Emergency Planning College
Understanding Crowd
Behaviours:
Simulation Tools

ii

This report has been commissioned by, and prepared for, the Cabinet Office
It forms part of the ‘Understanding Crowd Behaviour’ research report series. 
The research was sponsored and funded by the Cabinet Office, as part of
the canon of civil protection literature and guidance, published on their UK
Resilience website:

(www.cabinetoffice.gov.uk/ukresilience)


The research was carried out, and the report series produced, by
Organisational Psychologists at the Centre for Socio-Technical Systems
Design (CSTSD) and the Centre for Organisational Strategy, Learning and
Change (COSLAC) at Leeds University Business School.

(lubswww.leeds.ac.uk/cstsd)
(www.leeds.ac.uk/lubs/coslac)


Rose Challenger, BSc, MSc
Researcher in Organisational Psychology

Professor Chris W. Clegg, BA, MSc, FBPsS, FBCS, FRSA
Professor of Organisational Psychology

Mark A. Robinson, BSc, MSc
Researcher in Organisational Psychology


The research was project managed by the Emergency Planning College,
on behalf of the Cabinet Office.

Mark Leigh BA, MSc, MA
Consulting Editor






For further information, please contact:
The Cabinet Office Emergency Planning College
The Hawkhills
Easingwold
York
YO61 3EG

Tel: 01347 825000
Email: epc.library@cabinet-office.x.gsi.gov.uk



© Crown Copyright 2009

ISBN 978-1-874321-22-4

Published: June 2009

iii












DISCLAIMER

Please note, the recommendations made in this report regarding
good practice for event preparation and crowd management are
an interpretation of best practice made on the basis of knowledge
and expertise gained from literature and interviews. They are
not definitive rules of event preparation and crowd management.

Contents

iv

Contents


List of Figures vi
Foreword vii
Acknowledgements ix
A Guide for Readers xi

Executive Summary 1

Simulation Tools Review 10

Legion 12

Applications of Legion 13
Legion Software 14
Approaches to Simulation 16
Assumptions Underlying Crowd Behaviours 19
Evaluation of Legion 21
Validation 21
Strengths of Legion 21
Weaknesses of Legion 21

Myriad II 23

Applications of Myriad II 24
Approaches to Simulation 25
Assumptions Underlying Crowd Behaviours 32
Evaluation of Myriad II 33
Validation 33
Strengths of Myriad II 33
Weaknesses of Myriad II 33

Mass Motion 34

Applications of Mass Motion 35

Contents
v

Approaches to Simulation 36
Assumptions Underlying Crowd Behaviours 38
Evaluation of Mass Motion 40
Validation 40
Strengths of Mass Motion 40
Weaknesses of Mass Motion 40

Future Simulation Tools 41

Key Learning Points 43

List of Figures

vi

List of Figures


Figure 1. Screenshot from a Legion Studio simulation of crowd 14
Movement in the Sydney 2000 Olympic stadium

Figure 2. Screenshot from a Legion 3D visualisation, showing a large 15
cityscape with simulated pedestrians and vehicles

Figure 3. The Myriad II modelling suite 26

Figure 4. Screenshot of a Network Analysis simulation, with different 27
colours used to indicate the optimal flow capacities for alternative exit routes

Figure 5. Example of a Spatial Analysis simulation, indicating variations 28
In crowd density throughout the given environment

Figure 6. Screenshot from an Agent-Based Analysis simulation, showing 29
the movement of individuals through a complex environment

Figure 7. Screenshot from a Mass Motion simulation of crowd movement 36
at Union Station in Toronto

Figure 8. Screenshot from a Mass Motion simulation of crowd movement 38
at Transbay Terminal in San Francisco

List of Tables

vii













Foreword

Foreword
viii

Foreword



I am pleased to be able to commend this guidance to
you. It was sponsored and funded by the Civil
Contingencies Secretariat, project-managed by the
Emergency Planning College and written by a team of
specialists in organisational psychology from Leeds
University Business School.  It  is  the product of a  year’s 
research involving a detailed literature review and
primary research with practitioners and specialists in the
field. It summarises our knowledge, articulates our
current understanding of good practice in crowd management and gives
planners clear direction, and supporting information, regarding the safe
assumptions that may be made about crowd behaviour. As such, this
guidance fills what had been a significant gap in our canon of guidance,
and contains information that will be of value to a broad cross-section of
the public safety and resilience community.




