion that I don't see how it could be relevant.

Far-fetched in the case of that E20 exit, yes, but not far-fetched at all in the case of Brisbane's M2. That's why I was surprised at first to find it. Switch Island: yes, most unlikely now, but could/should have been done at construction time. They are British 1970's motorways after all.

The freeflow arrangement would simply remove the lights and change markings. There is ample weaving space for a relatively low speed road.

At the E20 exit: yes. At Brisbane's M2: no - far too much weaving. At Switch Island: totally unworkable - two 3-lane carriageways converge then split to 7 lanes in <100m.

Switch Island - I don't see how you're imagining braiding would fit in. You need hundreds of metres between each merge and diverge, whereas we're talking about a space of, as you say, <100m. It looks to me like space is a decisive reason for the signalised merge, i.e., type 1A. The alternative would be a completely different design of the complex, as Chris mention, rather than tagging on braiding.

Brisbane - this should be braided. There is braiding for a presumably less busy junction to the south.

That said, what I was denying was that a signalised merge saved money on the width of the road and associated structures, i.e., I thought you were saying the existing E20 spur bridges over Selinevej and under the railway could be narrower with a signalised merge rather than a freeflow lane gain. I was not meaning to deny that a signalised merge is cheaper than braiding, nor that that might sometimes justify the signalised merge (even if we disagree on the application to these specific cases).

So yes, in the case where we are choosing between extreme alternatives - signalised merge or braiding - 1B is a possibility. I hadn't really thought about that as it seemed an unusual choice, like choosing between a flat roundabout and a stack. But some cases are like this, e.g., Brisbane. So thanks for the clarification.

[Peter Freeman]

Switch Island - I don't see how you're imagining braiding would fit in. You need hundreds of metres between each merge and diverge, whereas we're talking about a space of, as you say, <100m. It looks to me like space is a decisive reason for the signalised merge, i.e., type 1A. The alternative would be a completely different design of the complex, as Chris mention, rather than tagging on braiding.

OK, special case here, which made it difficult to appreciate as part of my argument.

I was not suggesting "tagging-on" braiding at Switch Island's signalised merge now. I was not even suggesting doing simply that at the original build time. I was trying to say that, at original build time, a decision was taken to have (virtually) no freeflow. That forced the adoption of the signalised merge. The arrangement downstream of the signalised merge is not simply one diverge that is too close (though just one is sufficient justification), there are multiple signalisations in even closer proximity downstream of that one (!).

Taken as a whole, the interchange could have originally been built properly as full freeflow, but instead it was decided to signalise it all, thus forcing the signalised merge. In that way, it is like a Type 1B: a deliberate decision not to optimally exploit the vast canvas available. Reason (I assume): to save money.

In retrospect, I should not have used this example - it muddied the water. However, Keiji proposed it as a signalised merge example, and, taking into account the current dismal status quo, it is indeed a very good example that works.

[Peter Freeman]

Here's a planned interchange upgrade in Indiana that has a nice example of a signalised merge. It's for the correct and justifiable reason: another signalised intersection, with multi-directional choice, occurs very soon downstream. The absence of competition between the two inputs will allow early and orderly lane selection, laminar flow, and higher throughput.
See at 1m21s in this video:
https://www.youtube.com/watch?v=u1luViuWqZE

This is exactly the same situation as in Melbourne's first and very successful signalised merge that I reported on, here:
viewtopic.php?p=1242371&hilit=ferntree+ ... d#p1242371
The location is here:
https://www.google.com.au/maps/@-37.902 ... ?entry=ttu
In that case, an M1 off-ramp meets Ferntree Gully Road - originally at a 3+1=3 merge, now at a 3+3=3 signalised merge.

[Owain]

Probably mentioned before but here's a signalised merge on the Leeds IRR. Have to show an old date as more recent ones are affected by roadworks.

https://www.google.co.uk/maps/@53.79890 ... ?entry=ttu

Yes. A classic. Looks like it's always been like that, so a brave decision by 'the Motorway City' back then.

More brought about by the scrapping of other plans - the westbound flyover before it was not original to the motorway and was done in the late 1980s if memory serves.

May even have been early 90s. Before then all westbound traffic on the A64 passed through the signalised farrago at Burmantofts Street.

The A643 is getting one here, as part of the works to redevelop the Armley gyratory. The stop lines have been painted on the carriageway, but I don't think the lights are there yet.

