Roundabout Signals under TFGAP.
Version of 15 October 2005.
Dave Barber.
TFGAP-A

In section one we discuss some general properties of roundabouts, because they are not well known. Although these intersections ideally do not require traffic lights, in section two we look at how roundabout signals might be implemented should they become necessary.


1. Introduction.

1A. General. The roundabout, a variety of circular intersection currently receiving much attention, normally exhibits the following properties, among others:

The number of legs need not be four, although a greater quantity may necessitate a larger radius for the circle. Any of the streets can be one-way. Some roundabouts support multilane operation.

Below is a typical roundabout. Some aspects, such as pedestrian treatments, have been omitted from the drawing; what remain are the vehicular paths. For greater detail see MUTCD section 3B.24. The reader is cautioned that signage and pavement marking for circular intersections vary widely.

At a successful roundabout, vehicles rarely need to stop. Rather, they merge smoothly into the circle. The curvatures of the pathways are designed sharp enough to ensure that all motorists travel at a low speed, typically 20 to 40 kilometers per hour. With that, a driver who does have to stop and wait can accelerate to prevailing speed and enter with little difficulty.

In this report, we term the conventional noncircular intersection, ordinarily governed by stop signs or traffic lights, a plain crossing.

1B. Regions of the Intersection. Below in magnification is the core of this roundabout.

Omitted from the illustration are the directional arrows from above, but added to the drawing are yield signs and the corresponding pavement markings (consisting of a row of white triangles) behind which cars wait to enter. Also included are literal symbols for regions of the intersection, which are decoded thus:

With this notation it becomes easy to iterate a driver's route. If he enters the intersection from, for instance, the south:

The loop is of course the path nc-nwc-wc-swc-sc-sec-ec-nec-nc. A driver who misses his turn can travel around the loop and try again; this motorist is said to take a redundant path because he travels some section of the loop twice.

1C. Yielding. It is crucial to the roundabout design that entering traffic yield to circulating traffic. If instead circulating drivers must give way to entering drivers, the intersection can lock up because there is very little room in the loop for storing stopped vehicles: regions nc, wc, sc and ec are typically no longer than two car lengths.

If two drivers are coming from opposite directions and both are turning left, their routes will cross twice. In the example pictured below, left-turning drivers are entering from north and south, and their paths meet at nwc and sec. With proper yielding, they will not be in the same place at the same time, and so will not crash.

Were the circulators-yield-to-enterers rule to apply, this left-turn double cross in roundabouts would practically guarantee that queues of left-turning traffic would eventually block each other during periods of heavy volume. Similar deadlock can occur even in the case where only one entrance calls for circulating traffic to yield to entering traffic, with all other entrances giving circulating motorists priority over those entering.

By contrast, the double cross is not a feature of the plain crossing, below. As discussed later, this affects the volume of throughput.


2. Signals.

2A. General. Engineers normally refrain from recommending construction of a roundabout intersection if expected volumes are so heavy as to require traffic lights. Indeed, any stop signal runs counter to the whole point of the roundabout, which is to keep vehicles moving.

Yet if vehicle counts increase sharply after a circular intersection has been completed, the best remedy may be to simply add signals to some or all approaches rather than to rebuild the whole intersection in some other form. Poor performance due to overloading can be exemplified by such dense platoons of traffic from one entrance that drivers on other approaches have little hope of finding a safe gap in circulating traffic and entering the intersection.

2B. Signals Upstream for Entering Traffic. Entrance metering is comparable to the ramp metering used on many urban expressways. Much as an expressway ramp can have a one-car-per-green signal located upstream of the merge point, the roundabout can have a one-car-per-green signal upstream (perhaps fifty meters) of the yield bar; a several-cars-per-green arrangement may also be feasible. Motorists will still have to yield when reaching the loop, but traffic from no one entrance will be able to dominate the intersection.

As with ramp meters, planners must ensure that entrance metering signals do not cause unacceptable backups elsewhere. Entrance meters need not be installed on every approach, but only on those that carry heavy traffic. One constraint is that the traffic light must be installed far enough upstream from the loop that motorists do not think that the green signal cancels the obligation to yield. For more on metering, see TFGAP-M.

2C. Signals at the Loop for Entering Traffic. For traffic entering the loop, the yield sign and yield bar can be replaced with a signal and stop bar. The figure below shows where those stop bars would be, although it does not picture the signals. Essential to note is that traffic, once in the loop, has no signal or yield requirement.

Because circulating traffic is not controlled, approaching traffic must yield even when it has a green signal. In the language of TFGAP-I, entrance to the loop is permissive but not protected. Motorists may find this frustrating, as a green light offers no guarantee that they will be able to proceed. There is a similar lack of guarantee at plain crossings, where drivers when making a left turn must give way to oncoming vehicles -- heavy volumes of conflicting traffic may block their advance. At a roundabout there is no way to give protection to entering traffic without somehow controlling circulating traffic, and that control raises the danger of lockups.

Metering. The ONE CAR PER GREEN sign often installed in metering applications is not appropriate here because this signal is not upstream of the loop, giving a buffer space. Instead, the signal is situated for approaching motorists who have already reached the edge of the loop.

Those drivers may believe that the sign, while establishing a regulatory maximum of one vehicle per green light, also declares one vehicle to be the legal minimum. Hence a driver who is at the head of the queue at the beginning of the green cycle could infer that he has an obligation to proceed before the light changes back to red. Believing that, he may additionally infer that circulating traffic will be held back, and thus it will be safe for him to enter the loop. Since in reality circulating vehicles are not controlled, a crash is likely to result.

