Вы находитесь на странице: 1из 19

SIGNALISATION OF ROUNDABOUTS

Based on work done by Marie-Therese Conway, MCOS

Summary

Majority of signal controlled roundabouts started out as priority controlled with signals installed at a later
date.

However, in recent years in the UK and Ireland purpose-designed signalled roundabouts have been installed
demonstrating some belief in the effectiveness of the junction arrangement.

The introduction of signals on a roundabout can:


• regulate traffic patterns,
• reduce the need for weaving and merging,
• remove gap acceptance problems and reduce speeds.

This paper examines four case studies in order to look at the potential of signalising conventional
roundabouts and comparing which arrangement should be installed pre-signalisation to optimise the benefits
from signalisation if required at a later date.

Some of the findings suggest that not all roundabouts perform better under signalisation and the fewer the
nodes that need to be signalised the closer the signalised geometry will be to priority operation.

INTRODUCTION

Russell, Luttrell and Rys (2002) compiled a brief history of roundabouts and how they have been used in
different countries.
Earlier versions of roundabouts gave priority to entries on to the roundabout but this caused problems with
locking up on the circulating carriageway.

Research in the UK led to the idea of ‘yield at entry’ (give-way rule) and this solved the locking problem,
improved capacity, reduced crashes and created a complete change in philosophy of roundabout design and
operation (Brown 1995).

Kimber (1980) made a study of the entry capacities of roundabouts at eighty-six public road sites and is
widely referred to as it produced formulae for capacity prediction.

It was found that the most important factors influencing the capacity of a roundabout are the entry width
and flare and that the angle of entry and the entry radius have small but significant effects on the entry
capacity. More recent research by Omari et al (2004) found in a study of delay on roundabouts in Jordan
that the entry width has the greatest influence whereas the circulating width has the least influence on
delay.

The ‘roundabout revolution’ has only been a comparatively recent phenomenon in the US.

Jacquesmart (1998) stated that there was a total reduction in accidents of 37% and in personal injury
accidents of 51% by using roundabouts.

The SIDRA model is used in the US for design but the European models, RODEL and ARCADY, are also used
(the latter will be used in the research reported here).

Akcelik (1998) suggests that the ARCADY model is over-sensitive to geometric characteristics such as
internal circulating diameter (ICD), approach half-width, entry width and favours SIDRA as a means of
modelling more accurately driver behaviour using gap acceptance modelling.
Akcelik and Besley (2001) highlight the potential of the newer microsimulation models for modelling complex
gap-acceptance situations experienced in urban areas.

Tan (2001) presented a paper which compared the capacity of roundabouts and signalled junctions.

Tan (2001) concluded:

(i) at a mini-junction the full capacity of roundabout design is always larger than that of signal
intersection design

(ii) at a small junction the full capacity of roundabout design is larger than that of signal
intersection design in most cases

(iii) at a moderately sized junction, the full capacity of roundabout design is larger than that of
signal intersection design when the ratio of turning left to going ahead traffic is relatively
low, otherwise the full capacity of signalised intersection design will be larger than that of
roundabout design

(iv) at larger junctions, the full capacity of roundabout design is less than that of signal
intersection design in most ordinary cases.

Sisiopiku and Oh (2001) compared the performance of roundabouts with four-leg intersections under yield
control, two- and four- way stop control and signal control for various traffic conditions using the SIDRA
model.
They found that the performance of roundabouts compared well to the performance of signalised
intersections with one lane approaches and heavy traffic volumes.

Lines and Crabtree (1990) looked at reducing delay and increasing capacity of roundabouts by signalising
one or more arms and recommended how such junctions should be modelled using the TRANSYT (Traffic
Network StudY Tool) model (Binning and Crabtree, 1999).

The TRANSYT model was developed by the TRL in the UK to search for the optimum fixed time plan with
which to co-ordinate all the traffic signals within the network to be modelled.

Two important criteria which need to be considered when examining signalisation of roundabouts but which
do not exist in the case of non-signalised roundabouts are:

(i) the design needs to allow more than one stream of traffic to enter the roundabout without
conflicting with each other
(ii) provide space where vehicles can stop and queue at the stop lines on the circulating
carriageway until all conflicting traffic has cleared.

While (ii) is important, it would be equally important to try to keep traffic stopping on the actual
circulating carriageway to a minimum to avoid locking up.