I'm hoping they'll do something about that bridge too - it seems to add miles to the journey of anybody wanting to cross the road!

[Peter Freeman]

Downstream of the Danish example, after the split, is a related curiosity: a signalised, narrow-angle, no-turns crossover -
https://www.google.com.au/maps/place/Co ... ?entry=ttu
Again, Nordic pragmatism reigns. A flyover (braid) would theoretically be better, but could the cost be justified? The location is not motorway, so no principles are being offended, and capacity is more than sufficient.

Such 'no-turn crossovers' occur in the UK too. One is on Heathrow Airport's Western Perimeter Road -
https://www.google.com.au/maps/place/Ho ... ?entry=ttu
I always notice it as I drive away in a hire-car.

Another one, this time involving three carriageways, is on the M60 at J26 -
https://www.google.com.au/maps/place/Ma ... ?entry=ttu
Again, no problems are evident from this arrangement.

Here's another one (USA) -
https://www.google.co.uk/maps/@42.44488 ... ?entry=ttu
And on the same off-ramp, just 500m north, there's another, completing an unusual left-turn short-cut -
https://www.google.co.uk/maps/@42.44661 ... ?entry=ttu

[Peter Freeman]

The A643 is getting one here, as part of the works to redevelop the Armley gyratory. The stop lines have been painted on the carriageway, but I don't think the lights are there yet.

That is a classic example of a merge requiring signalisation. All three of the converging northbound approaches are queued at am peak for long distances upstream - the A643, a capable road in itself, as far back as the M621. The merge is, currently, the jam-point. Slight subsequent relief just before the roundabout can be seen.

After this merge signalisation, if the gyratory is not satisfactorily modified, the jam point will simply move 250m downstream (north) to the gyratory entry signal. Unfortunately, what I see of the gyratory upgrade plan is next to useless: minor tinkerings only, which will make little difference to traffic flows or congestion (though small improvements for pedestrians and cyclists).

It's hard to believe that the Armley design failure has been tolerated for so many years, impairing the LIRR whose first (tunnel)stage was such a brilliant move in the late 1960s. Now the impediment will be prolonged, and baked in.

[Owain]

The A643 is getting one here, as part of the works to redevelop the Armley gyratory. The stop lines have been painted on the carriageway, but I don't think the lights are there yet.

That is a classic example of a merge requiring signalisation. All three of the converging northbound approaches are queued at am peak for long distances upstream - the A643, a capable road in itself, as far back as the M621. The merge is, currently, the jam-point. Slight subsequent relief just before the roundabout can be seen.

After this merge signalisation, if the gyratory is not satisfactorily modified, the jam point will simply move 250m downstream (north) to the gyratory entry signal. Unfortunately, what I see of the gyratory upgrade plan is next to useless: minor tinkerings only, which will make little difference to traffic flows or congestion (though small improvements for pedestrians and cyclists).
It's hard to believe that the Armley design failure has been tolerated for so many years, impairing the LIRR whose first (tunnel)stage was such a brilliant move in the late 1960s. Now the impediment will be prolonged, and baked in.

It turns out I was wrong about the lights not being in yet - I noticed them (still with their florescent covers on) while walking over the bridge the other day:

There will still be an awkward kink where the A643 passes under the railway bridge, but the gyratory is being enlarged on an epic scale!

[Peter Freeman]

... but the gyratory is being enlarged on an epic scale!

The widenings of approaches and the circulatory could hardly be more un-epic. Deleting the semi-hamburger cut-through proposal was a relief, but that change left little that's worthwhile.

Oh, silly me - I forgot the 'new trees' that will be planted (in the ridiculous, huge, 'private' island that continues to spoil any chance of building a proper intersection).

Not epic. Small-minded and disappointing.

[traffic-light-man]

It turns out I was wrong about the lights not being in yet - I noticed them (still with their florescent covers on) while walking over the bridge the other day:

The stop line looks strangely far back from the merge there. I wonder if it's to do with fitting everything in before the VRS starts for the bridge support?

[Owain]

It turns out I was wrong about the lights not being in yet - I noticed them (still with their florescent covers on) while walking over the bridge the other day:

The stop line looks strangely far back from the merge there. I wonder if it's to do with fitting everything in before the VRS starts for the bridge support?

I noticed that strange distance too. My first thought was that it was to allow for the inclusion of pedestrian/cycle crossings*, so that the bridge could be demolished, but I don't suppose they'd do that on a such a busy route.

