DEVELOPMENT OF SHARED SPACE MICRO-SIMULATION MODELLING
AND APPLICATION FOR SCHEME DEVELOPMENT AND ASSESSMENT
Samya Ghosh,
Regional Director-AECOM, UK
Katy Thorpe
Principal Consultant-AECOM, UK
1
INTRODUCTION
Southall is a district located within the London Borough of Ealing and has the
largest Asian community in London. It is often referred to as ‘Little India’ and
is renowned for the specialist retail and culinary opportunities, attracting
visitors from near and far.
The main shopping street in Southall is The Broadway. It is a bustling
shopping parade which runs along an approximately one kilometre section of
the A4020 Uxbridge Road, which is a key east west strategic corridor route
into London. The Broadway is therefore required to fulfil both Movement and
Place functions.
Figure 1: Project Corridor: Regional Context
The main catalyst for the £7m Southall Broadway project was the worsening
road safety statistics in the London Borough of Ealing (LBE), with 41 recorded
collisions involving pedestrian injuries over a three year period from January
2008 to December 2010. Other relevant issues included overcrowded
footways; frequent traffic congestion and lengthy queuing; severance with
poor accessibility and connectivity for the busy shopping and restaurant area;
and a strong desire for economic regeneration, particularly with the advent of
Crossrail.
One of the project challenges was to produce a scheme that could balance
the required place and movement functions. It was considered desirable to
enhance the place function to reflect the unique character of the location and
its importance as a centre of economic, cultural, community and commercial
activities. However the proposals would also have to maintain the movement
function of getting people, goods and services from A to B, with an Annual
Average Daily Flow of 30,000 vehicles; up to 60 buses per hour; and 10,600
pedestrian crossing movements surveyed on a Saturday between 11 AM and
7 PM.
Based upon these requirements the agreed scheme objectives were to:
 Reduce the number of collisions and casualties along The Broadway;
 Provide a vibrant streetscape which creates a sense of place whilst
balancing the needs of all road users;
 Facilitate better movement of pedestrians through adopting shared
space principles and increasing footway capacity;
 Reduce street clutter and adopt a co-ordinated approach to materials /
street furniture;
 Provide more reliable journey times by smoothing the flow of traffic
through traffic calming and increased awareness of pedestrians;
 Encourage economic growth and have a positive impact on land values
through regeneration of the town centre and by facilitating commercial
activity along the street;
 Improve the environmental performance of the street through planting
and lighting enhancements; and
 Reduce crime and make the area feel safer through improved natural
surveillance, sight lines and street lighting.
To achieve these objectives the scheme would be required to encourage
change in both driver and pedestrian behaviour to rebalance between the
‘movement’ and ‘place’ roles whilst maintaining the corridor’s traffic flow
capacity and bus journey time reliability.
Previous attempts to address the problems on The Broadway had achieved
limited success and LBE were keen to develop a different and innovative
solution. Project Centre was commissioned as lead designers for the
scheme, and AECOM was selected to undertake the modelling. AECOM
worked closely with LBE and Project Centre from 2010 to 2014, assisting with
design development; scheme assessment and agreeing temporary traffic
management measures used during scheme construction.
The final design subdivides the scheme area into two distinctive zones –
‘Boulevard Zones’ and ‘Street Zones’ – that repeat themselves along the
route, with transition zones provided at each end of the scheme.
The Boulevard Zones create places for easier informal pedestrian crossings.
Analysis had identified that pedestrians frequently ignored controlled crossing
points, following more direct desire lines across the road or crossing on a red
man, behaviour which was believed to contribute to the high pedestrian
accident rate. The provision of Boulevard Zones recognises this inclination
and seeks to safely accommodate it within appropriate locations. In these
areas footways were widened and a narrow carriageway running lanes were
provided with a wide central median strip. The carriageways were raised to
signal the change in zone, re-enforced by changes in surface materials,
encouraging drivers to slow down and facilitating safer uncontrolled
pedestrian movement. Street furniture was placed close to the kerb in an
ordered and symmetrical manner to encourage the perception of narrowed
carriageways and encourage cautious driving behaviour. No parking or
loading is permitted in these areas.
The Street Zones provide space for functions such as bus stops and parking
and loading bays. In these locations a wider single carriageway was provided
so traffic can overtake stopped buses at bus stops and pedestrians are
discouraged from attempting to cross.
Boulevard Zone
Street Zone
Figure 2: Typical representation of Boulevard and Street Zones
The design comprises some controversial elements, including the removal of
four signal controlled pedestrian crossings and a 800m stretch of eastbound
bus lane, with the provision of a bus gate where the bus lane terminates.
Further changes were proposed to traffic circulation on local streets linked to a
multi-storey car park access.
The scheme benefits of wider footways, narrowed carriageways, slower traffic
speeds and urban realm improvements to the Place function and pedestrian
movement are intuitive, readily understood and believed. However, given the
bus lane removal and reduced speed limits it is understandable that there
were concerns raised regarding the likely impact on traffic movement and bus
journey time reliability. To assist with the design process and provide
evidence to stakeholders that the place-making elements could be
accommodated without compromising the movement function for buses and
general traffic, a traffic model was required. This model included a suitable
interface with the pedestrian flows to ensure appropriate representation of the
scheme operation and road user interactions.
