CAPACITY AND QUEUE MODELLING IN UN-SIGNALISED

ROUNDABOUTS

R. Camus, M. Dall’Acqua, G. Longo

Department of Civil Engineering
University of Trieste
Piazzale Europa,1 – 34127 Trieste

The paper focuses on an approach to evaluate roundabouts performance. It

refers to the facility as a succession of merging, weaving and diverging
segments and allows estimating capacity and delays. First results of this
application are discussed as regards mainly capacity and delays.

1. INTRODUCTION

The paper deals with the problem of un-signalised traffic roundabout’s

performances evaluation, mainly focusing on capacity and Level of Service
(LOS) attributes. Many different capacity models have been developed and
presented in the scientific literature, but the existing methods do not lead to
homogeneous results, mainly under peak traffic conditions. These models
mainly analyse a roundabout facility as a simple succession of some stop or
yield controlled independent intersections. For every one of these junctions,
capacity and Level of Service evaluation could be developed at a macroscopic
or at a microscopic level. These two main approaches are very different from
both a methodological and an operational point of view; at the time, neither
the macroscopic approach nor the microscopic one has never completely
guaranteed a satisfied level of agreement with data recorded in the field.
Different models give different results, both under light and congested traffic
conditions. So, a universal methodology that could well perform under
different traffic or environmental conditions seems not to be really available.
In this context, this paper presents a different approach to analyse roundabout
performances. In particular, a roundabout facility could be studied as a
succession of merging, weaving and diverging areas instead of a simple
succession of some stop or yield controlled independent intersections. Of
course some of this operations could be neglected, depending mainly on the
facility size and vehicles interactions.
The most important targets of this paper are the followings:
§ To develop a general framework for roundabout capacity and LOS
estimation;
§ To propose a methodological procedure that could be set on a median
level between a macroscopic and a microscopic approach and that
could include (on the basis of a data based parameters calibration)
both the two main existing approaches;
§ To apply a mesoscopic model for merging areas that could be able to
analyse vehicle interactions and quantify entry capacity and queues
dynamics.
The paper is organised as follows: the next section briefly describes major
basic sets for the roundabout performance evaluation and reports also a brief

© Association for European Transport 2004

review on the Statistic and Probabilistic approaches. Then, the third section
presents the main features of the proposed approach. Some applications to a
case study are illustrated and main results are discussed in the fourth section.
Finally some conclusions are drawn.

2. BACKGROUND

Modern roundabouts usually operate under the off-side priority rule; in this
case, the entry capacity for an approach leg depends on the circulating flow.
The relationship between entry and circulating flows could be managed at a
macroscopic level or on a microscopic one; to the first methodologies refers
the so called Statistical methods, to the second one the so called Probabilistic
ones.
Main differences on these two approaches are widely described and
discussed on [3], and here are briefly recalled.

2.1. Statistical approach

This approach refers to the interactions between traffic streams (entry and
circulating streams, sometimes also diverging streams) by regression analysis
on a wide data recorded set.
The entry-circulating relationship is expressed on the basis of the geometrical
parameters of the facility. Main fall of this approach is that the capacity
formulation (and also the delays formulation) could not be easily generalised.
As an example of the Statistical method, the entry capacity model developed
in Germany is reported below.
Let
K2 be the entry capacity for the minor stream (vph);
Qc be the circulating flow (vph),
entry capacity relationship follows:
K 2 = C + D ⋅ QC (vph);
where C and D could be chosen according to the following table.
Entry/Circulating lanes N° C D
1/1 1218 -0,74
1/2 or 1/3 1250 -0,53
2/2 1380 -0,5
2/3 1409 -0,42
Table 1 – Coefficients in the German model.

In the following picture the entry-circulating relationship is described in the

case of 1 entry lane and 1 circulating lane roundabout.

© Association for European Transport 2004

 1400

ONRC=1218
1200
Minor flow entry capacity (vph)

1000

800

600

400

200
SC=1646
0
0 200 400 600 800 1000 1200 1400 1600 1800
Major flow (vph)

Figure 1 – German linear model.