Bruce Mann
Director
Civil Contingencies Secretariat

List of Tables

ix













Acknowledgements

Acknowledgements
x

Acknowledgements


The authors would like to extend their thanks to a number of individuals and
organisations for their help and support throughout this research project: -

 Temporary Assistant Commissioner Chris Allison – Metropolitan Police
 Simon Ancliffe – Movement Strategies
 Professor Edward Borodzicz – University of Portsmouth
 Dr John Drury – University of Sussex
 Superintendent Roger Evans – Metropolitan Police
 Gerrard Gibbons – Liverpool City Council
 Superintendent Roger Gomm – Metropolitan Police
 Edward Grant – University of Derby
 Andrew Jenkins – Arup
 Glyn Lawson – University of Nottingham
 Susan Lees – Liverpool City Council
 Mark Leigh – Emergency Planning College
 Dr Rob MacFarlane – Emergency Planning College
 Susan McAdam – Liverpool City Council
 Chief Inspector Peter McGrath – Lothian and Borders Police
 Andrew McNicholl – Liverpool City Council
 Chief Inspector Peter Mills – Sussex Police
 Krisen Moodley – University of Leeds
 Erin Morrow – Arup
 Sergeant Kerry O’Connor – Metropolitan Police
 Superintendent Phil O’Kane – Lothian and Borders Police
 John Parry – Liverpool City Council
 Professor Stephen Reicher – University of St Andrews
 Mike Richmond – Richmond Event Management Ltd; The Event Safety Shop Ltd.
 Ian Rowe – Arup
 Professor Keith Still – Crowd Dynamics
 Sue Storey – Nottinghamshire County Council
 Alastair Stott – Liverpool City Council
 Steven Terry – London Fire Brigade
 Clara Yeung – Arup

 The Emergency Planning College, Easingwold
 Crowd Dynamics Ltd
 Liverpool Culture Company
 Metropolitan Police, New Scotland Yard
 Metropolitan Police Public Order Training Centre, Gravesend
 Lothian and Borders Police

A Guide for Readers
xii













EXECUTIVE
SUMMARY

Executive Summary
2

Executive Summary


 This research was sponsored and funded by the Cabinet Office, as part of the
canon of civil protection literature and guidance, and is published on their UK
Resilience website (http://www.cabinetoffice.gov.uk/ukresilience.aspx).

 For ease of reading, the research has been divided into a series of four, inter-
related reports, namely: -

o Understanding Crowd Behaviours: Guidance and Lessons
Identified

o Understanding Crowd Behaviours: Supporting Evidence

o Understanding Crowd Behaviours: Simulation Tools

o Understanding Crowd Behaviours: Supporting Documentation


 This Executive Summary provides an overview of the whole research project
(i.e., of all four reports), summarising the Research Aims, Methodology, Key
Messages, Good Practice Guidelines, Lessons Identified and
Recommendations for Further Research.

 For completeness, this Executive Summary is included at the beginning of
each report.

 In addition, a separate guide has been prepared for readers of the reports,
which aims to help identify which reports may be of most relevance and use.

o Understanding Crowd Behaviours: A Guide for Readers


 We recommend that anyone with a professional interest in crowd
behaviours should read this Executive Summary.

Executive Summary
3

Research Aims

 To review – and identify gaps in – existing research, theoretical literatures,
and available knowledge on crowds and their behaviour, in both normal and
emergency situations.

 To review how the leading simulation software tools accommodate crowd
behaviours, and consider how approaches to modelling and simulating crowd
behaviours might be enhanced for the future, incorporating both psychological
and technical concerns.

 To identify ways forward for the field of crowd management, particularly in
relation to planning for very large scale crowd events, which will take place
over consecutive days and across multiple locations.

 To produce a set of professional guidelines for emergency planners and
responders, specifying reasonable assumptions which can be made with
regard to crowd behaviours in normal and emergency situations, against
which current assumptions can be tested, and with which future planning can
be informed.



Methodology

 A rigorous methodology was undertaken during this research, to gain a wealth
of information regarding crowds, their behaviours and methods of simulation,
from a wide range of sources (see Understanding Crowd Behaviours:
Supporting Documentation, ‘Research Methodology’, pages 43 to 56).

 In-depth literature reviews examining over 550 academic papers, books and
official reports were carried out (see Understanding Crowd Behaviours:
Supporting Evidence, ‘Part 3 – Review of the Literature’, pages 54 to 242).
These specifically concerned: -

o The key theories of crowd behaviours, with particular focus on the
underlying assumptions and rules governing human behaviour, in both
normal and emergency situations.

o Relevant disasters and mishaps involving crowds, with particular
emphasis on crowd behaviours, and the often interconnected nature of
contributory factors.

o The key methods used to model and simulate crowd behaviours.

Executive Summary
4

 In addition, three of the leading simulation techniques currently available were
reviewed – through utilising accessible literature and conducting interviews
with both users and creators of the tools – focusing on their underlying
behavioural assumptions and rules (see Understanding Crowd Behaviours:
Simulation Tools).