2
LITERATURE REVIEW
While looking for a robust definition of shared space, the following definition
strikes out as the closest to the perception of majority of the practitioners and
end users.
“Shared space is an approach to modern street design which attempts to
break down demarcations between cars, pedestrians and other road users to
help promote the sharing of street space (Imrie, Kumar 2011)”.
It is based on the belief that motor vehicles are simply too dominant in the
road space, and policies associated with this concept are often designed to
improve pedestrian movement and comfort.
As is stated in the DfT’s Local Transport Note (2011) initiatives included
under a ‘shared space’ policy may include:
 removing priority of vehicles over pedestrians,
 removal of traffic lights,
 or introducing physical/psychological features which encourage lower
speeds.
Such policies have proven somewhat successful in continental Europe; and
there have been an increasing number of trials for associated schemes on UK
road space.
It is important to note that, as the MVA Consultancy describe in their 2010
report for the DfT, ‘shared space’ is not simply a design type characterised by
‘standard’ features; but more-so a separate design approach which may
incorporate a range of physical and psychological initiatives.
Many of the earliest shared space schemes occurred in the Netherlands; such
as The Laweiplein in Drachten (figure 1) which, in 2012, was estimated to
accommodate approximately 22,000 vehicle movements per day (Moody and
Melia, 2014). This was previously a traffic signalled controlled junction before
being turned into a public square with a roundabout. The overall findings were
deemed to show that the scheme had a positive response on safety with no
injury accidents in the 3 years up to 2006. This is compared to pre-2001,
where there were significantly more injuries when the junction was under
controlled conditions (Firth & Buchanan, 2001).
Figure 1. Shared space in Laweiplein, Drachten, the Netherlands
It was not until much later that similar schemes became apparent in the UK;
and it was not until 2009 (in Portishead, near Bristol) where local authorities
started trialling schemes which involved turning off the traffic lights. Firth &
Buchanan (2011) present a detailed account of the findings from this trial.
“When conventional traffic lights were removed, the maximum crossing times
reduced by over 20% on average; and the paper concluded that the scheme
resulted in less congestion and allowed greater capacity along the road
space”. However, the lack of statistical tests evaluating the roles of other
factors in the analysis could be perceived as a problem with the study.
The report by MVA consultancy for the DfT (2010) analysed 10 shared space
schemes in attempt to explore the way pedestrians use space at each of the
sites, and by using regression formulae, to determine by how much of this
pedestrian use is influenced by factors which were introduced as part of the
shared space schemes. It concludes that the new design characteristics did
lead to a higher percentage of pedestrians using the carriageway compared to
what would normally be expected. However, “the extent of the effect on
pedestrian movement by individual shared space characteristics was
extremely small, and a large-scale revamp of design characteristics would be
needed in order to achieve desired results”.
Figure 2. Elwick Square in Ashford, Kent, UK (before and after
implementation of shared space design)
Moody & Melia (2014) conducted a study on Elwick Square (figure 2); where
there was the construction of a new ring road system with a level surface
between the roadway and the footway, as well as the removal of street clutter
such as traffic signals, kerb and road markings. They analysed pedestrian
behaviour across the square; finding that even with the new layout, people
tended to use the informal crossings which still existed in the area instead of
their desired line. Also, “in the majority of cases (72%), the pedestrian still
gave way to the vehicle, and 80% of the people surveyed said they felt safer
when crossing the previous layout”. This negative response from the public
was also present in the Firth & Buchanon (2011) study of Bristol and its
surrounding areas; where 2/3’s of drivers passing through one of the junctions
surveyed in the scheme believed that the junction was safer when under
controlled circumstances.
However, criticism of shared space came from totally unexpected quarter. The
general opinion to shared space concept from people who are partially sighted
or blind has been negative. This is especially important as the most vulnerable
groups of society should be involved in all aspects of policy implementation;
from selection to design. The introduction of level surfaces (with no
demarcations between the road and the footpath) and the removal of
controlled junctions could both be seen as being problematic for people with
impaired mobility. Interview conducted by Matthews et al. (2011) as part of his
study on the “blind and visually impaired in the context of shared space
revealed that these vulnerable groups were unanimous in their opinion that
controlled junctions were vital for allowing them to cross the road”.
A study from Childs et al. (2010) attempts to obtain an optimum elevation
height for demarcations between the road and the footpath, but comes up
inconclusive. They believe that there will always be problems for people with
impaired mobility; with elevation in kerbs causing a hindrance in traditional
road spaces, and the lack of detection of delineated surfaces in areas where a
shared space policy has been implemented.
Imrie & Kumar’s (2012) account of shared space and its effect on people with
sight loss rejects the idea that it is simply design issues which causes this
group of people to dislike the concept of shared space. They believe that
there is a ‘trust gap’ between the policymakers and the visually impaired,
since this particularly vulnerable group of people are not involved in any
policymaking. This report revealed the significant lack of participation by
people with sight loss in all earlier stages of policymaking, with them only
starting to be consulted in the latter stages of the process (such as the design
phase).