As reported in the figure, SC represents the mainline capacity (no inflow

entering from the approach leg); on the other hand, ONRC represents the
entry capacity highest limitation (when there is a permanent queue on the
approach leg and major flow is null). Entry capacity could not overcome this
limitation (that depends only on the geometry of the merge area); usually entry
capacity is reduced by the circulating flow: in particular, when the major flow
reaches its maximum value (SC), entry capacity is reduced to zero.

2.2. Probabilistic approach

Probabilistic methods refer to the interaction between single vehicles,

following the gap-acceptance theory.
This approach is more structured from a theoretical point of view, and it could
manage also a lane-by-lane analysis; on the other hand, it needs the
calibration of the headway's distribution formulation (for both the major and
the minor approach) and of the microscopic parameters. This calibration
procedure is a very troublesome task, mainly for the extremely variable traffic
flow conditions a roundabout could manage.
As an example of this approach, the entry capacity model developed in
Germany following the Gap-Acceptance theory is reported below.
Let
K2 be the entry capacity for the minor stream (vps);
Qc be the circulating flow, (vps);
T ,To, ∆, to be microscopic parameters (s);
nc, ne be the circulating and entry lanes number,
following equation and parameters values hold:
Q n
K 2 = ((1 − ∆ ⋅ c ) nc ) ⋅ ( e ) ⋅ e −QC ⋅( t0 −∆ ) (vps).
nc T0

© Association for European Transport 2004

 Microscopic parameters (s) Values
Critical Gap (T) 4.12
Follow-up time (To) 2.88
Mainline minimum headway (?) 2.10
to=T-To/2 2.68
Table 2 – Coefficients in the German model.

The following picture reports the entry-circulating relationship in the case of 1

entry lane and 1 circulating lane roundabout, relating to parameters calibrated
in German roundabouts.

1400
ONRC=1250
1200
Minor flow entry capacity (vph)

1000

800

600

400

200
SC=1714
0
0 200 400 600 800 1000 1200 1400 1600 1800
Major flow (vph)

Figure 2 – German exponential model.

2.3. Performance analysis

Usually roundabout's performance is evaluated using three basic measures,

which are:
§ capacity estimation when all approach legs are congested;
§ capacity estimation far away from the saturation in each approach leg;
§ the Level of Service (LOS), (delays and queue lengths).
The first represents the ability of the roundabout to process traffic under heavy
traffic conditions, when all approach legs have queues; it represents the
summation of entry lane capacity over all approach lanes on all legs.
Geometrical and behavioural considerations aside, this measure depends
mainly on the Origin/Destination flow pattern.
In a more general situation, some approach legs may be no saturated; in this
case, the flow approaching the facility (that is the demand flow) is lower than
the entry capacity, that is defined in [8] as “the maximum rate at which
vehicles can reasonably be expected to enter the roundabout from an
approach during a given time period under prevailing traffic and geometric
conditions”. This one (maximum entry flow under no congested conditions) is
the second measure of performance characteristics for a roundabout.
Delays and queues suffered by each lane on each roundabout’s leg are the
third basic set useful to performances analysis. The second and the third set

© Association for European Transport 2004

of measures depend both on the O/D flow pattern and on the total entry flow
for each leg (and so on circulating flows around the central island).
In the congested roundabouts legs, entry capacity calculation requires an
iterative process and so the solution is not so simple; in an under-saturated
case, the solution is immediate applying the capacity formula.
As it can be argued from these observations, entry capacity for each
roundabout leg is not a fixed value; of course, it depends on the geometrical
features of the merge area, but it depends also on the circulating flow and on
the entering traffic flow condition.
Many models have introduce some coefficients to correlate in the best way
the entry capacity to the circulating flow; in some cases these coefficients
depend on geometrical features of the merge area, in some others on entry
and circulating lanes number, in some others on vehicles interactions: as a
consequence, there is not a common and wide accepted criterion to the entry
capacity formulation.
In the next section, first steps of a different approach are reported. Entry
capacity is correlated to the circulating flow on the basis of macroscopic
parameters; the main idea of this research study is that these parameters
depend only on the maximum flow that the circulating lanes and the entry
lanes could manage (in other words, their capacity when there are
respectively no entry flows and no circulating flows).
In such a way, the entry-circulating relationship could be managed referring to
the physical capacity of the conflicting streams; this approach has some
advantages, which are:
§ capacity as a traffic parameter is well-known and easy to calibrate;
§ capacity limitations are rightly present in the model and clearly visible;
§ these ones could be the bases to a universally accepted procedure to
approach entry capacity estimation.