 27 semi-structured interviews were conducted with a wide range of individuals
acknowledged to be experts in the field of crowds and crowd behaviours,
including leading academics, experienced police officers, and key crowd
event and management practitioners (see Understanding Crowd
Behaviours: Supporting Evidence, ‘Part 4 – Expert Interview Findings’, 
pages 243 to 275).

o The interviewees were specifically chosen for their wealth of
experience, ranging from a few to over 30 years. The majority had
over ten years’ experience in the field.  

o They had a range of roles and responsibilities, including overseeing
public order at major events, emergency planning, operational planning
and safety management.

o Experience of major crowd events amongst the interviewees included
Notting Hill Carnival, The Matthew Street Festival, Glastonbury,
Liverpool Capital of Culture 2008, Hogmanay, New Year’s Eve in 
London, large scale marches in London (such as Stop the City, Stop
the War, May Day protests), and events at Wembley Stadium.


 In addition the lead author of this report: -

o Attended two crowd-related courses held at the Emergency Planning
College, on Crowd Dynamics, and on Public Safety at Sports Grounds
and Events.

o Spent a day with police officers at the Metropolitan Police Public Order
Training Centre, Gravesend, and a day with Lothian and Borders
Police during a visit from the Queen.


 Particular attention has been paid to examining very large scale crowd events,
which will take place over multiple days and across multiple sites (see
Understanding Crowd Behaviours: Supporting Evidence, ‘Part 1 – Very
Large Scale Crowd Events’, pages 10 to 21), focusing on: -

o The differences between very large scale, multi-day, multi-site events
and other, more frequent or one-off events, specifically with regards to
preparation and crowd management.

Executive Summary
5

o The new and additional risks that arise in light of these differences and
the findings of this research, which will need careful and rigorous
analysis and mitigation by appropriate professionals.


 Analysis has also been undertaken of the problems occurring at the opening
of Heathrow Terminal 5 (see Understanding Crowd Behaviours:
Supporting Evidence, ‘Part 2 – A Cautionary Tale: Heathrow Terminal 5’, 
pages 22 to 53), since this provides an excellent recent example of a major
infrastructure and operational investment which was badly planned and
managed. There are important lessons to identify from this case study.



Key Messages

The key messages to take away from this report are: -

 A great deal is known about crowds and how to plan for and manage crowd
events. However, this has not been captured and articulated in a single
guidance document until now.

 Key advice for successful crowd management includes: -

o Thorough planning and preparation, using a wide range of “what if...?” 
scenarios, including unexpected scenarios.

o Adoption of a system-wide approach.

o Coordination between all agencies involved.

o Utilisation of personnel who have plentiful first-hand knowledge, skills
and experience in planning for and managing crowd events.

o Communication with the whole crowd – both audio and visual –
particularly in emergency situations.

o Leadership and guidance to initiate crowd movement in emergencies.

o Acknowledgement that seemingly small problems occurring in
combination can have a significant impact on event success.


 Nevertheless, there are significant gaps in our understanding of crowd
behaviours and in the current capability of crowd simulation tools.

Executive Summary
6

 These gaps are exemplified by the special circumstances of very large scale,
multi-day, multi-site crowd events, which will be very different to more
frequent, one-off events in a number of ways and, therefore, are likely to
involve new or additional risks which will require careful analysis and
mitigation.

 In particular, focusing on these very large scale, multi-day, multi-site events,
there is a need to consider the potential risks surrounding: -

o The different types of crowds and their likely behaviours.

o The behaviours of non-ticket holders who will be attracted to the
events, for a range of motives (both legal and illegal).

o The boundaries – i.e., the scope and scale – of the system we are
trying to plan for and manage.

o The range of “what if...?” scenarios that need to be considered.

o The knock-on effects of an incident over consecutive days.

o The importance of coordination between all agencies, across
widespread geographical locations.

o The need to ensure all personnel – from all agencies and in all
locations – are consistently and effectively educated, trained and
briefed, for both normal and emergency circumstances.

o The development of new capabilities and facilities for simulation tools,
in order to accommodate the above issues.


 There are also some important lessons to identify from the experiences of the
Heathrow Terminal 5 opening, in particular that: -

o Combinations of failures in preparation and management can come
together to create major inconvenience to the users of new facilities.

o These factors include apparently mundane failures such as delays in
the completion of the building programme, corner-cutting in training
and familiarisation, initial software problems with new computing
facilities, a failure to listen to the end users, and so on.

o These can happen on such a scale as to represent a public relations
debacle for the companies and authorities concerned and for the UK
more generally.

Executive Summary
7

o Careful preparations need to be made for such everyday
contingencies.



Good Practice Guidelines

 A comprehensive set of good practice guidelines has been collated and
established for all professionals and practitioners involved in the field of
crowds, including crowd events, crowd management, crowd control and
emergency services (see Understanding Crowd Behaviours: Guidance
and Lessons Identified, ‘Guidelines for Good Practice’, pages 10 to 39).
These guidelines focus on: -

o Good practice for crowd management.