The capacity of a roundabout can only be maximised if traffic can enter and leave the roundabout as
quickly as possible so traffic should meet at least two green signals in a row i.e. to be able to enter the
roundabout and pass through at least the first circulating carriageway stop line without having to stop.

Most signalised roundabouts benefit from a short cycle time as the shorter the cycle time the shorter the red
time leading to lower queues on the circulating carriageway.

Generally a cycle time range of 30 – 90 seconds is used.


The size of the roundabout will also influence its throughput; the larger the size the more the roundabout
will resemble any normal series of signalised junctions. The smaller the roundabout the more important
signal node coordination becomes.

The traditional saturation flow calculations from Kimber at al (1986) are recommended for calculating the
saturation flow (presented below) at stop lines on signalised roundabouts.

2080 − 140d n − 42d g G + 100(w − 3.25)


S = Eqn 1
f 
1 + 1 .5 
r 
where

2080 passenger car units (pcu) per hour is the saturation flow for a non-nearside 3.5 m wide lane
with no turning traffic,
dn is a dummy variable = 1 for nearside and 0 for non-nearside lane,
dg is a dummy variable = 1 for uphill and 0 for downhill,
G is the gradient,
w is width
f is the proportion of the flow that is turning and
r is the radius of turn.

The major contributors to affecting saturation flows are the w, f and r terms.

In terms of the analysis conducted here f is considered to be fixed.

So in order to maximise the saturation flow on a stop line r and w need to be maximised.

It should be noted that lanes can only be widened so much and checks need to be done to see if an extra
lane would be more efficient than widening the existing lanes.

Maximising the r term has the effect of straightening the approach to the stop line which may bring the
decision maker back to the fundamental decision of whether to use a signalled intersection or a roundabout
intersection.

This work using four case studies, looks at the potential of signalising conventional roundabouts and
comparing which geometric arrangement should be installed pre-signalisation to optimise the benefits from
signalisation if required at a later date.

METHODOLOGY

Analysis of both conventional roundabouts and signalised roundabouts was required in the analysis.

The former was done using the ARCADY model Binning (2000) and the signalisation analysis was done using
TRANSYT (Binning and Crabtree, 1999)

A description of both follows.

ARCADY

ARCADY (Assessment of Roundabout Capacity and Delay) is a computer program for predicting capacities,
queues and delays at roundabouts.

Single roundabouts with between 3 and 7 arms can be modelled including grade-separated roundabouts.
The program uses empirical formulae for calculating the capacity of each entry arm as a function of the
circulating flow crossing in front of the entry.

The operation of the roundabout as a whole is calculated on the basis that the entries to the roundabout are
linked by the common circulating carriageway.

Queues and delays are calculated using time-dependent queuing theory.

The first fundamental feature of ARCADY is the equation relating the entry capacity to circulating flow.

For standard roundabouts this is given by Kimber (1980) as

Qe = k (F − fQc ) Eqn 2

where Qe is entry capacity, Qc is circulating flow and F and f are the intercept and slope coefficient
respectively and are calculated using the formulae below.

 1  
k = 1 − 0 .00347 (φ − 30 ) − 0.978   − 0.05 Eqn 3
 r  

F = 303x 2 Eqn 4

f = 0.21t D (1 + 0.2x 2 ) Eqn 5

where

1 + 0.5
tD =
(
1 + exp D − 60
10
) Eqn 6

v + (e − v )
x2 = Eqn 7
1 + 2S
where

v is the approach half width (m),


e is the entry width (m),
l is the average effective flare length (m) r is the entry radius (m),
D is the inscribed circle diameter (m),
φ is the entry angle in degrees and
S is the sharpness of the flare.

The second fundamental feature of ARCADY is the way in which flows from each arm on the roundabout are
combined.

The circulating flow is derived from the entry flow and turning proportions from previous arms.

The third fundamental feature of ARCADY is its time-dependent queuing theory which, because of the time-
varying nature of traffic flow, is used to calculate queue lengths and delays.
Time-dependent queuing theory treats all regions of traffic intensity by means of a technique based on
probabilistic theory.

The program operates by first dividing the total time for which the capacity and queue length / delay
calculations are needed into a number of smaller, equal-time intervals or segments.

During each time segment, demand flows and capacities are assumed to be constant.

These segments are then considered sequentially.

In calculating the capacities, queues and delays at a roundabout, the program treats the entry arms as a set
of inter-connected junctions.