Entrance metering, on a several-cars-per-green basis, can still be effected when signals are at the edge of the loop. Needed is a vehicle detection system that counts the vehicles entering the loop on green, and that changes the light to red after either a predetermined number of cars have passed or a predetermined time interval has elapsed. If sometimes an extra vehicle squeezes through on a late yellow signal, the overall effect of metering will not be lost.

Appearance of the Signal. Because the roundabout differs greatly from the plain crossing, we suggest that a signal placed at the edge of the loop for entering traffic use neither the traditional arrow or dot, but rather a new symbol exclusive to roundabouts. Here is one possibility, the key:

It is a greatly simplified depiction of the roundabout, showing only one entrance (the approach being used by the driver who sees the signal) and the loop. Any attempt to fit all the legs of the intersection into a 300mm lens would result in crowding; plus, municipalities would be inconvenienced by having to stock three-leg, four-leg, five-leg and other versions of the lens. Instead of squeezing all this detail into a traffic signal, street departments can display full routing information on guide signs installed along the approaches. Another reason to show only one leg of the roundabout is to emphasize that this signal governs the juncture of the circulating road with only the immediate leg: other approches will have their own controls.

To be consistent with TFGAP-I, the signal should use:

As the flashing green light is a new idea, being introduced with TFGAP, an educational sign reading YIELD ON FLASHING SIGNAL can be installed. If the signal is installed for metering, a NO TURN ON RED sign should be added because motorists will attempt a right turn on the red signal, which defeats the whole purpose of metering. If the traffic light is erected for some other reason and the turn on red will be permitted, the red signal should flash.

Unsuitability of Arrows and Dots. The ordinary arrows of plain crossings, as pictured below, are ambiguous when applied to a roundabout.

Among many reasons is that some motorists will regard the roundabout as one big intersection, and others will consider it as several tee intersections in rapid succession. For instance, at driver entering at sa might think a left arrow is specifically for vehicles leaving at wl -- this is the "global" point of view. Meanwhile, another driver entering at sa could believe that the arrow is for any driver who will veer left at the next split and proceed into ec rather than el -- this is the "local" point of view.

Dots are no more successful. The clearance phase requires a yellow signal, but the permissive nature of loop entrance calls for it to flash. Together, these would lead to the flashing yellow dot which already has a much different meaning, telling drivers to use caution but not in any way directing them to yield.

2D. Signals at the Loop for Both Entering and Circulating Traffic. Below is illustrated a roundabout with stop bars (pictured) and signals (not pictured) for all motorists at every merge point. Because circulating traffic can be stopped, phasing sequences must be carefully selected or lockup will occur.

First we should look at the merge points in detail, for example where sa and sc combine to form sec. If sa gets a protected signal (steady green), then sc must receive either an outright stop signal (steady red), or some sort of yield signal (under TFGAP flashing red or green); by symmetry this statement continues to apply if sc and sa are exchanged.

Because signalized roundabouts are rare and many drivers will have to figure out what to do, we suggest that only steady red, green and yellow phases initially be used, with NO TURN ON RED signs on each traffic light. With this simplification, if sa has a green or yellow signal, then sc must have a red one, and vice versa. This is the most basic scheme, although not the most efficient. As drivers become familiar with it, engineers may consider introducing red or green flashing signals to increase throughput.

If no flashing signals are used, the signal faces could be ordinary dots and the intersection would function correctly. An improvement, however, would be to display for circulating traffic these signals:

and for entering traffic these:

The key pointing right and the key pointing down are the same shape, in different orientation. Such lights will remind all motorists that they are indeed in a roundabout and not a plain crossing.

During times of light volume, the roundabout can operate as if unsignalized by setting all the circulating signals to steady green and all the entering signals to flashing green. When the flashing green is used, the flashing yellow will be required for the clearance phase, and then the key signal must be used instead of the dot.

A Sample Schedule. The following table displays one possible roundabout phasing plan. Each row shows all the signals that would be lit at once, while each column shows the color of a single signal throughout the progression. The principal phases are odd-numbered, while the even-numbered phases are used only for clearance intervals lasting between three and six seconds.

In this plan the only drivers who would be blocked, once in the loop, are those taking a redundant path; they should number very few. If minimal u-turn traffic is also expected, which is likely, other scheduling patterns may be useful. Still, a heavily-signalized roundabout will probably not be as efficient as a heavily-signalized plain crossing. A major cause of this is the double crossing of simultaneous opposing left turns.

In the pattern below, not counting clearance phases, signals for circulating traffic are green about 75% of the time, while signals for entering traffic are green only about 25% of the time. This pronounced difference stems from the need to prevent lockups.

StepSignals for vehicles attempting to leave regions . . . Comments
scsaeceancnawcwa
1 wa may enter, leaving by any leg
2 clearance
3 na may enter, leaving by any leg
4 clearance
5 ea may enter, leaving by any leg
6 clearance
7 sa may enter, leaving by any leg
8 clearance
Return to step one.


3. Advance Warning Signs.

A1A2

When a roundabout has its signals for entering traffic at the edge of the loop, as described in sections 2B and 2C above, the combination of signs A1 and A2 pictured above can be replaced by sign B1 below. For simplicity it depicts dots instead of keys, and omits the yellow light entirely. An educational plaque reading SIGNALIZED ROUNDABOUT may be added. Alternatives are signs B2 and B3.

B1B2B3

Occasionally there may be some reason to emphasize that a roundabout is not signalized. One instance occurs when traffic signals at the intersection have been removed and replaced with ordinary yield signs; another instance is when in a series of roundabouts along a highway, nearly all have signals and drivers must be alerted to the exceptions. Sign C1 takes care of this. Meanwhile, sign C2 is the logical extension for the peculiar case when motorists must stop before entering the roundabout.

C1C2