 For example, concerned with: thorough planning and
preparation; minor risks combining to create major problems;
multi-agency teamworking; utilisation of experienced personnel;
cross-agency coordination; strategies for communicating with
the crowd; differentiation of different types of crowd; and
awareness of different behaviours from different types of crowd.


o Good practice for emergency situations and evacuations.

 For example, concerned with: leadership and guidance during
an emergency situation; initiating crowd evacuation as quickly
as possible; strategies for communicating with the crowd and
providing information; and awareness of how individuals are
likely to behave during an emergency.


o Good practice for crowd simulation techniques.

 For example, concerned with: trying to model more accurately
crowd movements and behaviours; incorporating different types
of crowd and crowd member; including family or other small
groups within simulation models, rather than just focusing on
individuals; and modelling interactions between crowds and
other groups, and between crowd members.

Executive Summary
8

Lessons Identified

 A comprehensive set of lessons identified has been produced (see
Understanding Crowd Behaviours: Guidance and Lessons Identified,
‘Lessons Identified’, pages 40 to 85), concerning: -

o Definitions and types of crowd.

o Assumptions about crowds – including crowd movement and self-
organisation, crowd behaviours in normal and emergency situations,
crowd disorder, and ways of improving crowd management.

o Ways in which crowds and their behaviours can be simulated.



Recommendations for Further Research

 Recommendations for future research and practice have been suggested (see
Understanding Crowd Behaviours: Guidance and Lessons Identified,
‘Recommendations for Further Research, pages 94 to 134), with the main
priorities concerning further work on: -

o The development of a rigorous risk assessment tool, which will enable
its users to identify the full range of risks associated with different kinds
of events and circumstances involving crowds.

o How new risks associated with the building and subsequent operation
of a range of new facilities and sporting events, over an extended
period, can be managed and mitigated – i.e., drawing on the lessons
that can be identified from an analysis of what is different about very
large scale, multi-day, multi-site crowd events, and of the multiple
problems which contributed to the problematic opening of Heathrow
Terminal 5.

o Stewarding and its impact on crowd behaviours. At present, there
appears to be no research investigating the interactions between
crowds and stewards, despite stewards undertaking a crucial role
during crowd events and often being the first point of contact for crowd
members.

o Individuals who wish to be part of an event but do not have tickets to
attend the event itself – i.e., non-ticketed event crowds – and the
impact which their behaviour has on the preparation for, and overall
management of, an event.

Simulation Tools Reviewed














Simulation Tools
Reviewed

Simulation Tools Reviewed
11

Simulation Tools Reviewed


 This section reviews three of the leading agent-based simulation tools
currently available namely: -

o Legion (for further information see www.legion.com)

o Myriad II (for further information see www.crowddynamics.com)

o Mass Motion


 As these are commercially available tools, access to detailed information was
limited. Therefore, for the following reviews: -

o Legion has been compiled from information available from the public
domain (www.legion.com) and from interviews with users.

o Myriad II has been derived from information available from the public
domain (www.crowddynamics.com) and from an interview with the
creator (Professor Keith Still, Crowd Dynamics).

o Mass Motion has been derived from an interview with the creator (Erin
Morrow, Arup).

Simulation Tools Reviewed














Legion

Legion ?
? 14

Legion Software


? The two main software tools within Legion are : -

? vLegion Studiou C ? ? t fpsue? ahfm? hg? efmUr? ‘fps? ‘ppr? C which enables
pedestrian movement within any defined space to be simulated (latest
version C Legion Studio 2006).





Figure 1. Screenshot from a Legion Studio simulation of crowd movement in
the Sydney 2000 Olympic stadium
(Taken from http://www.legion.com/case-studies/sydney-olympics.php)







? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ᔁ ? ? ? ? ? ? ? ? ᔍ ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ᔁ
most widely adopted, most powerful and most
? ࠈ? ? ? ? ? ? ? ? ? ? ᔖ ? ? ? ? ? ᔎ ? ? ? ? ? ? ? ? ? ᔒ ? ? ? ? ? ? ? ?

(Quote retrieved from
http://www.legion.com/software/studio-2006.php)

Legion
15

o ‘Legion 3D’, which, when used in combination with Studio, can be used
to visualise any simulation model in a three dimensional environment
(latest version – Legion 3D 2006).


Figure 2. Screenshot from a Legion 3D visualisation, showing a large
cityscape with simulated pedestrians and vehicles
(Taken from http://www.legion.com/services/3Danimation.php)





 Legion also now has a software package called ‘Aimsun for Legion’ which 
enables the interface of pedestrians and traffic to be simulated, through an
alliance between Legion and ‘Aimsun’, the leading traffic simulation software
company. This software package enables the accurate simulation of vehicles
and people at road crossings, for example.

Legion ?
? 16

Approaches to Simulation


? Observations of crowds and how they move is crucial to building up a
comprehensive knowledge base of different environments, event types and
crowd behaviours, so that more realistic, accurate models can be designed.

