Crafting Green Waves - Radical, efficient, cheap, effective ways to prevent congestion and reduce trip time - 2021 Edition
- Toggle chapter buttons
Overview - Free flow, no bus lanes, parallel conflicts, choice of priority, quality operation
- Toggle page buttons
Crafting Green Waves, prevent congestion, install cross-overs, choose queue or toll

Preventing congestion permits free flow for all, allows signal linking that reduces delay for the bus, is better than a bus lane, and the alternative of a bus lane seriously congests all other traffic without reason.

Having turn arrows at intersections wastes time for all traffic, creates unsafe driver challenges, and is better replaced by cross-overs before the intersection, that move the turns to service roads. .

Free entry to congested roads manages demand by queue delay, yet does not prevent giving all traffic the choice of jumping the queue for a toll. Getting most people to jump the queue saves time and keeps the toll as low as possible.

These things serve the travelling public, nobody does them, but job security should be at risk.

Method, gate with signals, no turn arrows, 90% jump queue

Use traffic signals to gate traffic and operate the street in green waves at the speed limit and remove bus lanes. In the counter-peak direction, smooth flow at minor crossings.

Increase intersection capacity by 70% by processing right turn conflicts on the intersections’ approaches in parallel with running the cross traffic, so averting intersection turn arrows.

Treat most traffic as a priority by giving them a choice between free entry and a toll. The toll must be set so low as to attract 90%.

Performance, 1/2 crashes, 5x faster, faster buses

Crashes would be halved. Emissions would be halved. Exampled trip times would reduce from 40 minutes to 8 minutes. Pedestrian crossing delays would be halved. Bus priority would be improved.

Conventional intersections are relatively dangerous, risk litigation and should be remodelled at modest cost. Customers hate congestion.

Map of Example Route.

The most well-known congestion in Melbourne is at the city end of the Eastern Freeway that terminates at Hoddle St and Alexandra Pde. Hoddle St runs south along the eastern side of the city area and Alexandra Pde runs west across the north side. For clarity, only the southbound traffic in the AM peak period will be discussed in this example yet it is also customary to model traffic for both directions in both peak periods and to consider off-peak traffic.

The route as mapped from the Eastern Freeway at the Yarra River in Alphington and down the congested Hoddle St - Punt Rd route to the river at Cremorne, now takes 40 minutes in the AM peak period and that duration would grow over time but can be reduced to 8 minutes with no future growth!

Hoddle St from the Eastern Freeway to Victoria Pde has 4 lanes, drawn as the middle arrow, including a peak hour bus lane. 40 buses per hour from the Doncaster area travel the freeway shoulder, down the Hoddle St bus lane, then to the Melbourne CBD along Victoria Pde. From the north, another 5 buses per hour travel south along all of Hoddle St and Punt Rd, but south of Victoria Pde there are 3 congested lanes with no bus priority.

Capacity, 2x volume, 5x speed

The traffic flow volumes and trip speeds for the Hoddle Street - Punt Road route are far below the desirable standards for the existing road width. The existing road capacity provided is only 44% of what can be achieved within the existing footprint. The existing trip time is 40 minutes when it could be 8 minutes.

We do not wish to drive in, nor to have the bus in congested traffic. The first step is to constrain flow to below congestion level and to accept that excess demand will readily transfer to the bus. For this, we gate or meter traffic down to practical capacity, reducing flow by only 10%. At practical capacity, we can and would link signals in the peak direction and produce a green wave, that not only gives the bus a free run between intersections, like a bus lane, but linking also reduces bus delay at intersections.

A bus lane reduces the road capacity by 25%, or 33% depending upon the number of lanes available, but with no extra benefits for the bus, so it should be removed. Does capacity matter? If we reduce capacity by 25%, less traffic will depart the queue at the entry gate and nominally the remaining queue will be delayed for 15 minutes (25% of an hour) for every hour of gating the queue. Bus lanes and the congestion that they cause are installed because the alternative of metering is traditionally an unrewarded effort. Bus lane removal is justified everywhere, gating traffic should be mandatory where there is congestion, and green waves should be a key performance indicator.

At Victoria St, the practical capacity for a “Two-Phase” intersection should be 5,240 vph (vehicles per hour), but the existing practical capacity is only 2,289 vph (44%). A Two-Phase (2Pi) intersection has no turn arrows at the intersection, as detailed later. The traffic signals recorded a congested 3,341 vph. Hoddle St is congested north of Victoria St, and queues shuffle back through the signals at Langridge St, Gipps St, Johnston St and the Eastern Freeway, a distance of 2km, and spill back on to the Eastern Freeway for a further 2km from 6.30AM to 9.30AM weekdays. A priority lane is provided on the left hand freeway shoulder for free use by buses and selected government cars to jump the queue.


It is customary to model traffic capacity for intersections using the program Sidra and to compare the highest recorded traffic volumes from the SCATS signal records. Sidra computes the “practical” capacity, that is the highest traffic volume where free flow can be expected, as 90% of absolute congested capacity. The diagram shows the southbound AM peak capacities for all of the Hoddle St - Punt Rd route. Victoria St-Hoddle St is the most critical bottle-neck.

Steps to prevent congestion, increase capacity, bypass queues, and broader application
- Toggle page buttons
Gate - Meter Traffic, Link Signals and Smooth Flow.

The purpose of metering is to prevent congestion. Operation of Hoddle St and Punt Rd should be improved to produce a green wave, where a platoon of traffic flows through sequential sets of traffic signals at the speed limit, in a compact bunch without dispersing, with the minimum of wasted time, braking, acceleration, and emissions. This would be at practical capacity for the critical intersection, being 90% of absolute capacity at Victoria St. Signals at Victoria St count traffic, know when practical capacity is reached, and would tell the upstream signals when to turn red in order to gate or meter excess traffic. This message is passed back up the line to the Eastern Freeway, where vehicles may queue without reducing route capacity, at threat by the necessary reservation of a priority lane to jump the queue. The time gap between the traffic platoons will be relatively clear of traffic, allowing easier access to side streets and properties.