3
SCHEME ASSESSMENT
Shared space scheme assessment is different from traditional transportation
major scheme assessment for the following reasons:
a) Shared Space schemes do not always reflect similar benefit streams to
traditional transport schemes;
b) There are many micro-environmental, lifestyle and ambience benefits
which can be appreciated qualitatively but are difficult to quantify in
analytical form;
c) Much of the behaviour in shared space depends on the behaviour of
drivers and pedestrians in a congested and conflicted situation, which
traditional modelling tools do not successfully represent; and
d) Shared space schemes impact more visibly on the local area in the
short term and provide a progressively wider impact in the longer term.
This can be difficult to predict and assess, especially in absence of
proven evidence from mature schemes.
With the above challenges in mind, an existing micro-simulation traffic model
was used to assess the impact of the scheme on traffic in a rational and
robust manner.
An existing base micro-simulation VISSIM model had previously been
developed by AECOM and approved via the Transport for London (TfL) formal
Model Audit Process (MAP). This model was used as a basis for assessing
the likely impact of the scheme proposals on the operation of the Uxbridge
Road corridor and surrounding local road network in terms of capacity and
delay to vehicular traffic.
The proposed Southall Broadway scheme comprises numerous changes to
the road layout. One initial task was to review the proposals in detail to
determine which features were likely to affect the highway capacity and
should therefore be represented in the proposed VISSIM model. Scheme
elements that either could not or need not be assessed with the support of
micro-simulation modelling included: wider footways; road safety impacts;
changes to parking and loading provision; and general urban realm
improvements. However, the following design components were agreed for
inclusion in the proposed modelling:
 Removal of eastbound bus lane;
 Removal of four existing signal controlled pedestrian crossings;
 Revision of junction layout at South Road / The Broadway signalised
junction;
 Implementation of eastbound bus gate at the west of the scheme with
linked pedestrian crossing;
 Implementation of ‘Boulevard Zones’ and ‘Courtesy Crossings’ which
are located on raised tables and are intended to encourage safe
uncontrolled pedestrian crossing;
 Implementation of 20 mph speed limit;
 Bus stop re-location;
 Beachcroft Avenue conversion to one-way southbound operation;
 Herbert Road conversion to one-way northbound operation;
 Extension of right turn pocket from The Broadway to Woodlands Road;
and
 Re-allocation of right turning traffic from Northcote Avenue to Dane
Road.
Whilst many of these changes were straightforward in modelling terms, some
required the development of bespoke coding, namely the representation of
the Boulevard Zones and a bus gating strategy that could intelligently react to
the build up of congestion within the study area to protect bus journey time
reliability. The modelling assumptions and methodology applied to the final
impact assessment for each of these elements is presented below.
3.1
Representation of Boulevard Zones
The current official TfL Modelling Guidelines do not include specific guidance
relating to shared space or courtesy crossings and a literature review did not
reveal appropriate quantitative analysis relating to comparable existing
schemes. Therefore the assumptions and modelling techniques for the
assessment were developed in conjunction with TfL modelling specialists as
part of the model development process to ensure that the potential scheme
impact was appropriately and robustly assessed.
It was recognised that it would be unrealistic to believe that a single model
could be developed to exactly represent the future operation of the scheme –
particularly as it would be so dependent upon the attitude and behaviour of
individual drivers and pedestrians, which would themselves vary and be
dependent upon many subtle conditions. The acknowledgement that any
model is only ever intended to be a simplified version of reality allowed a
pragmatic approach to be adopted. A modelling methodology avoided oversimplification of the proposals whilst being practical to adopt within required
time and budgetary constraints, and was agreed and used to undertake a
range of sensitivity tests. This allowed stakeholders to understand how
susceptible the model outputs were to changes in modelling parameter
assumptions, and to therefore have confidence that the likely scheme impact
would lie within a given range.
The Southall Broadway urban realm proposals include the removal of the four
signalised pedestrian crossings along The Broadway and the provision of four
‘Boulevard Zones’, three of which contain ‘Courtesy Crossings’. The Design
Note produced by the scheme designers Project Centre describes the
philosophy of the Boulevard Zones as follows:
‘It is the intention of the design to provide large areas of the street where
pedestrians can make informal crossing of the carriageway, across narrow
carriageway running lanes and a wide central median strip where pedestrians
can wait between crossing each carriageway. These spaces have been
termed as the ‘Boulevard zones’. Material detailing in terms of surface finishes
and levels will provide a clear strong signal to vehicle drivers that these
spaces are pedestrian orientated. This combined with a 20 mph speed limit
will provide an environment where pedestrians can negotiate their passage
from one footway to the other.’
The thinking behind the Courtesy Crossings was also explained as follows:
‘...the design proposed includes an additional more recognisable courtesy
crossing similar in form to that of a zebra crossing where material detailing
including tactile paving will indicate to both drivers and pedestrians that a
more recognisable crossing facility exists... The intention is that these more
recognisable crossings provide an additional facility for those who are not
confident about crossing the carriageway via the central median on its own.’
3.2
Existing traffic and pedestrian data
Traffic Flows
Traffic turning flow data at seven junctions and extensive video footage was
collected along The Broadway during typical weekday and Saturday
conditions as part of the base model development. The proposed modelling
assumed no change to total vehicular inputs, although
a number of sensitivity tests were performed regarding manual redistribution
onto side roads where new movements were permitted.