3. THE PROPOSED APPROACH

Even if each roundabout facility could be divided into input, weaving and
output segments, the critical element seems to be the input one where
different traffic streams compete for space, then influencing the facility
performance. Moreover this element is always present while the other ones
could be neglected for example in small roundabouts. For this reason, in this
first step, the methodology focuses mainly on merging areas. In particular the
model presented in [2] has been used to describe traffic dynamics on
roundabout merging areas. This model has been originally developed to
describe the behaviour of on-ramp lanes of a highway facility and it may be
considered as a mesoscopic model because it falls in the middle between the
macroscopic approach of Highway Capacity Manual [8] and the microscopic
approach proposed by the Gap-Acceptance theory.
The idea consists in using the same methodology to the evaluation of the
entry capacity and queuing dynamics for each roundabout leg.
Referring to [2] for the complete description of the model, here some features
are briefly recalled.

3.1. Main definitions

© Association for European Transport 2004

According to HCM procedure, traffic streams are analysed referring to a time
step t, usually a fraction of hour to model peak phenomena.
Let SD(t) be the system demand and MI(t) the mainline input, ONRD(t) be the
on-ramp demand and ONRI(t) be the minor inflow (on-ramp), SC be the
mainline capacity when minor flow is null and ONRC be the minor stream
capacity when the major is absent, MO1(t) be the maximum flow assigned to
the major stream due to the presence of the minor inflow and ONRO(t) be the
maximum flow assigned to the minor stream due to the conflicting main flow.
Moreover, the mainline capacity does not change across the merge area and
of course SD(t)= SC and ONRD(t)= ONRC. Of course SC and ONRC depend
upon geometric and functional features of the roundabout and they could be
estimated according to HCM procedures.
In the roundabout case, mainline flow is the circulating one, whilst the on-
ramp flow is the entry one on each approach leg.

3.2. Absolute priority to the major flow: first model elaboration

The mainline capacity within the merging area could be divided into the
capacity assigned to the two conflicting flows; the portion of the major flow
capacity assigned to the minor flow could be quantified by a single
macroscopic parameter, β, as follows:
ONRO (t ) = β ⋅ SC (t )
where β is a macroscopic parameter that depends on major flow volume and
on the hypothesis about the yield relationship between the major and minor
flows. Furthermore, the entry capacity ONRO(t) could be quantified as follows:
ONRO(t ) = MIN ( β ⋅ SC (t ), ONRC (t ))
In the hypothesis of absolute yield to the major flow, only the minor stream
could be involved in queue: if the entry flow is smaller than the entry capacity,
then the minor stream merging into the roundabout could be the entire entry
demand flow (no queues). On the other hand, if the minor inflow is higher than
the entry capacity, approaching vehicles will wait before merging into the
major flow: in such a case, the demand flow could not merge entirely into the
roundabout, and the outflow pattern from the roundabout leg is limited by
entry capacity. In the next time interval, vehicles that could not cross over the
yield line have to be added to the next minor inflow in order to dynamically
evaluate the queue evolution in the time.
Time interval length choice is an important task to perform an acceptable
description of conflicting flows behaviour in the merge area; a fraction of hour
is recommended, i.e. 1 minute could be sufficient to model peak traffic
conditions.
As an example, an exhibit of the function β(MI(t)) and of the entry capacity are
reported below.

© Association for European Transport 2004

 1,0

0,9
0,8

0,7
0,6
β (MI(t))

0,5
0,4
0,3

0,2
0,1
0,0
0 5 10 15 20 25 SC 30
MI (vpm)

Figure 3 – Proposed model, function β (MI(t)).