The procedure follows the following steps


(i) the slope and intercept are calculated for each entry (referring to relationship between Qe
and Qc )
(ii) the profile of the demand flows are derived for each time segment where the traffic flow is
input as turning flows in vehs per hour with heavy goods vehicles taken into consideration
as a percentage
(iii) the entry flow, circulating flow, entry capacity and final queue-length are calculated for each
arm in each time segment
(iv) the geometric delays incurred by traffic are calculated
(v) the final queue lengths in one time segment are taken as the initial queue length of the
next.

One of the outputs from ARCADY is the ratio of flow to capacity (RFC) for each arm in each time segment.

Taking the RFC of the worst time segment will give an indication of the junction’s operating efficiency.

There is a 15% margin of error and therefore an RFC of 0.85 is assumed to indicate an arm operating at
capacity.

A value of RFC greater than this indicates the junction is operating over capacity and anything less is within
capacity.

Calculated queues and queuing delays are average values.

Queuing delays are given for each time segment in vehicle-minutes.

TRANSYT

TRANSYT searches for the optimum fixed time plan with which to co-ordinate all traffic signals within the
network to be modelled.
The main aim is to search for a set of signal timings which, when applied to the network of signal controlled
junctions, minimises the stops and delays of the vehicles at the stop lines.

TRANSYT has two main elements:


• the traffic model and
• the signal optimiser.

The model predicts a value of Performance Index (PI) for the network for any fixed-time plan and set of
flows.

This is a measure of the overall cost of traffic congestion and is usually a weighted combination of the total
amount of delay and the number of stops experienced by traffic.
The optimisation process adjusts the signal timings and checks, using the model, whether the adjustments
reduce the PI or not.

Cycle time is not automatically considered by the signal optimiser but a separate program called CYOP can
be used to evaluate the range of cycle times from which the optimum can be selected.

In order to compare TRANSYT and ARCADY results the most relevant output from TRANSYT is the degree of
saturation (DOS).

DOS = ratio of arrival rate to saturation flow-rate for a particular traffic stream and is normally expressed as
a percentage, taking into consideration the effective green time.

The full equation for DOS is as follows:

Total flow on link * cycle time * 100 (%)


DOS = Eqn 8
Saturation flow * effective green time
TRANSYT uses a value of 100 to represent over-saturation and under these conditions a queue will grow for
as long as the specified flow conditions exist.

However, signal modelling is assumed to be 90% accurate so for this analysis a DOS of 90% will be used to
represent capacity level.

TRANSYT has been proven suitable for modelling traffic behaviour on a signalised roundabout and capable
of producing timings that provide good co-ordination and avoid locking up of the junction.

By using various facilities available in the TRANSYT program such as link weighting, the shared link facility
and queue penalties, the program can be used to model traffic behaviour on the roundabout, control queue
lengths and find signal timings such that queues do not form to block upstream junctions.
One disadvantage of the TRANSYT model is that vehicles arriving at the stop line are assumed to have an
uninterrupted passage to the stop line.

One of the simplifying assumptions within TRANSYT is that it queues all vehicles in a vertical stack on the
stop line and assumes traffic can move freely once it receives a green light.

By weighting the circulating carriageway links TRANSYT can give the maximum amount of green time to the
circulating links and penalise the entry links.

This will ensure that circulating traffic will not block up on the roundabout.

Queue limiting can also be used whereby a maximum queue length can be specified e.g. 6 passenger car
units and a penalty input so that if the link exceeds the queue length it will have a significant effect on the
performance index.

Weighting and queue limiting are used in the following case studies.

These measures are signal changes not physical changes and if having done an analysis these measures do
not improve a particular over- capacity situation, at that point in the analysis physical changes to the
network may need to be considered.

RESULTS

Four different roundabouts were examined and analysed for operation under

• priority control (without signalisation) and


• with signalisation applied.

Each roundabout needs to be considered separately in order to determine if it can be signalised or not.

Two tables of results will be presented for each roundabout

(i) the first table will be the ARCADY results for the unsignalised situation in which case the
RFC values should be less than 0.85 to be within capacity while the
(ii) second table shows the TRANSYT results for the signalised scenario showing the results as
DOS.

Bandon Road Roundabout Case Study

The Bandon Road roundabout is an existing at-grade roundabout on the N25 (major road), a schematic
diagram of which is presented in Figure 1.

Figure 1. Bandon Road Roundabout

The N25 is to be upgraded and with the growth predicted within Cork city, the existing at-grade
arrangement will not have the capacity to cater for future flows.