Observed cycle flows were very low and, in agreement with TfL, were not
included in either the existing or proposed models.
Pedestrian Flows
A pedestrian crossing survey was conducted along the scheme area, with
data divided into ten zones, again both on a weekday and Saturday. At the
existing formal crossing points the counts were subdivided by direction and
into pedestrians crossing on a green man and those crossing on a red man.
Analysis of the pedestrian flow data showed that, in addition to high demands
at the four formal crossings there were smaller but notable peaks at the short
median strip by Dane Road and by St Georges Avenue. This demonstrated
that a reasonably high level of uncontrolled crossing was already taking place.
It was noted that drivers were not observed to give priority to pedestrians
crossing in these locations, rather that pedestrians were gap seeking between
the vehicles. This provided some evidence that in the proposed scheme,
where pedestrian crossing distances would be shorter, vehicles would be
unlikely to feel compelled to cede priority to pedestrians.
It was not anticipated that the scheme should significantly impact pedestrian
crossing demand or desire lines as land uses adjacent to the corridor which
either generate or attract pedestrians would not change. Therefore the
existing crossing data was considered to provide a good indication of the
pedestrian crossing demand that should be reflected in the VISSIM modelling.
Three of the Boulevard Zones are positioned in approximately the same
locations as existing controlled pedestrian crossing facilities. In these
locations it was agreed that pedestrian surveys at the relevant crossing would
provide a good indication of likely crossing volumes. There was no existing
pedestrian crossing provision in the most western Boulevard Zone, positioned
between Lancaster Road and Dane Road. It was considered that the
provision of a median strip may encourage higher pedestrian crossing
movements in this area, and therefore crossing demand in the proposed
modelling may be higher than existing.
In the VISSIM model, within each defined time interval, vehicles or
pedestrians enter the links at a frequency based on a Poisson distribution.
Consequently the common ‘bunching’ of pedestrians into social groups is not
replicated. For example, an observed flow of 5 pedestrians in a 5 minute
period could in reality represent 2 groups of people and therefore result in up
to 2 instances of vehicles incurring delay by giving way. However this might
be modelled in VISSIM as one pedestrian a minute which could result in up to
5 instances of vehicles incurring delay by giving way. Sensitivity tests were
run with reduced flows in order to better understand the potential impact this
limitation could have upon the level of modelled delay.
3.3
Pedestrian Speeds
VISSIM (at the time of model development) has four default pedestrian
desired speed distributions, namely:
 IMO-M 30-50 (3.5 – 5.8kph);
 IMO-F 30-50 (2.6 – 4.3kph);
 Fruin 1 (2.1 – 6.6); and
 Fruin 2 (2.1 – 6.6).
Within the TfL VISSIM input file template there is also a bespoke speed
distribution with a range of 4.0 to 6.0kph.
Chapter 18 of the Highway Capacity Manual (HCM) states:
‘Pedestrian walking speed depends on the proportion of elderly pedestrians
(65 years of age and older) in the walking population (1). If 0 to 20 percent of
pedestrians are elderly, a walking speed of 1.2 m/s [4.32kph] is recommended
for computations for walkways. If elderly pedestrians constitute more than 20
percent of all pedestrians, a 1.0 m/s [3.6kph] walking speed is
recommended..’
The average walking speeds of TfL Distribution 1, ‘IMO-M 30-50’ and ‘IMO-F
30-50’ are 5.0kph, 4.65kph and 3.45kph respectively. As the TfL distribution
was considered to be most likely to be based upon empirical data collected at
relevant sites, this speed distribution was selected.
3.4
Pedestrian Models
Single link test models representing a simple ‘shared space’ were used to
compare pedestrian and vehicle delays using pedestrians either modelled
within VISSIM (where pedestrians are defined as another ‘vehicle type’) or
using the add-on pedestrian modelling VISWALK module; and priority control
represented using either priority markers or conflict areas. Analysis of the
outputs suggested that the different techniques resulted in relatively small
differences. It was decided that the VISWALK pedestrian module option of
VISSIM should not be used so that the whole modelling could be undertaken
on a single platform.
3.5
Boulevard Zones / Shared Space
The Boulevard Zones are designed to encourage a greater volume of
uncontrolled pedestrian crossing and reduced vehicular speeds. Whilst
pedestrians would not have right of way it was expected that the slower
speeds, shorter crossing distances and the feeling of ‘shared space’ should
encourage pedestrians to feel safer when crossing amongst vehicles.
Determination of crossing location
The nature of Boulevard Zones means that a pedestrian can cross at any
point in the Zone. To simply represent a range of potential crossing locations
each Boulevard Zone would have a number of discrete allocated crossing
points (modelled using links)spread out evenly along the Zone.
Pedestrians were coded to enter the ‘footway’ link at either end of the
Zone, both to the north and south of The Broadway. If the pedestrians arrived
at the first crossing opportunity on the Boulevard Zone with a sufficient gap in
vehicular traffic then the pedestrian would choose to cross the road at that
point. However, if the gap to the next vehicle was insufficient then the
pedestrian would continue walking to the next crossing point. The pedestrian
would continue to bypass crossing points if there was not space to cross until
they either reach the final crossing point (at which point they would cross
there when safe) or they reach a Courtesy Crossing, where they would have
right of way and therefore would cross at that point.