Having prevented congestion, signals along Hoddle St would be timed or linked towards Victoria St, being the peak direction, so that the platoon flows as smoothly as possible in a green wave. Linked signals must have the same cycle time so that the linking would repeat each cycle.

In the opposite or counter-peak direction, we are stuck with the peak direction timing, that may cause frequent disruption to the flow. Some smoothing counter-peak should be done at minor crossings by timing signals independently for each direction. So pedestrian crossings need to be staged and minor streets converted to “left turn out only” to smooth traffic flow in the counter-peak direction, reducing the number of stops and the CO2 emissions.

Then the outcomes are: minimums of trip time, CO2, and particulates; a reduction in associated noise; and easier access; that would all greatly improve the amenity of arterial streets. Congestion anywhere should never be permitted to occur, on the basis that the time it wastes is more costly than the alternatives. Congestion can best be removed by metering, that is much better than road pricing. Metering was successful for the Melbourne Commonwealth Games.

Two-phase, no turn arrows, cross-over on all approaches, radical capacity increase

Installing two-phase intersections on a four-lane road gives it the capacity of a six-lane road. Consideration for the customer can not lead to processing right turns in series with the through flows at the intersection when the intersection is the critical constraint, not to mention that it is quite dangerous as has been proven in Utah, and liability attaches. This basic but radical fact was known to our brightest minds yet they have failed to act on it, world wide! Every intersection turn arrow is wasting people’s time, including arrows on the cross road.

The queue delay on the Eastern Freeway is controlled by the capacity of the bottle-neck at Victoria St. If this bottle-neck is made worse by gating traffic, an extra 10% of traffic per hour will be delayed and all traffic will have to wait 10% of an hour longer in the queue, so capacity is very important. The best option increases capacity by 70%, by processing the right turn conflicts on all intersection approaches, in parallel with the cross traffic, instead of in series at the intersection, to create a “two-phase” intersection. A second immediate but inconvenient way to increase capacity by about 30% is to ban a pair of turns at the critical intersection. Drivers could turn left, then U-turn, or even find an alternative route of the same length instead. A third way that would increase capacity by 33% is to remove the bus lane that is no longer required, since there are free flow conditions for the bus. The bus should stop in the through lane for ease of operation. The number of intersections converted to two-phase operation should be a key performance indicator.

A further benefit of two-phase intersections is that signal cycle times can be reduced from the current 160 seconds to 60 seconds, reducing delays to all modes, particularly pedestrians. Both bus lane removal and two-phase intersections are extremely desirable.

Queue Jumping, choose queue delay or 90% jump the queue for a small toll

It is highly desirable for most traffic to jump the queue because of the obvious time saving but also because that would save fuel and cut emissions. Additional lanes, toll gantries and lanterns for the priority traffic would be required. A new priority lane should be installed on the north approach to the freeway, on the basis that provision should be made for priority traffic at all major queues, including at freeway ramp meters. The proportion of traffic jumping major queues should be a key performance indicator as should the amount of delay at major queues to priority traffic.

Currently the freeway shoulder is marked as a priority lane to be used by buses and government cars to jump the queue leading to Hoddle St, and it is controlled by a separate signal lantern. Highly valued trips made by community and business leaders, tradesmen, sales and deliveries are best made by private transport and constitute the most important and greatest quantity of traffic. They can not be avoided nor shifted to other modes without loss of value and they must be taken during the peak period. These trips warrant priority. Most vehicles now have e-tags so it is practical to toll the priority lane and permit all traffic to choose whether to enter the green wave via the queue or to pay a toll and enter via the priority lane(s). The free entry option should be retained, unlike the usual road pricing practice and the price must be set so low as to attract 90% of traffic, avoiding over-charging and minimising waste of time. There would be a relatively small increase to the queue delay.

Outcomes - The Impact of Metering, Two-phase Intersections and Queue Jumping

The bar chart compares (a)the existing state, with (b)being metered, implying removal of the bus lane and installing a green wave, then with (c)the added conversion of all seven major intersections to two-phase, and finally with (d)the further added queue jumping. The critical capacity shown in green at Victoria St is 3,300vph now, it increases to 4,500vph with removal of the bus lane, and increases further to 5,200vph with the addition of two-phase intersections.

The trip time shown in blue from the Yarra River at Alphington to the Yarra River at Cremorne is currently 40 minutes, it reduces to 25 minutes with removal of the bus lane, reduces further to 16 minutes with the addition of two-phase intersections, and finally reduces to 8 minutes for the 90% of express traffic that jumps the queue, but regresses to 23 minutes for the 10% of queued free entry traffic.


Because of the reduction of trip time, people get off the bus and back into their cars, so the demand shown in magenta that is now 4,200vph, increases to 5,300vph with removal of the bus lane, increases to 5,800vph with the addition of two-phase intersections, but then decreases to 5,400vph with the option of queue jumping. Note that a current capacity of 3,300vph and a current demand of 4,200vph means that 900 vph are added to the queue on the freeway. Also note that the queue delay increased by 7 minutes for the 520vph (10% of capacity) choosing not to pay the toll and that constrains demand down to 5,400vph. They may well be part of the 1,200vph for people that currently ride the bus but have switched to cars.

Emissions are strongly related to average trip time. CO2 emissions shown in brown are currently 12.5tonne/hr, they increase to 13.2tonne/hr with removal of the bus lane, reduce to 10.8tonne/hr with the addition of two-phase intersections, and reduce to only 5.9tonne/hr with queue jumping. Despite the much lower traffic volume, emissions for the existing traffic are only slightly lower because of the current slower trip speed. But queue jumping, by 90% of traffic when metered, drops CO2 emissions by an astonishing 53%! Jumping the queue is more important for emissions than capacity.

Currently the queues build to 2,000m on Hoddle St and spill back by 2,000m onto the Eastern Freeway. After metering, there will be no queues on Hoddle St, but 3,800m queues on the Eastern Freeway, reducing to 2,800m with the addition of two-phase intersections, and further reducing to a slow moving 800m queue with queue jumping.