This junction will be upgraded to a grade-separated roundabout and this is what will be analysed here.
As can be seen from Table 1, arms B and C have much higher approach widths, 7.3 m and 9 m respectively
than the others.

The circulating carriageway is 3 lanes wide.

The ICD of this roundabout is 113 m and is the second largest diameter roundabout examined here.

Table 1. Characteristics of Bandon Road roundabout


Arm v (m) e (m) l (m) r (m) D (m) φ (degs)
A 4.5 10.0 25.0 50 113 36
B 7.3 10.3 15.0 30 113 46
C 9.0 9 0.0 30 100 53
D 3.5 7 6.6 20 113 51
E 3.5 7.5 30.0 25 113 40
The evening peak hour traffic flows have been identified as the worst case scenario and it is for this period
the analysis is conducted on the flows expected in 2024.

Demands and turning proportions were calculated from count data and extrapolated to that year.

Table 2 shows the turning movements

Table 2. Turning movements for Bandon Rd roundabout


From / To A B C D E
(veh/hr)
A 0 557 347 250 0
B 555 0 1001 10 0
C 517 542 0 779 0
D 10 10 10 0 0
E 205 0 602 10 0

Table 3 shows the queue and delay information for the unsignalised situation for the 15 minute time
segment demonstrating the worst case scenario in terms of RFC.

Table 3. Queue and delay information for Bandon Road roundabout for peak 15 min time
segment

Arm Demand Capacity RFC Queue Delay


(veh/min) (veh/min) (vehs) (veh.min/
time segment)
A 21.1 28.75 0.73 2.7 40.0
B 28.63 29.95 0.96 15.0 206.0
C 33.6 36.72 0.92 9.6 137.3
D 0.55 12.72 0.04 0.0 0.7
E 14.94 17.15 0.87 6.0 85.0

There is a significantly high turning movement of 1001 veh/hr from B to C but high turning movements also
exist for C to D (779) and E to C (602).

As can be seen from Table 3, arm B has a RFC of 0.96, a queue length of 15 vehs and a delay of 206
veh.min/time segment.

This is the highest RFC value noted for any arm during any time segment in the period from 16:45 – 18:15.

Arms C and E are also over capacity with RFCs of 0.92 and 0.87.

Although they are over capacity, they are not causing a major problem at the junction except for the
proximity of a signal junction 65 m south of the junction on arm C.

If queues on arm C of the roundabout build up, they are likely to cause blocking back at the signalled
junction to the south of the roundabout.

At this point in the analysis, signalisation was introduced on all entries except D.

A cycle time of 60 secs was chosen to start with and the first set of results produced; the DOS results are
presented in Table 4.
Table 4. Signalled Bandon Road roundabout results
Arm Link DOS (%) DOS (%) DOS (%) DOS (%)
C=60sec C=90sec Queue Final
limiting results
applied
A 11 43 34 45 28
12 102 151 190 63
B 21 94 185 236 94
22 37 32 39 37
C 31 31 32 30 31
32 14 15 14 29
33 121 121 117 98
D 41 2 2 2 2
E 51 105 113 128 58
52 75 81 92 41

At this stage the results were put through the CYOP programme which evaluated the optimum cycle time
and the results suggested that the cycle time was too low so the test was run again for a cycle time of 90
secs as any higher would have been outside the guidelines for signalisation of roundabouts.

This had the effect of increasing the DOS significantly.

The next run included for queue limiting i.e. the maximum number of passenger car units allowed to queue
on a link) and severe penalties were applied to links which exceeded specified queue limits.

The aim of this process is to allow traffic to flow freely on the roundabout itself. This is necessary to reduce
the DOS on the circulating carriageway.

At this point, 4 of the 10 links were still causing concern; 12, 21, 33 and 51.

Double cycling was tried at this point i.e. where some entries would have a cycle time double that of the
network.

This did not have the desired effect of reducing the DOS on the 4 links.

What did have an effect was for some of the nodes to have a cycle time half that of the network.

This improved links 12, 33 and 51.

Once all the amendments above had been made i.e. changes to signal settings, the next step is to consider
geometric changes so that saturation flows are increased.

To reduce the impact of the constraint of the signal junction so close to the roundabout on arm C, a
dedicated left turning lane was provided (CD – high turning movement 779).