25,00

ONRC
20,00
ONRO (vpm)

15,00

10,00

5,00

0,00
0,0 5,0 10,0 15,0 20,0 25,0 SC 30,0
MI (vpm)

Figure 4 – Approach leg entry capacity ONRO(t).

The model presents a linear relationship between the minor flow entry
capacity and the major flow; this kind of relationship is not a new one, but
there are some original ideas which are:
§ there is a clear up limitation to the entry capacity;
§ mainline capacity has a peculiar impact on the merging area capacity
estimation; in fact to both the minor and the major flow is assigned a
portion of the mainline capacity;
§ entry capacity formulation needs only two parameters, SC and ONRC.
The β parameter gives the measure of the mainline capacity that could be
suitable to the approach leg of a roundabout. Furthermore, it is the only
parameter strictly dependent on the circulating flow; as a consequence, it
reflects vehicles behaviour in the merge area at a macroscopic level (as the
Gap Acceptance theory does by a microscopic level using many traffic
parameters).
The hypothesis of absolute priority to the major flow seems quite far from the
real traffic behaviour and the presented model overestimates entry capacity

© Association for European Transport 2004

prediction. For both these reasons, the first model elaboration is not a correct
approach to roundabout’s performance evaluation; in the next section the
model is refined as regards both the conflicting steams capacity estimation
and the introduction of the limited priority merging in the major flow.

3.3. Limited priority to the major flow: second model elaboration

The second model differs from the first one in the definition of MO1(t) and
ONRO(t), which are the portions of the mainline capacity assigned to the
major and minor flow respectively. According to the notations reported above,
the main idea of the model is the introduction of two time dependent
parameters, α and β: α=α(ONRI(t)), quantifies the mainline capacity reduction
due to the minor flow influence while β=β(MI(t)), quantifies the minor flow entry
capacity reduction due to the major flow influence.
The limited priority merge phenomenon could be modelled by the introduction
of the parameter βmin; this parameter could quantify the residual minor flow
entry capacity, even under high major flow volumes. The model needs the
definition of three parameters: the circulatory roadway and the approach leg
capacities (SC(t) and ONRC(t)) have to be calibrated, and then the βmin
parameter has to be assigned. If βmin=0, major flow benefits the absolute
priority.

1,0

0,8

0,6
β (MI(t))

0,4

0,2

0,0
0 5 10 15 20 25 SC=27,4 30
MI (vpm)

β min=0 β min=0.15 βmin=0.3

Figure 5 – Proposed model, second elaboration, β (MI(t)) function.

Noting with MF(t) and ONRF(t) the outflows from the merge area for the
mainline and the minor flow respectively, following equations hold:
ONRF(t) = MIN(β*ONRC(t), ONRI(t));
MF(t) = MIN(C - ONRF(t) ; ΜΙ(t)).
Queue dynamics are studied in the same way as in the first elaboration, but in
this case the problem is a little more complex because of the limited priority
assigned to the major flow: some queues could develop on the circulatory
roadway and the tail of the queue could reach the preceding merge area. Due
to the small distance between successive merge areas in roundabouts, queue

© Association for European Transport 2004

growth in the mainline could happen very easily: as a consequence,
roundabout capacity could fall down immediately.
This model presents the same benefits of the first one; in addition, some other
else which are:
§ limited priority merge could be modelled and so the facility could be
analysed as a multi-junction node;
§ capacity formulation needs only three macroscopic parameters;
§ the model could be easily adapted to the major flow absolute priority
condition;
§ the parameter β is strictly connected to vehicles behaviour, as
microscopic parameters do in the Gap Acceptance theory;
§ entry capacity prediction is very satisfactory.
On the other hand, the model could lightly overestimate the entry capacity
because of the limitation on βmin. After a comparison between the predictions
of the model and of the Gap Acceptance theory, it was found that reducing
βmin parameter values is equivalent to increase the Critical Gap value; in
particular, βmin=0 is equivalent to a three seconds Critical Gap. Vehicles
waiting to merge in the major flow could sometimes need higher Critical Gap
values, and that’s why it is necessary to extend the model to a more general
formulation; that is what follows in the next section.