? Talking to those involved with crowds on a daily basis C e.g., event
organisers, crowd control personnel, and station managers C is very helpful
for learning about crowds and how they move in particular environments.

? Modellers need to be experienced and to have observed numerous crowd
events, so that they are able to provide qualitative input into the model.

? Legion personnel have developed a good knowledge of how crowds behave
and move from observing previous/ similar events, and from talking to
experienced event planners and station managers.

? Legion is an agent-based simulation tool, with environment layouts based on
computer-aided design (CAD).

? Each agent can be seen to move through the environment C from an origin to
a destination, weaving through the crowd and performing various activities
and behaviours on the journey C as an individual capable of making
independent decisions.

? Agents move through the environment according to the principle of least effort
C i.e., with minimal time, minimal costs (dissatisfaction and discomfort),
minimal congestion and maximum speed.







? ? ? ? ? ᔒ ? ? ? ? ? ? ? ᔎ ? ? ? ? ? ? ᔁ ? he movement of
pedestrians footstep-by-footstep ? in a
quantitatively verifiable manner ? calculating
how individuals interact with each other and with
the physical obstacles in their environment. ?

(Quote retrieved from
http://www.legion.com/software/studio-2006.php)

Legion
17

 Agents also have the ability to make decisions based on their environmental
circumstances. For instance, if an agent gets off a train and there is a choice
of three escalators, they may consider proximity, how busy each is, and
where each leads in relation to their destination before deciding which
escalator to take.

 Agents are assigned through the available space in the simulated
environment. They move around the space randomly, coming together to
congregate and form groups at particular areas, and then moving apart again,
just as is observed in reality.

 Legion can account for the multi-purpose use of areas within the environment,
for instance, using a particular area for both queuing and as a passageway.

 Random elements of behaviour can be introduced to make the simulation
more realistic, e.g., entity size, speed, age, and luggage.

 Accordingly, there are different algorithms within Legion to accommodate
different types of crowd member, including: -

o Size of people

o Walking speed

o Degree of knowledge of environment or journey to be taken

o Size and quantity of luggage

o Disabilities


 The impact of chance events – for example, the late arrival of a train, or the
closure of an exit – on crowd movement can also be assessed.

 The demand from the crowd to be modelled in the simulation – in terms of
crowd profile, e.g., size, age – is determined by forecasts, knowing the event
schedule, knowing venue capacities, and from observations and knowledge of
the demand at previous or similar events.

 Questions which can be answered through Legion simulations include: -

o Will the venue be able to cope with the projected demand?

o What is the average queuing time at facilities during peak periods?

o Will queues in front of facilities, such as ticket windows or cash
machines, hinder the main crowd flow?

Legion
18

o Which operational scheme enables customers to experience optimal
service?

o In the case of an emergency, can the venue be evacuated safely and
in sufficient time?

o What are the likely crowd densities at bottleneck points, such as main
entrances or stairways?


 Outputs include Fruin’s levels of service, journey times, rates of movement, 
rates of people moving through critical areas of the environment, and number
of people waiting.

Legion
19

Assumptions Underlying Crowd
Behaviours


 Legion is comprised of different user groups – i.e., different types of crowd
and crowd member – which are based on observations of real crowds. These
user groups include :-

o UK commuter

o Hong Kong commuter

o Stadium leavers

o Tourists


 A UK commuter crowd typically has characteristics including: -

o A faster walking speed – mean speed of 1.3 metres per second.

o Small entities – i.e., with little luggage or baggage.

o Travel swiftly from A to B – i.e., good knowledge of their route, good
awareness of the best routes to take, use the same origin and
destination on each journey, and rarely need to buy tickets.


 A tourist crowd, by contrast, typically has characteristics including: -

o A slower walking speed.

o Larger entities – i.e., more baggage.

o Less familiarity with the environment, therefore take longer routes, and
frequently stop to consult signs or maps.


 However, Legion software does allow these user groups to be modified to
include whatever characteristics are most appropriate for the particular group
to be simulated. For example, modifications can be made in terms of walking
speed, age, entity size, size and volume of luggage and restricted mobility
(e.g., pushchairs, wheelchairs).

Legion
21

Evaluation of Legion


Validation


 Legion is continually being validated by observing and analysing crowd
movements at numerous locations and in varying situations (e.g., Berrou et
al., 2005). For example, validation studies have been carried out at the
Monaco Grand Prix (2000, 2001, 2002), in the London Underground and in
Grand Central Station, New York.



Strengths of Legion


 Legion is able to accommodate multiple types of crowd and multiple types of
crowd member.

 Individual agents can be randomly assigned characteristics and attributes to
more accurately represent the population to be simulated.

 Legion has a wide range of uses and is applicable to a wide variety of market
sectors.

 Legion is very easy to observe and understand, therefore making it very easy
to convey a situation to a lay audience.

 Legion is user friendly, with good presentation graphics, which greatly appeals
to customers.