To finally bring the junction arrangement within capacity (final set of results are presented in Table 4) two
final measures were considered:

• to reduce the radius of turn on each of the offending links


• and/or widen the carriageway at the stop line.
Most of the lanes are 3.5 – 4 m and to widen any further would run the risk of the two lines of vehicles
attempting to queue in one lane and so this strategy was not attempted.

The focus then was on the reduction/removal of the entry deflection, effectively straighte ning the links.

Hill Street Roundabout Case Study

The Hill St Roundabout differs from the Bandon Rd roundabout in that it is smaller at 75m ICD, is at-grade
and has a two-lane circulating carriageway.

Three future traffic scenarios (low, medium and high) were developed for the design year of 2019 (the
medium case is used in this analysis as it is thought to be the most likely of the three).

Two different shaped roundabouts were tested; a conventional circular arrangement and an oval shaped
roundabout, to see which would be more useful should the roundabout be signalised at a later date.

The circular version is shown in Figure 2 and the oval shape in Figure 3.

Figure 2. Circular version of Hill St Roundabout


Figure 3. Oval version of Hill St Roundabout

The characteristics of the Hill St Roundabout are presented in Table 5 and the turning movements are
presented in Table 6.

Table 5. Characteristics of Hill St Roundabout


Arm v (m) e (m) l (m) r (m) D (m) φ (degs)
A 7.0 11.0 8.7 20 75 48
B 4.5 10.0 42.0 20 75 35
C 4.5 9.5 31.0 20 75 43
D 7.0 10.0 8.6 30 75 36
E 7.5 11.0 10.3 50 75 35

Table 6. Turning movements for Hill St Roundabout

From /To A B C D E
(veh/hr)
A 0 127 171 432 622
B 284 0 0 45 439
C 285 0 0 68 92
D 451 91 22 0 35
E 269 355 167 126 0

The highest turning movement is from A to E at 622 veh/hr but movements from A to D, B to E and D to A
are also relatively high in the range of 432 – 451 veh/hr.

In this situation, as in the first roundabout examined, the evening peak period is considered to be the critical
period and therefore the roundabout was modelled for 15 minute segments within the 16:45 – 18:15 period.
The results of the time segment showing the highest RFC ratios are presented in Table 7.
Table 7. Queue and delay information for Hill St roundabout for peak 15 min time segment
Arm Demand Capacity RFC Queue Delay
(veh/min) (veh/min) (vehs) (veh.min/
time segment)
A 24.72 28.96 0.85 5.5 81.6
B 14.04 20.84 0.67 2.0 30.1
C 8.13 14.25 0.57 1.3 19.4
D 10.95 19.21 0.57 1.3 19.5
E 16.76 29.02 0.58 1.4 20.3

From the table, it can be seen that A is showing the highest RFC value of 0.85 followed by B at 0.67 with the
three remaining arms having RFC values of around 0.57.

A similar analysis was done for the oval shaped roundabout but the RFC values were 1.56, 1.06, 1.08, 1.06
and 0.55 for arms A, B, C, D and E respectively. The oval shape therefore performed worse than for the
circular scenario.

However, should traffic grow to a higher level in the future and signalisation is required, further analysis
was done to see which shape would perform best in this higher traffic scenario.

Signalisation was introduced at this point.

In the circular arrangement, only one arm A is over capacity (capacity, RFC = 0.85).

Therefore as a first step, this arm alone is signalised. However, as can be seen in Table 8, this had the
effect of putting another two lanes over capacity, those on Arm D, and it also put Arm C very close to
capacity.

Table 8. Signalled Hill St circular and oval roundabout results

Arm Link DOS (%) DOS (%) DOS (%) DOS (%) DOS (%) DOS (%)
Circular Circular Circular Circular Oval Oval
C=60sec C=60sec C=120sec C=120sec C=60sec C=30sec
A A, C, D A and C A A, C, D A, C, D
signalised signalised signalised signalised signalised signalised
A 11 129 129 132 131 356 178
12 83 78 80 79 219 110
B 21 44 44 44 44 38 41
C 31 88 91 93 28 82 37
D 41 98 297 54 54 267 133
42 98 276 54 54 247 124
E 51 58 51 57 57 52 51
52 49 44 49 48 44 43

At this point, signalisation was also introduced to C and D as well as on A to try to reduce the over capacity
problems.

The results for this test are presented in Table 8 but the problem has become even more exacerbated on D.

At this point the cycle time was evaluated and a longer cycle time of 90 sec was tested.