3.4. Final model elaboration

First step of the model is the right assignment to the βmin parameter. Two
alternative conditions are the followings:
§ if βmin>0, second elaboration of the model has to be followed;
§ if βmin=0, a three parameters model have to be introduced.
These parameters are βmin, m and γ; the two new parameters are necessary to
reduce the β(t) function to the composition of two straight lines, intersecting at
the point A (XA;YA), as reported below.

βmin=0

Figure 6– Proposed model, β (MI(t)) function for βmin=0.

Coordinates of the point A depend on m and γ parameters, according to the

following equations: XA= γ∗C, YA=m* XA.
New formulations of the β (MI(t)) function for βmin=0 follow these equations:

© Association for European Transport 2004

β(MI(t))= MI(t)*(YA-1)/XA+1, for MI(t) < XA;
β(MI(t))=YA*(1-(X-XA)/(SC(t)-XA)) for XA= MI(t) =SC(t).
The graph below reports the complete model formulation for both the case of
βmin=0 and βmin>0.

1,0
0,9
0,8
0,7
0,6
β (MI(t))

0,5
0,4
0,3
A
0,2 A
0,1
0,0
0 5 10 15 20 25 SC 30
MI (vpm)

βmin=0,3 βmin=0 βmin=0 m=0,01 γ =0,6 βmin=0 m=0,02 γ=0,4

Figure 7– Proposed model, final elaboration, β (MI(t)) function.

The last model is the most complete and the careful calibration of the
parameters involved could guarantee good results on both entry capacity and
queue dynamics estimation.
If it is necessary to use the three parameters model, probably m and γ are the
most difficult parameters to be calibrated; accordingly to the idea that they
represent vehicles interactions, the reduction in m and γ values is equivalent
to reduce the minor flow entry capacity.
A very wide spectrum of traffic conditions could be covered by the model, both
in the approach leg and in the circulatory roadway. In particular, it is very easy
to quantify queue lengths, also if there are over saturation conditions during
the time interval set. This is a very important point, because many queuing
and delay models developed for un-signalised intersections (and for
roundabouts), could not be used when entry demand flow is higher than the
entry capacity (saturation index >1). On the contrary, the proposed model
could manage both some overflow condition both on the roundabout’s legs
and on the circulatory roadway. Furthermore, the model could also measure
the global capacity of the roundabout when all approach legs present queues.
As a consequence, the model is able to give the well-known performance
measures usually applied to the roundabouts evaluation case.
Now the problem is: which is the relationship between the results of the
proposed model and the output of the existing ones? In other words, is the
proposed model able to reproduce the results of the existing models simply by
useful calibration of its parameters?
In order to give an answer to the previous questions, the proposed model (all
the three cases) and two existing approaches (one belonging to the statistical
methods and the other belonging to the probabilistic ones) have been applied
in order to estimate capacity of a real case study in Italy. So in the following

© Association for European Transport 2004

sections these applications to a single lane roundabout will be described.
Results are given for all the three elaborations of the model. Entry capacity
and queuing evaluation has been made for each approach leg. Time interval
step has been fixed in 1 minute.

4. APPLICATION AND COMPARISON: A CASE STUDY

Proposed models have been applied to a single lane four-legs roundabout

placed in the North East of Italy, near the city of Treviso. Its layout is quite
similar to the following figure [9], while the real dimensions are shown in the
next table.

Figure 8– Geometric elements of the roundabout [9].

A B C D
Design features
eastbound northbound westbound southbound
Entry radius (m) 12 12 12 12
Entry width (m) 4 4 4 4
Approach width (m) 3.5 3.5 3.5 3.5
Exit radius (m) 12 12 12 12
Exit width (m) 4.5 4.5 4.5 4.5
Departure width (m) 3.5 3.5 3.5 3.5
Circulatory roadway width (m) 7
Inscribed circle diameter (m) 30
Apron (m) 0
Splitter Island width (m) 7
Table 3 – Case study: geometric values.

© Association for European Transport 2004

The following tables show respectively the actual peak flows which have been
measured and their O/D fractions. It could be noticed that they are little
unbalanced demand flows.