 Lots of statistical information is produced, relating to factors such as levels of
service, journey times, rates of flow, and densities, which can be used to
assist with event preparation.



Weaknesses of Legion


 At present, Legion is less successful at modelling groups of people within a
crowd.

Legion
22

 Legion does not account for the psychological state of an individual, for
instance, levels of stress or emotion.

 The process of simulating a complex environment is very time consuming.

 Much computing power is needed to support the model.

 More qualitative elements need to be built into the Legion software, for
instance, concerning the choice of routes.

Legion
23













Myriad II

Myriad II
24

Applications of Myriad II


 Myriad II was developed by Professor Keith Still and colleagues at Crowd
Dynamics.

 Myriad II is a general purpose crowd analysis tool, applicable to all market
sectors and crowd events studied thus far.

 The main purpose of Myriad II is to test how, when and where the system will
fail, so that preventative measures can be taken and contingency plans be
produced to cope with potential problems during a crowd event.

Myriad II
25

Approaches to Simulation


 Myriad is unique because it models different environments in one, integrated
modelling suite, comprising: -

o Network analysis

o Spatial analysis

o Agent-based analysis


 These three different types of analysis are integrated in one environment,
enabling the best possible model to be produced for the particular situation to
be simulated.

o For example, network analysis would be most appropriate for modelling
the parts of the environment where there are few complex interactions
– e.g., simple roads or corridors – whereas agent-based analysis would
be more appropriate for more complex interactions – e.g., in
concourses.

o Myriad II is able to replace the appropriate parts of the network model
with an agent-based model, and data can then be passed between the
two. Thus, the overall simulation integrates a network model and an
agent-based model, to represent the parts of the environment without
complex interactions and with complex interactions, respectively.


 Real time counts – e.g., flow rates, densities, and ingress and egress rates –
can be taken at an event and added into a Myriad II model. This enables
early assessments of crowd dynamics, such as flow rates and densities, to be
calculated and, subsequently, for appropriate measures to be taken to alter
these dynamics, thereby improving crowd safety.

Myriad II
26

Figure 3. The Myriad II modelling suite
(Taken from http://www.crowddynamics.com/technical/)

Myriad II
27

Network Analysis


 The network analysis system is design to simulate crowd movements through
large complex spaces.

 It comprises a ‘buckets and pipes system’, where real-time crowd flow rates
along differing routes – i.e., pipes – can be monitored and the time taken to fill
and empty specific areas – i.e., buckets – can be assessed, along with the
time taken for the system to potentially over-fill and, subsequently, fail.

 Network analysis is a basic modelling tool, therefore all individuals should be
able to use it.


Figure 4. Screenshot of a Network Analysis simulation, with different colours
used to indicate the optimal flow capacities for alternative exit routes
(Taken from http://www.crowddynamics.com/technical/)

Myriad II
29

Agent-Based Analysis


 The agent-based analysis system is more appropriate for use in complex
environments and interactions, and is designed to simulate density, speed,
agent location and space utilisation in a given environment.

 The system is comprised of individual, autonomous agents, each with
individual speeds, densities and attributes. Thus, each agent is capable of
scanning, seeing and reacting to the environment.

 Each agent attempts to get from A to B in the fastest possible time according
to the principle of least effort – i.e., avoiding high density areas, and covering
the shortest distance in the shortest time, with maximum speed.

 Due to the complex nature of this analysis system, it should only be used by
experienced modellers.


Figure 6. Screenshot from an Agent-Based Analysis simulation, showing the
movement of individuals through a complex environment
(Taken from http://www.crowddynamics.com/technical/)

Myriad II
30

The Myriad II Suite


 The Myriad II suite is designed to test boundary conditions as opposed to
specific circumstances, in order to assess how different crowd compositions
are likely to affect the fundamental parameters in the model – i.e., flow rates,
density, ingress, circulation and egress – in both normal and emergency
conditions.

o For example, what is the likely impact on these parameters if 5% of the
population are elderly? What is the likely impact on these parameters
if 2% of the population are aggressive?


 In order to test these boundary conditions, Myriad II uses a flux algorithm,
which enables the impact of the numerous potential variations in crowd
composition to be more easily assessed.

o The more specifically characteristics of individuals within a model are
defined, the more combinations of different crowd compositions are
possible, and the more models are therefore needed to test out the
different possibilities.

o Using a flux algorithm, however, means that only one model is needed
to assess these different possibilities. The flux algorithm enables every
member of the crowd to possess the different characteristics to be
tested at some point during their attendance at the event.

o Thus, the overall speed-density distribution and composition of the
crowd is maintained, but specific attributes are randomly shuffled
between individuals within the crowd.




“You can’t predict to the nth degree what any 
individual might do, and I don’t think there’s 
any point in trying to.  There’s a huge benefit 
in understanding the limits of crowds, such as
flow rates, density and the point of collapse,
but there’s no additional benefit in breaking 
that down further and further.”