The decision whether to cross and the consequent change in pedestrian
routing across the road was controlled using detectors, fixed routing decisions
and VISVAP control.
Figure 2 includes figure of a pedestrian unable to cross at each crossing point
and travelling to the final crossing point. Figure 3 shows a pedestrian
travelling past the first crossing point but deciding to cross at the second
crossing point due to the lack of approaching vehicles.
Pedestrian approaches
first crossing point.
Figure 2: Pedestrian Original Crossing Intent
Due to vehicles also approaching
the crossing point the pedestrian
decides not to cross at this point
and continues along the pavement.
Figure 3: Pedestrian Amended Crossing Route
Pedestrian priority
A range of pedestrian priority assumptions were coded to explore more
positive and negative outcomes.
In order to robustly assess the potential operation of the Boulevard Zones
within the ‘Worst Case’ VISSIM model, two pedestrian types were coded:
 ’Pedestrian’ (50% of the composition); and
 ‘Pedestrian with priority’ (50% of the composition).
Priority markers were applied on the footway and carriageway links to ensure
that vehicles on the carriageway give way to pedestrian type ‘Pedestrian with
priority’ and only pedestrian type ‘Pedestrian’ give way to the vehicles. The
different pedestrian types were easily differentiated by colour in the
simulation.
In the ‘Medium Case’ VISSIM model, three pedestrian types were coded:
 ’Pedestrian’ (20% of the composition); ‘
 ‘Pedestrian with priority 1’ (50% of the composition); and
 ‘Pedestrian with priority 2’ (30% of the composition).
Priority markers were applied on the footway and carriageway links to ensure
that vehicles within the Boulevard Zones but outside of a Courtesy Crossing
give way to pedestrian type ‘Pedestrian with priority 1’ only. Therefore within
Boulevard Zones only 50% of pedestrians have priority. At Courtesy
Crossings vehicles only give way to 80% of the traffic, (that is Pedestrians
with priority 1 and priority 2).
Two sensitivity tests were undertaken on the ‘Worst Case’ scenario; one
which assumed that pedestrians had right of way 75% of the time and one
which assumed they had right of way 25% of the time. These tests were
again undertaken by simply changing the pedestrian composition. This rule
was not applied to buses, which always had priority over pedestrians within
Boulevard Zones except at the Courtesy Crossings.
3.6
Courtesy Crossings
Courtesy Crossings were treated similarly to Zebra Crossings within the
VISSIM modelling procedure.
Pedestrians in the Worst Case scenario were assumed to have right of way
over vehicles when crossing the road despite there being no obligation for
drivers to stop. In the Medium Case scenario 80% of pedestrians were given
priority at the Courtesy Crossings. Only pedestrian type ‘Pedestrian’ was
coded to give way to the vehicles in both the Boulevard Zones and Courtesy
Crossings.
3.7
Sensitivity Tests
The range of scenarios modelled to provide an understanding of the range of
potential scheme impact included:
 A ‘Worst Case’ scenario to provide an extremely robust, if potentially
unrealistic, assessment of the potential impact of the Boulevard Zones,
assuming that pedestrians assert priority over vehicles in these
crossing areas with vehicles ceding priority along the route. All
pedestrian flows observed within the study area were allocated to
either a Boulevard Zone or the proposed signal control pedestrian
crossing by the bus gate, according to the zone. Sensitivity tests were
run on this scenario with different proportions of pedestrians assuming
priority over vehicles. Different proportions of drivers giving way to
pedestrians were tested, namely:
o 25%;
o 50%; and
o 75%.

A ‘Medium Case’ scenario which applies reduced pedestrian inputs to
reflect that a proportion of pedestrians are still confident enough to
cross away from the Boulevard Zones; some pedestrians arrive in
groups; and some drivers will not stop for all pedestrians at the
Courtesy Crossings. Specifically it was assumed that
o 50% of pedestrians who are currently crossing in a street zone
in the existing situation will remain happy to cross in a street
zone (and not divert to a Boulevard Zone).
o One in ten pedestrians will arrive at the crossing point in a group
and therefore the flows are reduced by 10%. The arrival rate
impacts on the model operation as drivers are modelled to
observe and give priority to waiting pedestrians.
o Vehicles will now cede priority to 80% of pedestrians at the
Courtesy Crossings.
The approach outlined above does not provide the most detailed or
aesthetically pleasing methodology for modelling shared space. However the
strengths of this approach were considered to be: the robust nature of
assessment; the ease of coding multiple sensitivity tests; and the confidence
this provided to stakeholders that the range of potential level of impact on
traffic capacity was being sensibly assessed.
3.8
Development and representation of bus gating strategy
A potential eastbound bus gating strategy was developed, using the VISVAP
VISSIM add-on module, with the objective of mitigating the impacts of
removing the eastbound bus lane by providing priority to buses and providing
a mechanism for managing traffic flows through the corridor during peak
periods.