*The provision for pedestrians/cyclists on the gyratory itself is being improved significantly.

[Peter Freeman]

It turns out I was wrong about the lights not being in yet - I noticed them (still with their florescent covers on) while walking over the bridge the other day:

The stop line looks strangely far back from the merge there. I wonder if it's to do with fitting everything in before the VRS starts for the bridge support?

Yes, those stop lines are too far back. Ideally, they should both be close to the actual merge, so that, when the signal turns red, the tail-end vehicles can clear past the merge-point before the other input gets its green.

To visualise the process, imagine the extreme case, where input-A's stop line is a distance upstream of the merge point that is equivalent to one whole release-platoon of vehicles. Then, the first vehicles released by input-B's green will immediately meet a solid queue of vehicles from A that haven't yet passed the merge point. This competition for space will last for the whole input-B green phase, with lane-selection chaos. Thus, the merge is not relieved at all.

For intermediate (non-extreme) cases, the further back the stop lines are, the less effective the signalised merge will be.

[traffic-light-man]

Yes, those stop lines are too far back. Ideally, they should both be close to the actual merge, so that, when the signal turns red, the tail-end vehicles can clear past the merge-point before the other input gets its green.

To visualise the process, imagine the extreme case, where input-A's stop line is a distance upstream of the merge point that is equivalent to one whole release-platoon of vehicles. Then, the first vehicles released by input-B's green will immediately meet a solid queue of vehicles from A that haven't yet passed the merge point. This competition for space will last for the whole input-B green phase, with lane-selection chaos. Thus, the merge is not relieved at all.

For intermediate (non-extreme) cases, the further back the stop lines are, the less effective the signalised merge will be.

Doing some frugal measurements, you're probably looking at a 5 second intergreen between the A58 and the A643, and a 6 second intergreen between the A643 and the A58 (not accounting for any high-speed increases), so there's only 1 second in it on one of the stage changes anyway.

Assuming there's somewhere for the traffic to go, I don't think that would be a problem - exit blocking caused by the gyratory itself would be the concern in my mind.

[jnty]

I think I had it in my head that large stop line-merge point distances could actually be more efficient. Surely giving traffic more space to speed up before the merge point means you can tighten up distances between alternate platoons, whereas if the line is right up at the merge point, you have to hold the other traffic sooner to allow the newly started traffic time to limp past the merge as it accelerates. I don't know if UK signal rules permit you to let traffic into the junction before conflicting traffic has fully cleared the conflict point, which is essentially what you'd have to do here to extract full advantage. But even if you can't, there's probably a sweet spot where you can minimise buffer time between the tail end of the last phase's traffic passing the conflict point and the next green.

[traffic-light-man]

I think I had it in my head that large stop line-merge point distances could actually be more efficient. Surely giving traffic more space to speed up before the merge point means you can tighten up distances between alternate platoons, whereas if the line is right up at the merge point, you have to hold the other traffic sooner to allow the newly started traffic time to limp past the merge as it accelerates. I don't know if UK signal rules permit you to let traffic into the junction before conflicting traffic has fully cleared the conflict point, which is essentially what you'd have to do here to extract full advantage. But even if you can't, there's probably a sweet spot where you can minimise buffer time between the tail end of the last phase's traffic passing the conflict point and the next green.

Guidance gives a minimum intergreen of 5s, which is the sum of both conflicting-phase ambers showing consecutively (i.e. without any 'all red' duration).

Concurrent ambers, where the two ambers show together for some duration are technically possible, though seldom seen in my experience. In fact, I think I've only ever actually noticed it in person once, which was on a site in Plymouth where it looked most bizarre, especially given it was a four arm cross-roads.

Having said that, I believe in the early days of signal control, fully concurrent ambers were quite common (including a 3s starting ambers), purely because some controllers back then wired the ambers across all phases to the same circuit, but those were lacking even by the '60s.

[Peter Freeman]

Yes, those stop lines are too far back. Ideally, they should both be close to the actual merge, so that, when the signal turns red, the tail-end vehicles can clear past the merge-point before the other input gets its green.

To visualise the process, imagine the extreme case, where input-A's stop line is a distance upstream of the merge point that is equivalent to one whole release-platoon of vehicles. Then, the first vehicles released by input-B's green will immediately meet a solid queue of vehicles from A that haven't yet passed the merge point. This competition for space will last for the whole input-B green phase, with lane-selection chaos. Thus, the merge is not relieved at all.