Conclusion, lose all congestion, lose bus lanes, lose turn arrows, jump for a very low toll

Take Away: Find competent traffic engineers who can (a) replace all congestion with green waves, (b) convert all critical intersections to two-phase, and (c) install priority lanes with queue jump options for 90% of traffic at all queues.

Bus lanes should be removed and metering and signal linking immediately installed to better benefit the bus. Signals at minor streets and pedestrian crossings should be independently linked in each direction, requiring minor works. Two-phase intersections should be installed at seven intersections along the route, starting with Victoria St, enabling increased capacity so reduced queue length and delay, but also reducing delays to counter-peak traffic and to other modes. A priority lane should be installed from the north, possibly by extending the metering northwards to avoid reducing capacity. Toll gantries should be installed over priority lanes and all traffic given the choice of entry via a queue or priority entry via a small toll. The toll might be $1.50 to save 15 minutes after a two-phase intersection is installed at Victoria St. Until then, the delay for the 10% free access vehicles would be 46 minutes, not 23 minutes and the toll might need to be $3, saving 38 minutes, instead of the $1.50.

Arterials, no congestion, no turn arrows, jump queues for all arterials, customer focus

Serious consideration should be given to the reasons why public institutions world wide do not give enough priority to their customers. Green waves, two-phase intersections and queue jumping should be managed and reported by the traffic control room as key performance indicators, but under the umbrella of the ABS. For good governance, the reader will get the point about Governments not being trusted to report upon themselves. The existing outcome of neglect is hard to deny. The State department should report on KPI’s for sites or routes nominated by each municipality, to keep the customer focus.

A suitable policy might include private enterprise traffic operations teams, charged with replacement of all congestion with metered lengths and green waves, enhanced in steps by smoothing minor crossings, two-phase intersections, and tolled queue-jumping. They should be tasked with publishing the strategy for each route, with costings, target dates and KPI’s for each step. We might start with Hoddle St-Punt Rd and follow up with Alexandra Pde-Elliott Ave. A green wave is particularly relevant during every construction phase.

All congestion should be prevented and green waves implemented in the peak direction. Off-peak, pairs of arterials should be linked in opposite directions, for example Maroondah Hwy inbound and Canterbury Rd outbound, because the linked direction is much smoother and faster. Two-phase intersections, priority lanes and queue jumping should also be implemented, where congestion currently exists.

Freeways, no flow breakdown, queue jump at ramp meters, diverging diamond at interchanges

Regular operation without breakdown during peak periods is highly desirable. The ramp metering should probably be more stringent to achieve fewer breakdowns but the problem needs much deeper insight. It is worthy of a separate investigation.

A secondary issue is that ramp meter queues are excessive, getting worse and indiscriminate. Tolled priority lanes would be self enforcing and should be installed at all ramp meters. Priority lanes at ramp meters have previously been tried for trucks but failed because of lack of compliance.

Another secondary issue is that conventional interchanges, including single point interchanges should be replaced by “Diverging Diamond” interchanges, that are the interchange version of a two-phase intersection. Diverging diamond interchanges reduce delay 24/7 and increase freeway capacity where the exit ramp is a constraint.

Freeway interchanges have 200% green time compared to 53% green time for existing intersections so interchanges are not critical and lane changing becomes the main concern. Freeway demand is already well managed by ramp metering but there are options for improvement. Of prime importance is the prevention of traffic incidents associated with lane changing but this requires more investigation and is not covered here.

DDI, Diverging Diamond Interchange, no turn arrows

A Diverging Diamond Interchange (DDI) is where arterial flows cross to the wrong side of the road, over the structure, and then cross back again. This is the preferred interchange option. It enables flows from both directions to enter the entry ramps by simply diverging without further signal control, except for ramp metering and pedestrian crossings, and also allows flows from the exit ramps to enter the arterial road, signalised in parallel with one of the arterial road through flows, or by added lanes. Pedestrians and cyclists should normally cross the structure in the centre, as shown here in blue, between protective barriers.

This sketch is overlaid on the Eastern Fwy - Hoddle St interchange and is a pair of 2-phase intersections. North is rotated to the right. Note the quite short clearance distances. Hoddle St, Eastern Fwy has a Diverging Diamond Simulation

- Toggle page buttons
Bus-lane, prevent congestion, link signals, smooth flow, remove bus lane

To avoid congestion-delay to buses in Hoddle St, a bus lane has been reserved for them, reducing the capacity available to general traffic to 75% of its previous value. The bus lane has been very effective in removing delay to buses, but its side effect is an increase in the AM peak period delay by 15 minutes for 5,300 cars per hour, causing, 1,100 drivers to shift away from cars, mainly onto the bus, reducing the demand to 4,200 cars per hour, the existing state. While the existing bus lane removes congestion-delay to buses, that benefit could also have come from metering. The green wave and smoothing at minor crossings with metering would also remove delays to buses caused by traffic signals, so metering is of greater benefit to buses than a bus lane. Smoothing is presented later.

In Hoddle St, modelling shows that 1,100 people chose to use the bus rather than cop the extra 15 minutes delay. It follows that buses were available to substitute for 20% of trips, and that 20% of people prefer to take the bus than to endure the congestion. That is their decision to make, and it is based on their low value of the cost of mode change. But getting rid of congestion can: save 90% of people an average of 15 minutes each; improve amenity for residents; and with queue jumping reduce CO2 emissions by 53%. This is a demonstrated net benefit that would apply to some degree for all congested arterial streets. So to get rid of all congestion, anywhere, through metering, that need have no negative impact on capacity, net benefits apply, and the benefits of green waves would apply for all vehicles, including buses. This thesis rejects the current practice of permitting congestion and installing a bus lane.

A bus lane was installed in Stud Rd, where there were only 5 buses per hour, and reduced its capacity by 33%. This created huge delays for the general traffic, and caused such a public reaction, that it had to be removed. A similar response is expected for the 3-lane section of Hoddle St and in the 3-lane Punt Rd section of the route, where there are only 5 buses per hour, and such a proposal would reduce the capacity of the road by 33%. But metering should be installed there on the basis of removing congestion, with the benefit of also providing priority for buses. Metering is simply using traffic signals to prevent congestion, and reduce emissions and trip time. All traffic can use any lane. Metering creates or retains a queue.