Professor Keith Still
Crowd Dynamics

Myriad II
31

 The key factors involved with successful crowd management which are
considered by Myriad II are: -

o Ingress, circulation, egress

o Design, information, management

o Flow, fill, fail

o Speed-density distribution

o Packing coefficient


 Currently, a new part of the Myriad suite – called “event planner” – is being
trialled. This system is designed to visualise risk as a dynamic, according to a
red-amber-green timeline. Risk changes in time, size, location and shape,
therefore it is important to have a more dynamic means of assessment.

Myriad II
32

Assumptions Underlying Crowd
Behaviours


 Influences on crowd behaviours can only manifest themselves in one of four
ways, namely, objective, motility, constraint and assimilation (Still, 2000).
These have been found to be very robust over the years and have needed
very little refinement. They cover all possibilities, e.g., event types, crowd
types, attributes, etc.


1. Objective

 What are the individual’s objectives?  How is he or she going to move?

 Objectives can only cause an individual to move in a certain direction
or to remain still, and depend on factors such as information, signage,
management and location geometry.


2. Motility

 The rate at which an individual can move.

 Motility is a function of human dynamics – i.e., acceleration,
deceleration, and speed of movement – and is dependent on factors
including route conditions, weather conditions and crowd composition.


3. Constraint

 Constraints are factors which act on the system to restrict crowd
movement.

 For example, increased crowd density and increased location
complexity both decrease crowd flow.


4. Assimilation

 This is the time it takes for people to take information onboard and
react to it.

 Assimilation depends on issues such as communication system,
management strategy and composition of the crowd.

Myriad II
33

Evaluation of Myriad II


Validation


 Myriad II is continually being validated against video footage and observations
made in the field.

 The assumptions made by Myriad II are constantly being refined as a result of
more up-to-date data and information. This refinement is vital to ensure a
more accurate model is produced.



Strengths of Myriad II


 The main strength of Myriad II over other simulation tools is the use of the
three integrated modelling tools in one environment – i.e., network analysis,
spatial analysis and agent-based analysis.

 This integration means that an environment can, for instance, be modelled
primarily as a network with more complex sections of the data integrated into
an agent-based model.

 It is this multi-scalar modelling, enabling different environments to be
simulated in one, integrated suite that makes Myriad II unique.

 Models can be set up quickly and results can be obtained quickly.



Weaknesses of Myriad II


 The complex nature of the Myriad II suite – incorporating network, spatial and
agent-based modelling – means that the user – i.e., the model builder – needs
to have a broad background in modelling, in order to understand the differing
modelling techniques and, therefore, to use to tool as appropriate.

Myriad II
34













Mass Motion

Mass Motion
36

Approaches to Simulation


 Mass Motion is a 3D agent-based simulation tool, populated by individual,
autonomous agents capable of making independent decisions in order to
achieve a goal.


Figure 7. Screenshot from a Mass Motion simulation of crowd movement at
Union Station in Toronto
(Image courtesy of Erin Morrow, Arup, creator of Mass Motion)





 Each agent has a position, an orientation, and a velocity.

 A key part of all Mass Motion modelling is that all agents have a goal – i.e., to
achieve a task in the minimum time possible, such as getting from A to B, or
exiting a building.

 Agents are aware of physical constraints around them, such as walls, and
have a cone of vision in which they are able to see other agents. Taking
these factors into consideration, each individual agent makes a best guess of
the way forward five times per second – i.e., almost continuously – as occurs
automatically in reality.

Mass Motion
38

Assumptions Underlying Crowd
Behaviours


 Mass Motion primarily distinguishes two different types of crowd: -

o A commuter crowd, where published data on walking speeds is used to
inform model parameters

o An evacuation crowd, where the rules are simpler.


 Within a commuter crowd, a further distinction is made between tourists and
expert commuters, and the percentages of these within a particular simulation
can be varied. Moreover, the percentage of tourists who need to follow signs
can fluctuate, as can the percentages who are just looking for a platform, who
want to do some shopping or who are seeking a bathroom.


Figure 8. Screenshot from a Mass Motion simulation of crowd movement at
Transbay Terminal in San Francisco
(Image courtesy of Erin Morrow, Arup, creator of Mass Motion)

Mass Motion
39

 Differences in individual agents are assigned randomly on a distribution curve.
For example, some people walk faster than others, some are more averse to
congestion, and some are keen to minimise the distance they travel.

 Mass Motion also utilises two kinds of rules for commuter crowds: -

o Higher order rules, where people are given an overall target, such as to
get to point A. Within that overall target there are additional rules, for
example, if area X is congested, deviate to area Y.

o Local rules, which are more reflexive and involve local decisions to
determine the best way for individuals to move through an
environment.


