Page 1
Crowd Dynamics: SimulatingMajor Crowd Disturbances
This is a joint work with Piper Jackson, PhD, Andrew Reid, PhD Student, Vijay Mago, PhD and Vahid , PhD
Valerie Spicer, PhD and Hilary Kim Morden, PhD StudentModelling of Complex Social Systems - MoCCSy
CCJA-ACJA October 2013
Page 2
Mathematicians
Criminologists
Computer scientists
Crowd management practitioner
Group Composition
Page 4
Literature review
• LeBon (1960) Group mind / psychological crowds
• Zimbardo (2007) De-individuation theory
• McPhail (1991) Crowd crystals
• Stott, Hutchison, & Drury (2001) Hooligans/ESIM
• Forsyth (2006) 6 factors of collective behaviour
• McHugh (2010) Emotions of body movement
Page 5
Modeling Project
• Social dynamics
• Macro factors – Fuzzy Cognitive Map (FCM)
• Micro factors – Cellular Automata (CA)
• Threshold analysis: Major crowd disturbance
Page 6
Crowd Psychology
A people behaviour: Disruptive
B people behaviour: Observers Participants
C people behaviour: Guardians
Page 7
Macro Factors• Effective social control mechanisms
• Police – city – transit • Structured environmental factors
• Road design – event location• Unfavourable situational factors
• Suitable target – podiums in the environment• Unstructured technological connectivity
• Text messaging – Twitter – Facebook • Volatile demographics
• Younger people – intoxication – gender distribution • High risk event
• Divisive event – non-family oriented
Page 8
Creating the Fuzzy Cognitive Map
• Group process – used surveys
• Requiring further definition of factors
• Started with 26 factors reduced to 6 factors
• Verified definitions and strengths with independent group member
Page 9
Creating the FCMEnter here: to (affected)
c1 c2 c3 c4 c5 c6 P(incident) Please Enter:
c1 1 Very Low
c2 2 Low
from c3 3 Moderate
(affecting) c4 4 High
c5 5 Very High
c6
Increases
Words: to (affected) Decreases
c1 c2 c3 c4 c5 c6 P(incident)
c1 c1 Cohesive Social Control Mechanisms
c2 c2 Structured Environmental Factors
c3 c3 Unfavourable Situational Factors
(affecting) c4 c4 Technological Connectivity
c5 c5 Volatile Demographics
c6 c6 Risk of Event
Colour to (affected)
c1 c2 c3 c4 c5 c6 P(incident)
c1 0 0 0 0
c2 0 0 0 0 0
from c3 0 0 0 0 0
(affecting) c4 0 0 0 0 0
c5 0 0 0 0
c6 0 0 0 0
Page 10
FCM – CA relationship
Page 11
Micro Interactions – CA model
• Each cell has a stable character• A type person• B type person• C type person
• Each cell has a disruptive risk• -1 ↔ disruptive• 0 ↔ observing - susceptible• 1 ↔ active guardianship
A (-0.8)
A (-0.5)
B (-0.1)
C (+ 1) C (+0.5)
Page 12
Disruptive to Guarding
Page 13
Fuzzy Transitions
• 9 rules: one for each combination:
{A, B, C} {Disrupting, Observing, Guarding}
• All rules applied fuzzily each iteration
• Takagi-Sugeno-Kang: Each rule is a mathematical function, e.g., f(x, y) = y - x
Page 15
CA transition rules
DeterioratingA, B Disruptive:-rn2
exponential negative
PreventingA, B Guarding,C Disruptive: rn2
exponential positive
BoredomB Observing:-rsp(s) linear inward
RespectingA, C Inactive,C Guarding: 0no interaction
Page 16
Results – Unfavourable FCM
Page 17
Results – Unfavourable FCM
Page 18
Results – Favourable FCM
Page 19
Results – Favourable FCM
Page 20
Results – More A Types
Page 21
Results – More A Types
Page 22
Results – Fewer A Types
Page 23
Results – Fewer A Types
Page 24
Future Directions• Model Adjustments to enhance precision
• FCM expansion – factor interaction
• CA modification – non-adjacent cell influences
• Data testing and further validation of model
• Verification with crowd control experts
Page 25
Crowd Dynamics: Simulating Major Crowd Disturbances
Valerie Spicer, SFU [email protected] Hilary Kim Morden, SFU [email protected] Patterson, VPD [email protected]
Andrew Reid, SFU [email protected] Piper Jackson, SFU [email protected]
Vahid Dabbaghian, SFU [email protected] Mago, SFU [email protected]
Page 26
Crowd Dynamics: Simulating Major Crowd Disturbances
QUESTIONS?