This did not improve the situation so a cycle time of 120 secs was tried in the case for A and C arms
signalised and in the case for only arm A signalised.
The latter had the best set of results overall as can be seen from Table 8, although there is still a
significantly high DOS on A.

Some of the arms could be flared even more but the two problem arms are already two lanes wide at entry
and with a two-lane circulating carriageway, entries cannot be widened to three lanes.

The oval arrangement did not prove much better because it is much narrower on the East-West axis so it is
not possible to achieve circulating links with adequate capacity while ensuring the entry links also stay within
capacity.

The two tests giving the best results for the tests on the oval shape are presented in Table 8 with A, C and
D signalised for 60 sec and 30 sec cycle times.

The latter gives the better set of results but even so there are significant problems on A and D, much worse
than in the circular case.

The main conclusion from this case study is that if the demand was higher than the medium growth scenario
presented here then the better shape in terms of efficiency is circular.

It can also be concluded that the ideal geometric arrangement suitable for signalisation would be of larger
dimensions.

The final conclusion is that signalisation of the existing arrangement would be unlikely to generate a solution
where all arms are within acceptable DOS levels without resorting to changes to the geometry of the
roundabout.

Carlow Inner Relief Road Case Study

Two roundabouts in close proximity form the basis of the next case study (see Figure 4).

Figure 4 Carlow Inner Relief roundabout layouts

They form part of a double / dumbbell arrangement, one with an ICD of 35 m and circular in shape, the
other is a five arm oval shaped layout.

40 m separates the two junctions and both roundabouts are tested for the 2010 pm peak hour.

The junctions, the Thompson and Paupish roundabouts, are first analysed in a normal priority type situation
and then with signals installed.
The characteristics of both junctions are presented in Table 9, the turning movements are in Table 10 and
the RFC values for the time segment during which conditions are most critical in the evening pm peak are
presented in Table 11.

Table 9. Characteristics of Paupish and Thompson roundabouts

Paupish
Arm v (m) e (m) l (m) r (m) D (m) φ (degs)
A 2.92 3.56 7.5 15.0 29.3 44.0
B 3.3 4.0 4.2 7.5 28.0 6.0
C 3.5 7.0 9.7 51.4 28.0 8.4
D 3.3 3.3 0.0 11.5 28.0 33.0
E 3.5 4.2 2.3 10.0 34.0 9.3
Thompson
A 3.5 4.1 2.0 35.0 35 19.0
B 5.7 5.7 0.0 30.0 35 30.0
C 3.5 5.0 3.0 27.5 35 23.0

Table 10. Turning movements for Paupish and Thompson roundabouts

Paupish
From /To A B C D E
(veh/hr)
A 0 0 4 79 215
B 0 0 1 35 122
C 7 7 0 21 525
D 28 47 63 0 111
E 157 210 802 78 0
Thompson
From /To A B C
(veh/hr)
A 0 192 781
B 252 0 378
C 1095 288 0

Table 11. Queue and delay information for Paupish and Thompson roundabouts for peak 15
min time segment
Paupish
Arm Demand Capacity RFC Queue Delay
(veh/min) (veh/min) (vehs) (veh.min/
time segment)
A 5.45 6.58 0.83 4.2 60.5
B 2.89 7.32 0.39 0.6 9.6
C 10.24 19.56 0.52 1.1 16.2
D 4.55 6.58 0.69 2.1 31.2
E 22.8 16.56 1.38 224.6 2667.2
Thompson
A 17.79 16.51 1.08 50.7 607.9
B 11.52 17.79 0.65 1.8 26.7
C 25.28 16.76 1.51 321.1 3857.6

The internal link between the two roundabouts i.e. A on the Thompson roundabout and E on the Paupish
roundabout have RFCs of 1.08 and 1.38 respectively.
Looking at the internal link between the two roundabouts, which is a single carriageway, the link was flared
more at both entries to eliminate the significant capacity problem.

The extent of the flaring was such that it meant the link should in fact be four lanes wide or a dual
carriageway.

Once the link was made into a four-lane carriageway, capacity on the link was no longer an issue.

Having sorted the problem on the internal link, the other arm, C, on the Thomson roundabout experiencing
a significantly high RFC of 1.51 was examined to see if signalisation would help to reduce the problem.

In order to control the flow and attempt to gain additional capacity, it was decided to signalise the problem
arm only initially.

In the analysis the Paupish roundabout was considered as one priority node separate from the Thomson
roundabout and the interaction between them was kept in mind.