O D A B C D
eastbound northbound westbound southbound Totals
A 0 240 120 300 660
B 243 0 194 243 680
C 309 124 0 247 680
D 230 288 173 0 690
Totals 782 651 487 790
Table 4 – Peak flows in the morning (vph).

O D A B C D
eastbound northbound westbound southbound Totals
A 0 36 18 46 100
B 36 0 28 36 100
C 46 18 0 36 100
D 33 42 25 0 100
Table 5 – Origin/Destination path (%).

5. PERFORMANCE ANALYSIS

Some different methods to evaluate un-signalised roundabout’s performance

have been tested; at the same time, also the proposed model (in its three
evolution steps) have been tested and results have been compared: in
particular, the methods analysed have been both the German models
described above. These models belong to the Statistical approach and to the
Probabilistic one. As a consequence, following results could be considered
useful as a test to underline differences in evaluating roundabouts
performances and to test the proposed model adherence to existing models.
Main results are reported below.
The first two tables show the global intersection capacity estimation, while the
following ones refer to the capacity and to the queue estimation for each leg.
Parameters involved (N°) Predicted
Compared Models
C CR βmin m γ capacity (vph)
Proposed Absolute priority 1646 1218 - - - 3800
model: 3 Limited priority 1646 1218 0 - - 2920
steps 3 parameters 1646 1218 0 0.05 0.422 2920
Linear German model Geometrical (2) 2920
Table 6 – Linear German model: results and comparisons.

Parameters involved (N°) Predicted

Compared Models
C CR βmin m γ capacity (vph)
Proposed Absolute priority 1714 1250 - - - 4015
model: 3 Limited priority 1714 1250 0 - - 3077
steps 3 parameters 1714 1250 0 0.05 0.379 2814
Exponential German model Geometrical (2); Temporal (3) 2810
Table 7 – Exponential German model: results and comparisons.

© Association for European Transport 2004

 Entry lane capacity Residual Queue (vph)
Compared Models
A B C D A B C D
Proposed Absolute priority 1014 1003 840 939 0 0 0 0
model: 3 Limited priority 786 780 637 718 0 0 43 0
steps 3 parameters 787 780 637 719 0 0 43 0
Linear German model 786 780 637 718 - - - -
Table 8 – Linear German model: results and comparisons.

Entry lane capacity Residual Queue (vph)

Compared Models
A B C D A B C D
Proposed Absolute priority 1067 1057 897 994 0 0 0 0
model: 3 Limited priority 824 818 677 757 0 0 3 0
steps 3 parameters 758 751 614 691 0 0 66 0
Exponential German model 759 752 610 690 - - - -
Table 9 – Exponential German model: results and comparisons.

The queue estimation is not possible by using the existing models and this is
one of the most important improvements due to the proposed approach.
Moreover it may be noticed that it is possible to calibrate the parameters in
order to reproduce the results of the existing models as regards both the
whole facility capacity and the capacity of each leg. These are only first
results, but they could be very interesting because it seems that the proposed
approach could represent a general framework for roundabout capacity
estimation, as it allows to reproduce the results of each existing approach
simply by calibrating the parameters.
The figures 9 and 10 show two examples of the dynamic variation of flow
patterns. In particular it is interesting to observe that capacities (ONRO(t) and
MO1(t)) are not constant due to the relationship between queue and flows of
competitive streams. Moreover the dynamic difference between the estimated
in-flow pattern (ONRI(t) and MI(t) in figure 9 and 10 respectively) and the
stream capacity (ONRO(t) and MO1(t) in figure 9 and 10 respectively), if
positive, represents the number of vehicles, which would like to transit along
the facility but can not due to other preceding vehicles; it could be noticed that
queue increases when the competitive flow is near to its capacity and vice
versa.

© Association for European Transport 2004

 25

20

15
vehicles

10

0
0 10 20 30 40 50 60
minutes

ONRO ONRI MI ONRD

Figure 9– Traffic flow patterns.

25

20

15
vehicles

10

0
0 10 20 30 40 50 60
minutes

ONRI MO1 MI SD

Figure 10– Traffic flow patterns.