 Different rules are used to underpin movement in evacuation crowds. For
example, crowds in an evacuation situation are less likely to consider
alternative options preferring to behave in the most obvious way, and are
more likely to follow others.

Mass Motion
40

Evaluation of Mass Motion


Validation


 Much work has been carried out in order to validate Mass Motion. For
example, at Toronto Union Station, city data and census data were used to
plot and simulate the flows of commuters. Each exit door and platform was
then surveyed to show that the simulation matched the actual pattern of crowd
behaviours to within a 5 % error.

 Other validation has been done on a more localised scale, for instance,
looking at how many people take an escalator during a certain time period.



Strengths of Mass Motion


 The main benefit of Mass Motion is its ability to simulate the way in which
crowd members think, for example, concerning entries and departures.
Consequently, usage and flow patterns, as a result of emergent behaviours,
can be modelled, without the need to input the pathways into the model
initially. No other simulation tools do that.

 A further strength of Mass Motion is its capacity to model the wider
environment, enabling the movement of an agent throughout the system as a
whole to be simulated, as opposed to movement at one specific location.



Weaknesses of Mass Motion

 To the best of our knowledge, Mass Motion is currently lacking in the following
areas: -

o It does not account sufficiently for groups of individuals moving around
a crowd event.

o Improvements need to be made with regards simulating people with
disabilities and large amounts of luggage.

o It does not consider individuals’ emotions and the impact which this 
can have on their movement and behaviour.

Mass Motion
41













Future
Simulation Tools

Future Simulation Tools
42

Future Simulation Tools


The interviewees agreed that key areas for future simulation tools to focus on
include: -

 Developing 3D agents and environments.

 Improving the speed and ease of use.

 Producing quantifiable data.

 Accurately reflecting the different characteristics of different types of crowd
and types of crowd member.

 Using research evidence to underpin the choice of characteristics and
behavioural assumptions for different types of crowd and crowd member.

 Simulating the behaviour of groups within crowds.

 Incorporating individuals’ emotions into simulation models, such as stress, 
frustration and patience.

 Modelling the interaction between people and traffic – crowds do not exist in
isolation, and it is important to examine the interface between crowds and the
different elements with which they interact.

Key Learning Points
43













Key Learning
Points

Key Learning Points
44

KEY LEARNING POINTS

– Simulation Tools –

 Real-time observations of crowds and how they move, in
addition to talking to experts involved with crowds first-hand
on a regular basis, is vital to develop a realistic simulation
model.

 Simulation tools can be used to assist with issues such as
design, safety and security, and strategic planning, for
market sectors including transport, retail, sports and the
public realm.

 3D software tools offer the most realistic visualisation of an
environment.

 The most realistic simulation tools are populated by
intelligent, autonomous agents, capable of making
independent decisions and reacting to environmental
conditions.

 The principle of least effort appears to be the most utilised
algorithm underpinning agent movement, where agents
move so as to minimise time, costs and congestion whilst
maximising speed.

 Different types of crowd, with different characteristics, are
acknowledged within the simulation tools, based upon
observations and experience rather than research literature.
For instance, commuter crowds, tourist crowds and
evacuation crowds.

o There does not appear to be a set number of crowd
types in each simulation tool – i.e., it is not possible to
say that Legion, for example, has X crowd types.

o The characteristics of these key crowd types can be
modified to accommodate the type of crowd required.

Key Learning Points
45




 Agents can be randomly assigned individual attributes, such
as size, gender, age, luggage, walking speed, disabilities,
and familiarity with the environment.

o There does not appear to be a set number of crowd
members types in each simulation tool – i.e., it is not
possible to say that Mass Motion, for example, has X
crowd member types.

o The simulation tools appear flexible and able to
accommodate differing types of crowd member

 Assumptions are made regarding likely crowd behaviours in
particular environments – based on observations and
experience of crowds – such as how early crowds will arrive
for an event, at what speed and in which direction
individuals are likely to move, and where people are most
likely to congregate.

o There does not appear to be a fixed number of rules
underpinning crowd behaviour – i.e., it is not possible
to say that Myriad II, for example, has X rules relating
to crowd behaviour

o The simulation tools appear flexible and able to adapt
in order to accommodate anticipated crowd
behaviours in specific circumstances.

 Simulation tools are continually being validated by
observing and analysing crowd events.

 A key weakness of current simulation tools is the vast
amount of time and computing power they require.

 Future simulation tools should aim to include: -

o Groups of people within a crowd.

o Emotions of individuals.

o Interface between people and traffic.

Published by the Cabinet Office and available from:
Online:
www.cabinetoffice.gov.uk/ukresilience
Mail, Fax & Email:
Emergency Planning College
The Hawkhills
Easingwold
York
YO61 3EG
Telephone: 01347 825000
Fax: 01347 822575
Email: epc.library@cabinet-office.x.gsi.gov.uk
© Crown Copyright 2009
ISBN 978-1-874321-22-4
Published: June 2009