Given the very high RFC value, the problem arm was first widened to two lanes so that the link could cater
for the prevailing excessive queuing.

When this was complete, the RFC of arm C was reduced to 0.91.

Signalisation of this arm resulted in the internal link between the two roundabouts again going over capacity
and therefore the internal link, A, was also signalised.

With the two arms signalised and for a cycle time of 90 secs, this finally meant that all arms on the Thomson
roundabout had DOS values of less than 84%.

Signalisation in this case proved useful in reducing over capacity on the links but alterations to the geometry
of the links were also necessary when RFC values were excessive.

It is thought that although the Thomson roundabout is considered small for signalisation the fact that it is a
three-arm junction with very simple traffic movements enabled signalisation to work well in this case.

The proposal therefore for this roundabout is to build the two roundabouts but with additional lanes on the
internal link and on arm C of the Thomson roundabout and when traffic levels rise signals can be added to
this existing geometry at a later date with good effect.

If the links were not widened as proposed here, it is likely there would be considerable problems in the
future, which signalisation alone would not be capable of solving.

Southern Link Road Case Study

This case study of the Southern Link Road (SLR) involves analysis of a grade-separated roundabout. The
schematic diagram of the roundabout is shown in Figure 5.
Figure 5 Southern Link Road roundabout

Arm D represents the slip road up and slip road down to the main carriageway running west – east and E
represents the same on the other side.

A is a link towards a town centre and C is the link from an industrial estate.

The geometric details of the roundabout are given in Table 12.

Table 12. Characteristics of SLR roundabout


Arm v (m) e (m) l (m) r (m) D (m) φ (degs)
A 7.6 9.5 30.0 35.0 116 40.0
B 6.0 6.0 0.0 290.0 116 15.0
C 7.6 9.0 20.0 42.0 116 28.0
D 7.5 7.5 0.0 45.0 116 27.0

Because of the attractor of the town on arm A in the morning peak period and also the potentially large
numbers of vehicles from the industrial estate on arm C in the evening peak, both the am peak and pm peak
are examined in this case study.

The turning movements for the am peak are presented in Table 13 and the corresponding RFC values are
presented in Table 14.

Table 13. Turning movements for SLR roundabout in am peak period


From /To A B C D
(veh/hr)
A 0 197 794 728
B 219 0 50 0
C 538 49 0 153
D 1205 0 376 0
Table 14. Queue and delay information for SLR roundabout for peak 15 min time segment in
am peak period
Arm Demand Capacity RFC Queue Delay
(veh/min) (veh/min) (vehs) (veh.min/
time segment)
A 31.42 44.45 0.71 2.4 35.6
B 4.92 8.88 0.55 1.2 17.9
C 13.53 32.97 0.41 0.7 10.4
D 28.9 28.38 1.02 37.6 468.5

The priority situation, as shown in Table 14, suggests that there would be significant queuing (38 vehicles)
on arm D heading for A and that the queuing would cause blocking back to the main carriageway below.

One proposal to solve this is to provide a dedicated left turning slip lane from D to A.

The roundabout operates within capacity for the majority of the day with the exception of the evening peak
period when there is an exodus of vehicles from the industrial estate (arm C).

This problem is given some consideration in Tables 15 and 16.

Table 15. Turning movements for SLR roundabout in pm peak period


From /To A B C D
(veh/hr)
A 0 219 538 1205
B 197 0 49 0
C 794 50 0 376
D 0 0 153 0

Table 16. Queue and delay information for SLR roundabout for peak 15 min time segment in
pm peak period
Arm Demand Capacity RFC Queue Delay
(veh/min) (veh/min) (vehs) (veh.min/
time segment)
A 35.87 49.48 0.73 2.6 39.0
B 4.5 8.68 0.52 1.1 15.7
C 22.3 25.21 0.89 7.0 101.3
D 2.8 24.24 0.12 0.1 1.9

The RFC values are presented in Table 16 and there it can be seen that arm C is just over capacity at 0.89.

Signalisation is considered in this case study to see if the need to provide the additional left turning lane
from D to A can be removed in favour of signalisation.

Part-time signals may be an option in this case for the morning peak period and this is examined here. The
cycle time for the network was taken to be 90secs and all arms were signalised in the first run.

The D arm at this point had a DOS of 108% with the DOS on all other arms significantly lower and below
capacity.

A cycle time of 60 secs was then tested but this also gave a DOS of 108% for arm D.