Where there is no existing bus lane, capacity should be maintained preferably by installing two phase intersections or temporarily by banning right turns at the critical intersection(s).


The figure compares the capacity for a 4-lane approach in Hoddle St, Melbourne: (a) with a bus lane and 3 congested lanes; and (b) a metered 4-lane approach without a bus lane. For the 4 metered lanes, a green wave can be applied, and the total capacity is 3,840 vehicles per hour. For one bus lane and 3 congested lanes, the buses have no congestion, but do not get a green wave, general traffic is highly congested, and the total capacity is only 3,240 vehicles per hour.

Meter, provide 10% spare downstream

The figure illustrates the process of metering where traffic volumes entering the length are limited by the upstream signals. The traffic signal system records each vehicle at the stop line, noting the time gap between vehicles. Spare capacity is calculated at the downstream intersection and used to limit the flow at the upstream intersection. The objective is to provide spare capacity of 3 right and 7 through at Victoria Pde. This 10% spare capacity requirement only applies to the critical intersection in a control group. Less critical intersections will require greater spare capacity, governed by an intersection further downstream. The simplest control is just of the 81 in the figure.

But a platoon will disperse because traffic is variable, so the faster drivers then learn that they are held up by red lights and the slower drivers will miss the end of the green and learn to keep up. The traffic engineer needs to remove any obvious causes of turbulence, such as lane changing, and then the signals engineer adjusts the time offset between signals to keep the platoon flow as lumpy as possible. This will be near the speed limit. Drivers should be advised of the ideal travel speed.


Easier access and shorter trip times will make trips by car more attractive, and they will increase in number, including from sources in the Hoddle St vicinity. So when the 1, 4, 2 & 7 in the figure grow significantly, they may also need to be constrained. Options are to meter them, to toll them, or ban them completely, or by time of day. Such moderation requires endless maintenance. The growth just doesn’t impact the flows, it also impacts car ownership and parking availability.

If 3 right spare to the city can not be achieved by metering, and right turn queues block the through lanes, a city entry toll for right turns to the city at Victoria St may be required, perhaps for the whole morning peak, but free entry to the city would still be available via a right turn at Langridge St, Gipps St, Johnston St, or even Albert St. The only requirement here is that the intersection at Victoria St is not congested.

It has long been held that practical capacity is 90% of absolute capacity and that has been adopted here. The need to meter to exactly 90% of capacity can be modified empirically by adjusting the metering until it produces a satisfactory green wave, trading off higher capacity against smoother flow.

Smooth, keep right turn into minor streets, remove right turn out and straight across

Common practice on major arterial streets to avoid disruptions to the flow by minor crossings, is to not have a median break opposite side streets, and to have U-turn breaks elsewhere. Pedestrian crossings are staged. Alternatively, a Restricted Crossing, U-turn layout as shown, is provided.

Converting minor intersections to “left-in, left-out, right-in”, also known as “Restricted Crossing, U-turn”, and staging pedestrian crossings, allows independent timing of the signals for each direction. Although the time savings is small, the experience of many stops over a short distance is frustrating, energy consuming, increases emissions, and applies for all modes.

Where median breaks exist at minor streets, the right turns from the major road can be retained, but the straight across and right turns from the minor streets need to be re-directed. With the arrangement shown, linking for southbound traffic, fixes timing of left-in, left-out and right-in to Elizabeth St, on the right, and of pedestrians and cyclists crossing southbound traffic. Timing is different for northbound linking, that controls left-in, left-out and right-in to Albert St, and of pedestrians and cyclists crossing northbound traffic. U-turn opportunities may need to be provided in Hoddle St to replace the right turns and straight across for the traffic from the minor streets .

History, faltering steps to removing turn arrows

Existing intersections, particularly those operating in excess of 90% capacity, can be converted to two-phase, within the existing footprint with capacity and delay benefits as well as safety. This conclusion is widely applicable because because it was tested on 30 of the most difficult sites. They were sketched and modelled in detail to retro-fit 2Pi within their existing footprint and the result was a vast improvement in capacity, delay and alignment for each of them. Further, in contrast to the conventional bias towards increasing the footprint size, all the designs tested for this concept used the existing footprint, existing number of lanes and lane widths, and still achieved a major increase in capacity with better alignments than existing, and increased safety.

“Continuous flow” intersections are used in Utah on the Bangerter Highway. They have cross-overs on the main approaches to the intersections, but not on the side road approaches. Federal funding is limited to the highway approaches. Turn conflicts with opposing traffic are resolved on the main approaches with a “cross-over”, saving time lost at the intersection due to one turn phase. These intersections have ~35% improvement in capacity and delay and ~50% improvement in safety. Cross-overs are presented shortly. .

Common sense will tell you that the delays caused by turn conflicts from the side road are of similar size and should also be removed, although this has not been done before. It can fit within the existing footprint, and is more driver friendly if all approaches are the same. There are different ways to resolve the right turn conflict, but the essential requirement is for the intersection to have only two signal phases, and it should be known as such, a “two-phase” intersection.

Cross-overs for Utah’s “Continuous flow” intersections are positioned at a distance from the cross road such that the right turners only stop once, so the right turners have continuous flow. This places the cross-over ~200m from the cross road. This is unsuitable for the urban context with access to properties; for short flaring; does not have the option of multiple cross-overs; gives undue importance to the right turns; some drivers miss their turns; yet the proposed two-phase intersections do not have these problems.

Increasing capacity improves the quality of service by reducing the delay at an intersection queue. In fully developed urban areas, it can be quite disruptive to acquire property to widen the road reserve. So it is highly desirable to get the most traffic capacity out of the existing number of traffic lanes. Approaches to major intersections often have short additional lanes, known as flaring, but to match the mid-block capacity, that would need to double the number of through lanes, and to have full utilisation of those extra lanes. Most flaring lanes are usually reserved to safeguard turning traffic, with its differential speed.