From these data it is quite easy to estimate both the dynamic evolution of
queues, the total lost time in queue (which is the area of the pattern diagram
above capacity) and maximum queue length in term of number of vehicles
(which is the highest value of the inflows ONRI(t) or MI(t) in the histogram).

6. CONCLUSIONS

The paper reports some first results of a new model for the un-signalised
roundabout performances evaluation. The proposed analysis refers mainly to
capacity and queue length. The development of a different approach to
roundabouts analysis seems to be necessary to overcome the uncertainty in
roundabouts evaluation due to many different models existing in literature.
In this context, the methodology described in this paper could be a general
framework for roundabout performance evaluation because:
§ it is in accordance with the HCM main principles regarding the merge
area analysis and, at the same time, the methodology is in well

© Association for European Transport 2004

 accordance also to the Gap Acceptance theory, which is the well-
known theoretical approach to the un-signalised intersections analysis;
§ comparative results show that the developed model could easily cover
capacity prediction results of both the two considered German models
(linear model, macroscopic approach and exponential model,
microscopic approach), belonging to the called Statistical and
Probabilistic approaches respectively, with reference to the global
capacity of the roundabout and to the single approach entry capacity.
Moreover, the model could approach in a more complete way the LOS
analysis problem, predicting queue length dynamics under both light and
heavy traffic conditions. This is a very important result, mainly because many
delays and queue prediction formulations suggested for the roundabout LOS
evaluation are not suitable in congested conditions.
As further developments, the comparison between the proposed and existing
models results should be extended in order to confirm the general applicability
of the model. Moreover, as the comparison with existing performances
prediction models is a necessary step, but of course it is not enough, another
step will be the calibration of the model parameters with reference to real
data, focusing in particular to the measurements of entry flows, queues
lengths and flows distribution matrix. Also the relationships between the model
parameters and the geometrical and functional features of each roundabout
leg should be further investigated.
Finally a possible third research step will be the average total delay estimation
for each vehicle grouped by entry and exiting leg (O/D flows). Total delay
main part is given by queuing delay on single-lane small roundabouts, but on
multilane roundabouts also the geometric delay has to be included in the total
delay computation. In such a case, it should be necessary to include also the
weaving and diverging movements for a more complete roundabout
performance analysis.

REFERENCES

[1] Akcelik R., (2004), A roundabout case study comparing capacity

estimates from alternative analytical models, Proceedings of the Urban
Street Symposium 2003, Anaheim, California (USA), 28-30 July 2003.

[2] Albanese M., Camus R., Longo G. (2003) Capacity and queue modelling
for on-ramp-freeway junctions, Transportation Research Record 1852,
Transportation Research Board, 256-264.

[3] Bernetti, G., Dall’Acqua, M., Longo, G. (2003) Unsignalised vs signalised

roundabouts under critical traffic conditions, Proceedings of the
European Transport Conference 2003, Strasbourg (F), 8-10 october
2003.

[4] Brown M., (1995), The design of roundabouts, HMSO, London (UK).

[5] Fisk C. S., (1991), Traffic performance analysis at roundabouts,

Transportation Research Board- B, Vol. 25, 89-102.

© Association for European Transport 2004

[6] Hagring O., (2003), Capacity model for roundabouts, Trivector Report
2003:7, Lund (S), 20 October 2003.

[7] Hagring O., Rouphail N. M., Sorensen H. A., (2003), Comparison of

capacity models for two-lane roundabouts, Transportation Research
Record 1852, Transportation Research Board, 114-123.

[8] Highway Capacity Manual,. (2000), Transportation Research Board,

National Research Council, Washington, U.S.A.

[9] Roundabouts: an Informational Guide, (2000), US Department of

Transportation, Federal Highway Administration, U.S.A.

[10] Taekratok, T. (1998) Modern Roundabouts for Oregon, Oregon

Department of Transportation.

[11] Wang R., Ruskin H. J. (2002) Modelling traffic flow at a single-lane urban
roundabout, Computer Physics Communications, 147 570-576.

[12] Wu N. (2001) A universal procedure for capacity determination at

unsignalised (priority controlled) intersections, Transportation
Research Board- B, Vol. 35, 593-623.

© Association for European Transport 2004