At this point, given that the DOS on D was still high, it was time to consider geometric changes.
Flaring to increase the number of lanes to two at entry seems to be the only alternative and this means
ensuring that when the roundabout is constructed that the slip road from D is sufficiently wide to split it into
two lanes when signalisation is required.

DISCUSSION
Lines and Crabtree (1986) demonstrated that TRANSYT, when used with appropriate care is suitable for
modelling traffic behaviour on signalised roundabouts and is capable of producing good coordination and
avoiding traffic locking up on the roundabout.

However, from this analysis it can be seen that when signalisation is used an increase in capacity is not
guaranteed.

Signalisation needs to go hand in hand with geometric improvements such as extra lanes on circulating
carriageways and flaring of entries to increase the density of traffic at stop lines.

As can be seen from the case studies examined above, not all roundabouts benefit from signalisation.

Generally it can be seen from the case studies presented here that larger roundabouts do benefit from
signalisation and this appears to be due to the fact that more stacking space is available at the stop lines.

Having said that, the Carlow Inner Relief road also performed well although its ICD was only 35m, a
diameter considered perhaps too small to be signalised.

The success in this case may be due to the non-excessive flows and that the whole junction was simple in
terms of layout.

On the basis of this research it is suggested that, given the range of characteristics that a roundabout can
have and the different balances in flows on entry arms, each roundabout should be considered individually.
The tests would involve examining whether in fact signalling will be useful at all, whether signalisation on its
own provides benefits, whether signalisation in conjunction with geometric changes is the answer or
whether geometric changes alone are required.

During the design phase of a roundabout it would be important to consider the potential solutions in the
future if traffic levels increase so that the optimum physical characteristics can be introduced during the
construction phase which may help when/if signalisation is required and found to be useful at a later date.

REFERENCES
Alcelik, R. (1998), ‘Roundabouts, Capacity and Performance Analysis’ Research Report ARR 321. ARRB
Transport Research Ltd. Vermont St., Australia.

Alcelik, R. and Besley, M. (2001). ‘Microsimulation and analytical methods for modelling urban traffic’.
Proceedings of Conference on Advanced Modelling Techniques and Quality of Service in Highway Capacity
Analysis, Trucker, California, July.

Binning, J.C. (2000). ‘ARCADY 5 User Guide’. Transport Research Laboratory, Crowthorne, Berkshire, UK.

Binning, J.C. and Crabtree, M.R. (1999). ‘TRANSYT 11 User Guide’. Transport Research Laboratory,
Crowthorne, Berkshire, UK.

Brown, M. (1995). ‘The Design of Roundabouts – State of the Art Review’, Transport Research Laboratory,
Crowthorne, Berkshire, UK.

Jacquesmart, G. (1998). ‘Modern Roundabout Practice in the US’, NCHRP, Transportation Research Board,
National Academy of Sciences, US.
Kimber, R.M. (1980). ‘The Traffic Capacity of Roundabouts’, LR942, Transport Research Laboratory,
Crowthorne, Berkshire, UK.

Kimber, R.M., McDonald, M. and Hounsell, N.B. (1986). T ‘ he Prediction of Saturation Flows for Road
Junctions Controlled by Traffic Signals’, RR67, Transport Research Laboratory, Crowthorne, Berkshire, UK.

Lines, C.J. and Crabtree, M.R. (1990). ‘The Use of TRANSYT at Signalled Roundabouts’, RR274, Transport
Research Laboratory, Crowthorne, Berkshire, UK.

Omari, B.H., Al-Masaeid, R. and Shawabkah, Y.S. ‘Development of a Delay Model for Roundabouts in
Jordan’, Journal of Transportation Engineering, ASCE, 130, 1, 2004, 76-82.

Russell, E.R., Luttrell, G and Rys, M. (2002), ‘Roundabout Studies in Kansas’, Proceedings of the 4th
Transportation Speciality Conference of the Canadian Society for Civil Engineering, CSCE, Quebec, June.

Sisiopiku, V. P. and Oh, H. (2001), ‘Evaluation of Roundabout Performance using SIDRA’, Journal of
Transportation Engineering, ASCE, 127, 2, 2001, 143-150.

Tan, J. (2001), ‘Comparison of Capacity Between Roundabout Design and Signalised Junction Design’,
Proceedings of 1st Swiss Transportation Research Conference, Monte Verita/Ascona, March.

Вам также может понравиться