In addition to flaring, a good strategy is to install a two-phase intersection, (2Pi), and to resolve right turn conflicts with the opposing through traffic on the approaches, within the flaring, timed to coincide with the cross traffic. This might require an additional set of traffic signals on each approach, but they are easily justified on the basis of a 70% increase in capacity, safer operation, lower delays, avoidance of property acquisition, better lane utilisation, and less turbulence. For prudence, these signals would only be required on the basis of traffic engineering analysis.

Two phase intersections are also called parallel flow intersections, continuous flow intersections, displaced right turns, or displaced left turns, but the descriptions are often inconsistent for the concept being used.

2-phase Intersections, extra capacity, better safety, reduced delay

Instead of sitting on the approach, doing nothing while the cross traffic runs, the right turners perform a “get-out-of-the-way” movement, by crossing the path of the opposing through traffic, thereby operating in parallel with the intersection cross traffic and avoiding the delay of a separate turn phase at the intersection. Separating these right turn conflicts away from the intersection simplifies driving decisions, and has an enormous benefit for safety, with only half the crashes, so all conventional intersections are now relatively dangerous, and should no longer be built on the bases of safety, capacity and delay.

The figure "Wasting Time" compares the allocation of time for the conventional 4-phase and the proposed 2-phase intersections, both with 120 second cycles. In practice, 4-phase intersections often need 120 second cycle times in order to limit the proportion of lost time: amber and all red. Longer cycle time increases capacity at the expense of delay. With 2-phase intersections they have no turn phases at the intersection, and capacity is increased by 69% for 120 second cycles, calculated from 90/53. But for the current proportion of lost time, 2-phase intersections should normally operate with 60 second cycles with only 51% more capacity and importantly, much less delay. Maximum delay for through cars is then one short phase of 36 seconds for 2-phase intersections, instead of 3 longer phases for 94 seconds for 4-phase intersections. The option of operating the 2-phase intersection at 120 second cycles with greater capacity is retained.


The simulation of the Victoria St - Hoddle St intersection carries 5,240vph southbound, it has a green wave, no residual queues and 57% more traffic than the existing congested intersection that has a volume of only 3,341vph southbound and 4km of queue, extending past four more intersections and on to the Eastern Freeway. If queue jumping is permitted, average trip time is then 9 minutes instead of the current 40 minutes.

Complex intersections with more than four legs can, and should be converted to four legs, where 2-phase intersections can be applied. The simulation of a 2-phase intersection with six approaches at Princes Hwy - Springvale Rd - Centre Rd shows how this can be done, although this intersection warrants a grade separation.

The safety, delay and capacity benefits of processing right turns on approaches should also apply for low volume un-signalised intersections.

Crossover, driver friendly, direct short cut, within existing footprint

A cross-over is a more driver friendly and much preferred option because the right turner enters the turn lane at the same location as usual; turners pass through the intersection only once; and the turn alignment looks, and is, a short cut with a better than normal alignment. This is particularly advantageous for trucks and at intersections with acute angles where split phases are currently necessary, because opposing turns cannot turn at once as a result of the geometry, so through and right from one direction then have to proceed together, less efficiently. The conflicting movements at cross-overs are simpler, slower and safer.

Note that the right turn lane(s) cross to a right hand service road half way towards the intersection, usually under signal control; and the alignment of the opposing through movement forms a plait with the right turn lane, staying within the current intersection footprint. The opposing through flow has high alignment standards. The right turn has sharper curves yet low off-tracking of 10cm if the turn radii are at least 30m. The right turners potentially queue twice but since the cycle time is shorter, the required total storage length is about the same, so intersection flaring does not need to be lengthened.


The simulation of Victoria St - Hoddle St shows cross-overs on the north and west approaches. A cross-over only interrupts one carriageway, so signals can be perfectly timed to link with the main intersection, and not disrupt the through movements. The right turners then cross in the shadow of the cross street operation. As shown in the simulation, signal timing often means that turners are not stopped at the second stop line, (it is continuous flow by chance).

Triple Crossover, saves time if the right turn is heavy

A triple cross-over is shown here is on the Hoddle St north approach so that a heavy right turn does not wastefully increase the phase time for the combined cross-over and Victoria St. The figure is rotated with north to the right. There is a heavy right turn from Hoddle St, with 40 seconds green, into Victoria St, with 10 seconds green. The turners have 10 seconds, and 6 lanes provided to cross the opposing Hoddle St through lanes, but only 2 lanes are needed to complete the turn into Victoria St at the main intersection, during the 40 seconds, and before the early cut off. Multiple cross-overs mean that turns are not critical for maximum capacity.

Pedestrians, shorter, faster, safer crossings with clear channels, fortuitous timing

Full signal control of all turns in slip lanes has better pedestrian and cyclist safety. Pedestrian crossings are shown in blue. Right turns would have an early cut-off, “ECO” and left turns a late start, “LS” to enforce sharing of the phase time in the road being entered. With both left and right turns in the same slip lane and under full signal control, pedestrians and cyclists in their own sub-phase, can then cross after the ECO from one approach and before the LS from the adjacent approach, without any impact on capacity and with greater safety..

This full signal control of pedestrians and clear channels for vehicles is safer than filtering right and left turners through pedestrians. The lower cycle times of 2-phase intersections, and staged crossings reduce delay to pedestrians and other modes, assisting compliance. Pedestrians cross fewer lanes and have fewer crossing stages than for conventional intersections. So that right turners are never stranded across through lanes at the end of the phase, right turns, from up the figure, are signed with a give way for the left turn, from down the figure, where they join the departure, labelled “GW”, because they are in the same phase. Line-marking similar to a roundabout as shown reinforces the priority.

Compromises - Only the cross-over option is driver-friendly.
- Toggle page buttons
P-turn, uncertainty to be avoided

Indirect turns are inefficient and confusing. Right turners in a P-turn: go through, perform a U-turn and then a left turn, as is shown in magenta on this sketch of Victoria St - Hoddle St. P-turn is to be firmly avoided because it is unusual; needs to be signed in advance; leaves open the option for people to turn right, illegally and unsafely, where they conventionally do; and incurs extra travel distance.