The proposed bus gate was originally modelled with the assumption that a
bus demand for the bus gate would be detected when the bus moved away
from the upstream bus stop. A seven second minimum green for the buses
was provided, timed to start approximately as the bus arrives at the bus gate.
A key capacity constraint of the eastbound Broadway link is the signalised
South Road crossroads junction, located at the eastern end of the scheme.
Following the removal of the bus lane, early model runs showed that the
extensive queues which develop on The Broadway from this junction could
impose delay on buses which the bus lane had previously allowed them to
avoid.
Therefore a bus gating strategy to detect when a queue is developing and
relocate it away from the Southall Boulevard area was required. The objective
would be to provide a ‘virtual bus lane’ for buses and have the additional
benefits of enhancing the Broadway Boulevard urban realm proposals by
reducing the queuing traffic; improving air quality in an area with a particularly
heavy pedestrian footfall; and creating traffic platoons through the zone to
facilitate uncontrolled pedestrian crossing and assist side road traffic to enter
The Broadway.
To achieve this objective the flow of general traffic was controlled at the bus
gate by adjusting signal timings to reduce capacity, so that the bus gate
becomes more capacity critical than the downstream junction. The existing
bus lane running up to the bus gate would then enable buses to by-pass the
relocated traffic queue.
The scheme proposals also included a new signal controlled pedestrian
crossing approximately 50 metres east of the bus gate. It was proposed that
the bus gate and the pedestrian crossing should operate as separate streams
with the same signal controller and should be co-ordinated to minimise bus
delay.
3.9
The Strategy Modelled in VISSIM
Two key amendments were made to the original modelled operation of the
bus gate:
 The move to the bus stage occurs one second after the bus is
detected to leave the stop. This creates a gap in general traffic flow to
allow the bus to travel into; and
 Additional detectors were placed downstream of the bus gate to
identify when the link is congested so that general traffic could be held
back at the bus gate, effectively relocating the queue.

Eastbound detectors were also placed on the carriageway to the east of the
pedestrian crossing to monitor the presence of buses between the bus stop
and crossing.
The coded signal control logic is outlined below:
3.10 Bus Gate Operation: Non-Congested Conditions
If downstream queues are not detected and no bus is detected the signal shall
stay green to general traffic.
 A bus demand was placed for the bus stage when a bus was detected
leaving the upstream bus stop. The bus demand would remove all
extensions to general traffic and allow the bus stage to be called
immediately. This ensured minimal delay to buses and held general
traffic back to create a gap for buses to travel into directly downstream
of the bus gate.
 A cancel loop downstream of the bus gate cancelled the bus demand
and returned to the traffic stage.
 The above was subject to phases serving their minimum green periods
(7 seconds) prior to the stage change.
 If a second bus was detected during the bus stage then the bus stage
was held to allow the second bus to proceed through the green signal.
 The bus stage was allocated a maximum green time of 40 seconds to
prevent excessive green times been provided.
Bus
detected
leaving
stop.
Demand recorded.
Bus Stop
1s
later,
provided
General Traffic (GT) has
had minimum green, the
move starts to bus stage
Once the bus has been
detected
leaving,
provided there are no
other buses between
Dets 100 & 200 start the
move back to GT stage
If there is another bus
following, providing the bus
green time isn’t greater
than 40s, the bus stage will
extend to let it through
Figure 4 – General Bus Gate Operation
3.11 Bus Gate Operation: Congested Conditions
Queue Detection and Selection of Timing Plan
 If downstream queuing is detected then the bus stage is called
irrespective of whether a bus has been detected. A fixed time plan is
called to restrict the green time provided to general traffic. This
effectively relocates the traffic queue to the west of the bus gate.
 The existence of downstream queuing is determined by assessing the
level of SCOOT congestion on three separate loops. These are
located approximately 100m, 200m and 300m west of the South Road
stopline respectively.
 SCOOT congestion is defined as the number of four second intervals
in a cycle during which the detector is continuously occupied. It is
expressed as a proportion of the cycle time, with the assessment
period set to 300 seconds (five minutes).
 If the value of SCOOT congestion is greater than 5%, congestion is
flagged in this location and appropriate action is taken based on which
loops are congested.
Bus Gate
Congestion Detectors
South
Road jn
Figure 5 – Congestion Detection Layout

Different green times were modelled based on which loops were
congested. Within the VISSIM control logic four congestion levels were
identified:
o Level 0 - If none of the detectors has congestion flagged.
o Level 3 - If Detector 3000 has congestion flagged (regardless of
status of 1000/2000).
o Level 2 - If Detector 1000 and Detector 2000 have congestion
(and Detector 3000 does not have congestion).
o Level 1 - Either Detector 1000 or Detector 2000 have
congestion, but not both.

A further ‘FTOpp’ variable was defined to determine the Fixed Time
plan to be run, which is assigned a value of 0, 1, 2 or 3. The above
congestion levels determine which Fixed Time plan to call.
o Fixed Time plan 0 does not automatically assign any green time
to the bus gate and thus Stage 2 is only called when a bus
presence is detected and a demand is placed (see the logic
outlined above).
o Fixed Time plans 1 – 3 provide increasing lengths of green time
to be assigned to the bus stage every cycle, and are required
when congestion is detected and result in general traffic being
gated into the Broadway Boulevard zone. The bus gate is set to
double cycle, with cycle times of 48 or 52 seconds, depending
on the time period. Timing plans vary between the time periods,
with the greatest capacity restraint provided in the Saturday
model, which generally suffers from higher levels of congestion.