For intermediate (non-extreme) cases, the further back the stop lines are, the less effective the signalised merge will be.

Doing some frugal measurements, you're probably looking at a 5 second intergreen between the A58 and the A643, and a 6 second intergreen between the A643 and the A58 (not accounting for any high-speed increases), so there's only 1 second in it on one of the stage changes anyway.

Assuming there's somewhere for the traffic to go, I don't think that would be a problem - ...

The locations where signalised merge is beneficial are those where there isn't somewhere to go. Or, more precisely, where the back of the next queue is not far downstream of the merge point. The issue to address is not how much traffic the signalised merge can transmit. It's how can the merge's detrimental effect on the next downstream queue be mitigated. See earlier posts for explanation.

... exit blocking caused by the gyratory itself would be the concern in my mind.

If the road was UNobstructed downstream of this merge point, it would be better without signals. Simply (a) let them merge (2+1 into 2), perhaps with some difficulty, or (b) first condense each of the 2-lane carriageways to single-lane, so that a side-by-side, no-conflict merge occurs. In both cases, they can then sort out lane preferences in the clear space beyond.

However, that is not the case at Armley. The gyratory is causing a queue that reaches back to (or near enough) the merge point being discussed. It is a queue with a significant proportion of vehicles that wish to change lanes. Unfortunately, they are in the lane that the uncontrolled merge placed them in, and are trapped in a situation where lane changing is difficult. This has a detrimental effect on the flow towards, and at, the gyratory.

[Peter Freeman]

I think I had it in my head that large stop line-merge point distances could actually be more efficient. Surely giving traffic more space to speed up before the merge point means you can tighten up distances between alternate platoons, whereas if the line is right up at the merge point, you have to hold the other traffic sooner to allow the newly started traffic time to limp past the merge as it accelerates. I don't know if UK signal rules permit you to let traffic into the junction before conflicting traffic has fully cleared the conflict point, which is essentially what you'd have to do here to extract full advantage. But even if you can't, there's probably a sweet spot where you can minimise buffer time between the tail end of the last phase's traffic passing the conflict point and the next green.

You're also thinking of this issue as "How do I maximise flow through a signalised merge?". That's not the point here (though it might be interesting in itself, elsewhere). The objective of a signalised merge is (usually) to avoid causing a problem** in the lead-up to the NEXT traffic obstruction (which, in this instance, is the signal to enter the gyratory). If there is no next traffic obstruction, then a normal (UNsignalised) merge is preferable.

It is true that a signalised merge can perform the function of balancing the queue lengths on the two inputs. And there may be, but not usually, situations where signalising a merge might increase overall flow. But these two possibilities are secondary issues or side benefits.

** So what is the 'problem' we're trying to fix?
Merge UNsignalised: If the queue for the gyratory reaches back to this unsignalised merge point, traffic squeezes together in a take-it-or-leave it fashion. So, vehicles from Gelderd Road will most likely stay in the left lane, and those from the A643 will likely end up in the right lane. Before reaching the roundabout, those wishing or needing to swap lanes have only 200m in a solid queue in which to do it. Much of the swapping will occur at the last minute (human nature), just as the roundabout light turns green. This disorganised swapping and jostling reduces flow onto the roundabout. That's the flow we're trying to maximise.
Merge signalised: On the other hand, if the merge is signalised (so it's no longer really a merge!), drivers can lane-select at these lights, in a leisurely and safe way. The 200m queue up to the gyratory now flows in an orderly fashion, with less lane swapping, and traffic enters the roundabout smoothly and densely.

[Peter Freeman]

... I don't know if UK signal rules permit you to let traffic into the junction before conflicting traffic has fully cleared the conflict point, which is essentially what you'd have to do here to extract full advantage. ...

I would expect and hope that the rules don't permit that.

There is a related situation that TrafficLightMan (?) and I previously discussed in another topic. This is where a slip-road in the lead-up to a Give Way line and sign, has a signalised pedestrian crossing. The similarity is that a green signal might suggest that you can nonchalantly proceed over the dashed line onto a main road that legitimately has traffic flowing on it. The solution, in Australia and possibly elsewhere, is for the ped crossing to have a 2-aspect traffic light - red and amber, but no green - accompanied by a red-background sign that says "Stop here on red signal". In this situation, with no lights on, you're still on the lookout for give-ways.

Your merge signal on Gelderd Road could be made like this. It would still need to include a dashed line and a give way sign.