Regarding this sketch, having two-phase intersections, 2Pi, is critical for capacity and delay. On the south approach, the 100 vehicles per hour right turn lane is replaced with a Q turn, shown green, to gain an extra through lane within a restricted footprint, a gain of 1,000vph. On the east approach, the right turn is replaced with a P turn, shown magenta, to prevent queuing on the tram track. These P-turn and Q-turn options with low volumes increase capacity even though traffic re-enters the intersection. The option of simply banning these two turns is worse because drivers could still make the same manoeuvres but without signing guidance and with longer detours. The option with the highest practical capacity, for the target pattern of traffic demand, will reduce delays. The existing layout for this intersection has a practical capacity of 2,289vph southbound; it is currently operated highly congested at 3,341vph; and the 2Pi design sketched has a practical capacity of 5,240vph. Sidra was used to calculate the practical capacities.


The simulation of the Victoria St - Hoddle St intersection includes the P-turn and Q-turn on the sketch. While in theory, only two U-turns are needed to process four right turns at an intersection, and the example shows one U-turn serving two right turns, this should be firmly avoided because it is not driver friendly. It is only proposed here because it is such a critical intersection, and might save nominally 3,000vph 15 minutes each.

Q-turn, counter intuitive to be avoided

Indirect turns are inefficient and confusing. Right turners in a Q-turn: turn left, then perform a U-turn, as is shown in green on this sketch of Victoria St - Hoddle St. Q-turn is to be firmly avoided because it is unusual; must be performed from the left; needs to be signed in advance; and incurs extra travel distance. A Q-turn has been used at Moorooduc Hwy - Cranbourne Rd intersection.

Parallel Flow, waste of space and disconcerting

This option is inferior to the other three options. Parallel flow, is a right turn from the conventional location, but into a near-side service road, during the side road phase, crossing back to the normal lanes some distance down the departure leg. This option uses right turn arrows from the main road when the cross traffic is running. A presentation of parallel flow is on Youtube. Either the cross back to the normal location interrupts both directions so can’t be linked both ways to the main intersection, or the cross back is to an additional median lane, that then merges. Both require yet another lane for the left turn from the side road being entered, that must diverge before the cross back. Requiring additional lanes within a restricted footprint seriously reduces capacity. Parallel flow is used at Punt Rd-Olympic Bvd.

Continuous Left Turn, dangerous with hoons, disjointed for pedestrians

A continuous left turn for the length of the right turn service road, shown here in magenta, is dangerous because drivers run through the red signal too often at pedestrian crossings in continuous slip lanes, particularly when the turn radius is good and the speed is higher, and the lane is better used as an additional through lane or right turn lane. Pedestrian crossings are longer and have more stages. Prefer the arrangements shown previously, designed to improve pedestrian safety and increase capacity.

Unsafe Two-Phase Intersections, cheap and nasty

Intersections where right turns filter through opposing traffic and pedestrians, may have only two phases but filtering is relatively dangerous and is often done to maintain capacity on the cheap. They are literally two-phase but only because some conflicting movements are not signal controlled as they should be. Pedestrian safety is decreased by having vehicles filtering through pedestrians, and having more and longer pedestrian crossings than with full channelisation and full signalisation.

More Details
- Toggle page buttons
Queue Jump for all, not just the privileged few

Demand should always be constrained by delay at the metering queue. Some people constrained by this delay will take the bus. This delay wastes time for those caught in the queue and the queue increases the CO2 emissions, both of which should be avoided to the greatest extent possible. People whose time is more valuable should be able to choose to jump the queue with a low toll. If the toll is set too high, less than 90% will jump the queue and some valuable trips will be delayed. If the toll is set too low, all priority traffic will be delayed by metering. It is difficult to set the toll to have the desired effect, so the toll needs to be flexible. 90% queue jumping drops the CO2 emissions by ~53%.

This free entry queue is a buffer between the toll being too high and too low. The toll should be set so that priority traffic was rarely delayed, and the delay in the queue would then constrain the demand, but only if necessary to prevent congestion. If all traffic jumped the queue, priority traffic would be delayed by the metering.


The conventional area pricing is coarser. If there was area pricing, a queue-jumping toll would still be required, unless the toll was overkill. Area pricing is less selective in toll location and so its constriction will reduce the level of service. It does not have as much equity and does not moderate local traffic within the cordon.

To overcome any political negatives, the toll should be installed for three months and then removed. Users should be polled to measure acceptance and if strongly endorsed, the tolled queue-jump reinstated. The southern half of a diverging diamond interchange of Hoddle St with the Eastern Fwy is shown. Three lanes from the Eastern Fwy are proposed to be tolled and two retained for a free entry queue. Priority lanes are not proposed in Hoddle St, south of the Eastern Fwy, just a green wave. Note that a new priority entry is also proposed from the north. The toll should be pitched low for maximum uptake, saving time, estimated at $1.50 to save 15 minutes, and reducing pollution.

If the trip via the queued lanes on the Eastern Fwy is reduced from 40 to 23 minutes, demand from the Eastern Fwy grows to 96 vehicles/cycle. Metered capacity of Hoddle St is 110 v/c total, apportioned with 38 v/c from the north, and 72 v/c from Eastern Fwy. The Eastern Fwy toll is set so up to 64 v/c queue jump, and 8 v/c will be discharged from the queue. The excess of demand (96) over supply (72), 24 v/c, is added to the queue. If the toll is set too low, the priority lanes will be limited to 72 v/c by metering. If the toll is set too high, the balance of the 64 will be discharged from the queue. Metering combined with road pricing reduces trip time and emissions but we need also to retain the equity of free entry.

Metering prevents congestion but produces or retains a metering queue. The most critical intersection for the southbound traffic in Hoddle St is at Victoria St, but the metering must be cascaded through successive intersections to constrain the queue 2km upstream at the end of the Eastern Fwy, otherwise the queued lanes plus the lanes required for priority traffic will create a greater restriction.

Queue jumping is also known as congestion charging, FasTrak, and high occupancy toll (HOT) lanes and is a form of road pricing where there can be a variable toll. High occupancy lanes are not being proposed, just tolled priority jumping of the queue, as happens now for a privileged few for free. User choice should at least be available for all, rather than a select few, and it can cut emissions in half. Huge benefits would accrue to the community from faster trips, and there is still free access.