Within the VISSIM model the algorithm checks which Fixed Time plan
to run every five minutes. The Fixed Time plan selection is determined
based upon the following rules:
o When the Congestion level is greater than or equal to the
current ‘FTOpp’ value, then the updated ‘FTOpp’ value (i.e.
Fixed Time plan number) is equal to the new Congestion level
value.
o When the Congestion level is lower than the preceding 5
minutes level then:




The ‘FTOpp’ value (i.e. Fixed Time plan number) is
decreased by a value of 1 (i.e. it cannot jump from 3 to 1
or from 2 to 0, etc.).
If the gating strategy had been activated (i.e. Fixed Time
plan 1, 2 or 3 is operating), the measured congestion
level had to be zero for two five minute intervals in a row
before Fixed Time plan 0 can be adopted and traffic
restraint ceased. The aim of this is to ensure that
congestion has successfully been reduced and avoid
switching the gating strategy on/off at short intervals.
When Fixed Time plans 1 – 3 are in operation, there is a seven second
window for the bus stage to be called every cycle. This will only occur
if general traffic has received a specific threshold green time which is
different in each timing plan.
The bus stage is then run for a minimum green time which is different
in each timing plan (the higher the congestion level, the longer the
green time).



Once this minimum green time has been achieved, if there is no bus
detected between the bus stop and bus gate, the move back to the
general traffic stage will commence. A maximum extension time is
coded to ensure that the possible extension of Stage 2 is limited.
The general traffic stage then runs until the next window for the bus
stage.
The communications channels (between the bus gate and the nearby
pedestrian crossing) are used to flag when a bus is travelling between
the bus stop and the pedestrian crossing and if any of Fixed Time
plans 1 to 3 are operating.
Pedestrian Crossing Operation – NON-CONGESTED Conditions
 Demand for the pedestrian crossing is checked. If the traffic stage has
been operating for a sufficient time (defined differently in each time
period) and there is no ‘Hold’ for buses (i.e. Marker 1 has not flagged a
bus presence between the bus stop and the pedestrian crossing) then
a move to the pedestrian stage can commence.
 If a ‘Bus Hold’ has been placed, the vehicle stage is held, but the
effective time of this ‘Hold’ is monitored. Once there are no buses in
the area the ‘Hold’ is cancelled. A maximum of 25 seconds extension
for the bus is allowed before the move to pedestrian stage occurs.
Pedestrian Crossing Operation – CONGESTED Conditions
 Cycle time is set to match the bus gate.
 In congested conditions the bus gate installation is usually prompted to
move to the bus stage at 20 seconds. At 28 seconds, if there is no
‘Bus Hold’ placed, the pedestrian crossing would start a move to the
pedestrian stage if demand has been placed. The reasons for this are:
i.
An eight second offset would allow all eastbound general traffic
which has just passed through the bus gate to cross the
pedestrian crossing without being held; and
ii.

It is considered less likely that drivers would resent being held at
a red signal without the presence of a bus if they can see
another red signal downstream. This is intended to help make
the gating strategy more acceptable.
If a ‘Bus Hold’ has been placed the vehicle stage is held and the
effective time of this ‘Hold’ is monitored. A maximum of 40 seconds
extension for the bus is allowed before the move to pedestrian stage
occurs. As soon as the bus is detected passing the pedestrian
crossing the ‘Hold’ is cancelled.
AECOM do not know precisely how TfL have configured and operate the bus
gate and pedestrian crossing. However, the traffic control logic in the Stage 5
VISSIM model was coded based upon discussion with TfL throughout the
model’s development to represent a bus gating strategy which could be
potentially be implemented.
The model outputs provided evidence that a bus gating strategy could be
developed and implemented to help to protect eastbound bus journey times
against the detrimental impact of the proposed removal of the eastbound bus
lane through the Southall Broadway scheme area whilst retaining existing
capacity for general traffic along the corridor.
3.12 Scheme Impact
The operation of the Boulevard Zones and Courtesy Crossings in terms of the
interaction between vehicles and pedestrians along these zones was clearly
critical to the delay incurred by traffic along the corridor. The propensity of
drivers to give way to crossing pedestrians determined the frequency and
extent of delays, not only to the drivers choosing to give way but to those
required to slow down or stop behind them.
The outputs were a surprise to many stakeholders, with the overall conclusion
that the scheme would provide:
 Reduced pedestrian delays;
 Minimal impact (< 1 minute) on bus journey times achieved via gating
strategy;
 Small increase (circa 1 minute) in eastbound general traffic AM & PM
journey times; and
 Large reduction in eastbound general traffic Saturday journey times
(circa 5 minutes).