Proportion Jumping, more & more save time

For time efficiency, and low CO2, as many people as possible should get priority entry. The practical limit to this is probably 90%.

Assuming we have not yet installed two-phase at Victoria St, with no queue jumping except for the bus, all traffic takes 25 minutes. But if priority vehicles jump the queue and take only 8 minutes, for a toll, then modelling with 50%, 70% or 90% jumping, the free entry vehicles would take 41, 44, or 46 minutes respectively, demand would decrease from 5,300 to 5,100, 4,800, and 4,600 respectively, and the queue length would decrease significantly.

Average trip times for all traffic would be 25, 16, 13, and 10 minutes respectively, so 90% is the best outcome. Priority trip time drops from 25 minutes to 8 minutes with queue jumping, saving 17 minutes, but the free access queue trip time increases, with an apparent saving of 38 minutes for queue jumping, so the toll might have to be $3 while there is no 2Pi at Victoria St. The practical trade-off is between frequency of metering priority traffic and quantity of trip time saved.

Demand and Equity, reduce trip time and they come from the bus

With 2Pi at Victoria St, for tolled queue jumping, the free queue travel time for the trip increases by 7 minutes for 10% of traffic, and for the tolled 90% of traffic reduces by 8 minutes. This is a net benefit and it provides a new express travel option for essential trips. Note that there will still be 28% more trips, from mode change, than currently exist, but an additional 11% will have gone back to their previous mode. This puts the queued 10% proposed to be used for demand management, into perspective. The target is a reliable 8 minutes, instead of the current 40 minutes.

Metering and two-phase intersections will increase capacity. Queues will clear quicker and travel times reduce. Then demand will increase by about 2% for each travel time minute saved. Because of metering and two-phase intersections, more and more mode change trips occur. Trips will be made by car instead of being on the bus, on bikes, by walking, made at other times, or not made at all. These mode change trips are of low marginal value to the maker to be made by car.

My bleeding heart says why should the poor people be subject to delay? But give me a break, these people are currently riding the bus.

Traffic Island Grading, attention to detail

Traffic islands must have kerbed cut-throughs for the vision impaired, as shown here, but use three full width tactile tiles and no internal tiles. It should have flat, 2% grades for self cleaning, not steep kerb ramps. All these improvements assist pedestrians, the vision-impaired, prams, wheelchairs, cyclists and traffic.

At major intersections, cyclists must be encouraged to mix with pedestrians rather than with car traffic, using wide cross-walks and cyclist lanterns.

Monash Single Point Interchange, heaps of lost time, lacks queue space & priority option

A Diverging Diamond Interchange, as sketched on top, would have a delay cost of only $5,012/hr compared to a delay cost of $11,115/hr for the existing single point interchange, for the same size footprint, using the existing structure. It would better accommodate ramp meter queues. Because of the much shorter clearance distances and fewer phases, the 2-phase DDI has: half the delay; twice the capacity; half the crashes; and fewer gridlocks, compared to a single point interchange. Converting all freeway interchanges to diverging diamonds should be given serious consideration for safety, delay and capacity reasons.

Two lanes have been provided for the right turns to the ramp meters, one of which should be a priority lane. The single lane left turn to the ramp meter can be either a priority lane or not, but not both.

Where there are ramp meters, the meter can be at the top of the ramp, with queues extending through the structure. Ramp meters limit the flow of traffic to the freeway to avoid flow breakdown. Meters at the top of the ramp reduce the required ramp length, and increase the distance to the next freeway exit, so reducing freeway turbulence. A video for Marvin Rd interchange in Washington State reported 60 DDI’s built since 2009, having greater capacity, with 50% crash reduction, and lower crash severity.

This single point interchange at Monash Fwy, Toorak Rd , previously considered efficient, and having only 3 phases, has double the delay of a diverging diamond interchange. Note the long existing clearance distances, covering the width of the Freeway underneath, that cause considerable lost time with amber and all-red and these long clearances are more likely to fail and be blocked.

Freeway Operation, excess breakdown, needs priority toll, needs anti-swooping, needs lidar

Congestion on freeways leads to flow breakdown, loss of throughput, consequent increased delays, and should never be permitted by the control strategy, but this is not consistently achieved.

Ramp metering already makes provision for high priority vehicles to jump the entry queue, but it needs to be further developed to include a priority toll, as is proposed above for arterial streets, to permit as many vehicles as possible to jump the queue, where the queue delay or queue length is significant. Allowing 90% traffic to use the priority lane for a small toll, reduces the ramp meter queue length by 90%, increases the queue delay slightly for the 10% free entry queue, slightly reduces the demand, reduces the queue CO2 emissions by 90%, and reduces the average delay by 90%. Both the priority and queued lanes need to be metered, but the priority lane should normally flow freely. This 90% reduction in queue length is important where the ramp meter queue is 1.5km long, and jumping is important when the queue delay is 30 minutes.

The need to prevent crashes is even more important for all safety, delay and reliability aspects. Contemporary understanding of freeway crashes is rapidly developing, and it is considered that two classes of freeway crash are prevalent: those at medium density, at moderate speed, associated with lane changing with 0.3 second gaps, with serious consequences, described as “swooping”; and those at high density, at low speed, with minor consequences, and described as “rear-end”. Control systems seek to avoid occurrences of high densities and rear-ends. Regulation of swooping needs to be invented.

Limiting the exposure to swooping is restricted by sub-par observation equipment. Professor Boris Kerner has defined a 3-phase model of traffic flow that better explains observed traffic behaviour than the widely accepted two-phase model. Not to be confused, the two phase model of traffic flow relates to the speed, volume, and density of traffic flow; but two-phase operation of signals is the number of primary signal phases at an intersection. John Gaffney and Matthew Hall of VicRoads have observed freeway traffic behaving according to the 3-phase model, including nucleations, derived from TIRTL measurements at 500m intervals on the Monash Freeway and associated road crashes. Refer also to John Gaffney’s Churchill Fellowship report.

Individual vehicle behaviour cannot be observed in detail from the TIRTL measurements. Better equipment to observe traffic behaviour is used by driver-less cars and should be used by the road operator for safety research and more generally for signal control. Jeff Hecht writes that: automotive lidar remains in flux; long-range microwave radar works better in bad weather; ultrasound is best for parking assist; short-medium range radar is good for cross traffic, rear-end collision and blind spot detection; optical cameras have good resolution and can detect traffic signals; and lidar can directly measure distance and speed for objects up to 300m. An example from Luminar lidar is shown, but it is likely that a combination of a few technologies can best observe traffic and will come from the car industry.

While the process described above is proposed to craft green waves on arterial roads, it has limited application for freeways. Melbourne’s freeway and tollway network carries 30% of the arterial road traffic although comprising only 7% of the arterial road network length. Progress is being made on managing freeway operation to achieve optimum speed, without congestion, and to establish reliable trip times.

Ramp metering to control freeway traffic has proven successful as described in VicRoads’ Freeway Ramp Metering Handbook. The method uses occupancy as a criterion to limit inflows and avoid congestion. Managing varying entry flows across adjoining ramps can spread the micro peaks, decrease freeway turbulence, and increase freeway capacity.

Freeway Operation

Speed control is used on freeways related to traffic incidents, including roadworks, and care is taken with the rate of change of speed and associated densities to avoid unsafe states. VicRoads practice is to reduce speeds in 10kph increments during incidents because of road safety. The Police practice of stopping vehicles on freeways, requiring a 60kph speed change increment has proven to be relatively dangerous and should be avoided. Better to direct offenders to an exit ramp.

It is necessary for Police to make safe accident sites on freeways, but the practice of closing the road for investigation beyond the time needed to clear the site, is excessive and can not be justified. Often the cause of the crash was not involved in the crash and has long departed.

- Toggle page buttons
Streamlining Hoddle Street, costly works with little impact

The original proposal for Streamlining Hoddle St, for a total of $41.8M, was checked by Treasury. It included two-phase intersections for seven sites: Eastern Fwy, $6.9M; Victoria St, $6.5M; Johnston St, $3.8M; Wellington Pde, $6.2M; Swan St, $5.1M; Alexandra Ave, $2.8M; and Gipps St, $2.5M. Smoothing works, $7.0M, were proposed for Truro St, Langridge St, Albert St, Hotham St, Freeman St, Rowena Pde, and Richmond Tce. Albert St was the example for smoothing. A new connection to Brunton Ave, $2.0M was proposed.

VicRoads reviewed these designs and have implemented different concepts but only at three of the intersections, Eastern Freeway, Johnston Street and Swan Street/Olympic Boulevard, yet has not adopted the green wave concept requiring metering, nor smoothing flow at minor crossings, nor tolled queue jumping. The cost increased from the estimated $16M to $110M for the three intersections. There is little evidence that the Government’s stated objective of improvements to the trip time along Hoddle St and Punt Rd were considered. Until the major constriction at Victoria St has two-phases, the capacity of the route will be unchanged. Current improvements will have inbuilt capacity constrictions that will then be tested. Unless metering removes congestion, express trips are not available. When high priority queue jumping is adopted, extensive queue delay and high CO2 emissions will reduce. The prescription is included herein. VicRoads have done some road works for capacity but none of the improvements for trip time.

Victoria Street, Hoddle Street, Two-Phase Simulation incorporates many radical concepts

The simulation has 50% more traffic than than the existing congested intersection. The sketch for the two-phase layout is below.

Eastern Fwy, Hoddle St, Diverging Diamond Simulation caters for all movements

The simulation has 50% more traffic than the previous congested interchange. The diverging diamond layout is shown below.

Alexandra ve, Punt Rd, Two-Phase Simulation, narow bridge with turn bans has been kept

The simulation has 50% more traffic than the existing congested intersection.

Bridge Rd, Hoddle St, Two-Phase Simulation, narrow west approach is a problem

The simulation has 50% more traffic than the existings congested intersection.

Swan St, Punt Rd, Two-Phase Simulation, need for a double right turn forces a P-turn

The simulation has 50% more traffic than the previous congested intersection. .

Johnston St, Hoddle St, Two-Phase Simulation, narrow widths permit cross-overs with give way control

The simulation has 50% more traffic than the previous congested intersection.

Springvale Junction, Two-Phase Simulation, complex but it can be signed

This six-leg junction is converted to four legs for simulation of two-phase operation. The U-turn is better done from the middle of the road. This site warrants a grade separation.

Stud Rd, Ferntree Gully Rd, good for cars but inspired changes for pedestrians

This was the first site designed for 2-phase operation and two simulations comapre the two-phase layout with the existing layout. It is an early stage of development of the 2-phase intersection concept, and is not actually a continuous flow layout, but just a 2-phase layout. The next stage of development of the 2-phase intersection concept, was the removal of the left turn bypass, replaced by additional turn lanes, and simplification of the pedestrian crossings with introduction of right turn priority, right turn early cut-off (ECO) and left turn late start (LS) that is now used generally.

Stud Rd, Ferntree Gully Rd, Melbourne, Australia, two-phase simulation by Dr Sarath Premachandra. Triple speed simulation. Check out the black truck that arrives at 2.08 and takes 1 minute to clear the intersection (20 secs at triple speed). This layout has 3 double right turns, 1 single right turn and 4 continuous left turns. The right turns get around in a quarter of the time. .

Continuous left is now considered dangerous for pedestrians and now 2 triple right turns, 2 double right turns, 2 triple left turns and 2 double left turns are proposed with both left and right turns signal controlled in the one slip lane. Pedestrians cross the slip lane after the early cut-off from one leg and before the late start from the adjacent leg. A further major problem with the simulated layout is that the continuous left turns take up an additional lane width on each approach.

Stud Rd, Ferntree Gully Rd Melbourne, Australia, existing layout simulation by Dr Sarath Premachandra. Triple speed simulation. Check out the black truck that arrives at 2.08 and takes 3min 48secs to clear, including two stops in the through lanes (1.16 at triple speed). Following this simulation, a second right turn phase from the west was added.