In the Worst Case scenario , (all observed crossing pedestrians move to the
Boulevard Zone to cross, drivers willing to give way to 50% of pedestrians at
any point on the Boulevard Zone and 100% of pedestrians at the Courtesy
Crossing), some detrimental impact was observed in modelled traffic flows
and delays westbound along the corridor. However it was deemed unlikely
that drivers would feel obliged to cede this level of priority to pedestrians.
Westbound modelled bus journey times showed an approximately 25 seconds
increase in the AM peak, 45 seconds in the PM peak and over two minutes in
the Saturday peak. The greatest detrimental impact was modelled in the
Saturday peak where the very high pedestrian crossing movements on the
Courtesy Crossing by Herbert Road resulted in a westbound queue from this
point which extended back to the South Road junction. The other peak
periods have significantly lower pedestrian flows in this location which is why
the detrimental impact modelled was considerably lower.
The more realistic assumptions in the Medium Case scenario resulted in
outputs suggesting that the scheme could be implemented with minimal
impacts on general traffic and bus journey times through the network, with the
sensitivity tests suggesting even lower levels of impact.
The Saturday Base model had the slowest existing journey times eastbound
of any of the modelled scenarios, with queuing extending back along the
whole eastbound link for the majority of the peak hour. The proposed model
outputs show benefits to eastbound general traffic journey times in the
Saturday peak in all scenarios, providing a reduction of approximately six
minutes.
Eastbound bus journey times in all three time periods show a small increase.
In the Worst Case scenario these were approximately 40 seconds in the AM
peak; 30 to 60 seconds in the PM peak and 45 to 60 seconds in the Saturday
peak. However increases were notably lower in the Medium Case scenario,
with 15 seconds increase in the AM peak, 20 seconds in the PM peak and 5
seconds in the Saturday peak.
These results were naturally welcomed by the scheme designers and
sponsors. However it was necessary to rationalise these conclusions so that
they could be justified to a wider audience. Model analysis and observation
showed that key contributory factors to the achievement of these benefits
were:
 Removal of delay points at signal controlled crossings (up to 26s lost
time / typical 96s cycle – i.e. 27% at each crossing);
 Removal of conflict from busy side road pushing in front of traffic;
 Short pedestrian crossing distances in the Boulevard Zones with single
lanes and wide median strip;
 Platooning of vehicles from signal control on the western and eastern
model extents providing gaps and crossing opportunities; and
 Bus gating strategy reacting dynamically to protect bus journey times.
Based on the above outputs it was concluded that the scheme would be
beneficial to all road users while enhancing the safety and ambience of the
project area and was therefore taken forward for implementation in 2014.
The model was also used iteratively to inform the Temporary Traffic
Management (TTM) proposals for the construction stages and played a vital
part in successfully delivering the scheme.
4
SUMMARY AND LESSONS LEARNT
This paper has presented the bespoke modelling techniques developed to
represent the shared space ‘Boulevard Zones’ and bus gating strategy
elements to support the development and appraisal of the Southall Broadway
Boulevard scheme in the London Borough of Ealing.
The shared space Boulevard Zones were modelled in VISSIM using priority
markers and routing decisions changed dynamically using VISVAP. A range
of pedestrian types were coded so that a variety of driver and pedestrian
behaviour scenarios could be tested. This technique was easy to apply and
allowed the development of a good understanding of the range of possible
outcomes.
An intelligent bus gating strategy to protect bus journey times following the
removal of the existing bus lane was developed and tested using the VISSIM
add-on module VISVAP. This was used to demonstrate to stakeholders that it
would be feasible to implement a mechanism for dynamically managing
congestion within the study area, with added advantages for pedestrians and
general traffic.
Our experience shows that there is no need for very advanced modelling
techniques to realistically appraise the shared space schemes. Rather a
traditional but robust micro-simulation tool can be improvised to realistically
assess the interface between various elements of the shared space and can
predict the operational challenges and benefits.
It was clear through our experience that regular communication between
design and modelling teams and the value of feedback are critical for success
of these projects.
The importance of stakeholder engagement was another key element
enabling us to incorporate all local factors in the assessment tool and hence
the model outputs could answer all the questions from the relevant and
affected stakeholders.
It is also vital not to assume at the outset that you know the right answer. It is
critical to understand the existing network, all the relevant parameters and
their sensitivities. This understanding will allow them to be appropriately
incorporated within the models which can then be used to practically produce
outputs that are considered robust by all stakeholders.
Initial observations show the scheme has achieved its objectives. Pedestrians
have no problem crossing the road, traffic is moving smoothly and buses are
operating without additional delays. Residents are extremely satisfied with the
end results; members of the public have been congratulating ward councillors
on the “wonderful” scheme and the deputy leader of the council was “full of
praise”. Finally, the success of the project was formally recognised by Cabinet
who agreed to “congratulate officers on this project”.
A detailed ex post study is currently being contemplated, where the objective
is to understand how the scheme is performing; whether it is meeting the
intended objectives and how accurate were the forecasts produced by the
modelling; and if there are notable differences then what are reasons for any
differences. The data collection for this work is scheduled from the second
week of September 2015 and the modelling and analysis through the Autumn.
The findings from this ongoing work will undoubtedly present some interesting
facts which can be taken forward for appraisal of similar schemes in the
United Kingdom.