Science (Communication) and Wikipedia - Potentials and Pitfalls
PhD dissertation QUA_SI 2014
1. Empirical Studies and Computational Results of a!
Proxemic-based Model of Pedestrian Crowd Dynamics!
Information Society Ph.D. Program
Department of Sociology and Social Research
February 6th 2014
Andrea Gorrini
Ph.D. candidate of Information Society
University of Milano-Bicocca
2. CSAI - Complex Systems and Artificial Intelligence research center!
L.Int.Ar - Artificial Intelligence Laboratory / Crowdyxity s.r.l.!
Department of Computer Science!
University of Milano-Bicocca (Milan, Italy)!
Supervisor: Prof. Stefania Bandini!
ITS - Institute of Transport Studies!
Monash University (Melbourne, VIC, Australia)!
Supervisor: Prof. Majid Sarvi!
!
RCAST - Research Center for Advanced Science and Technology!
The University of Tokyo (Tokyo, Japan)!
Supervisor: Prof. Katsuhiro Nishinari!
Empirical Studies and Computational Results of a
Proxemic-based Model of Pedestrian Crowd Dynamics
3. Empirical Studies and Computational Results of a
Proxemic-based Model of Pedestrian Crowd Dynamics
! Interdisciplinary Needs:
* Crowd Dynamics
* Proxemic Theory
* Methodological Approach
• Empirical Case Studies
• Modeling and Simulations
February 6th 2014
OUTLINE
4. The investigation of pedestrian crowd dynamics
is a complex field of study that requires
interdisciplinary efforts (e.g., social science,
computer science, traffic engineering, applied
mathematics, urban planning).
The use of computer-based simulations allows to
investigate those scenario that are difficult to be
directly observed (what-if scenarios), and to
provide practical results to enhance the spatial
efficiency of mass gathering-transit facilities, both
in terms of service, comfort and safety:
• Kiss Nightclub fire 2013, Brazil
• Love Parade 2010, Germany
INTERDISCIPLINARY NEEDS
Motivations
5. 13.05.2013
Since the pioneering study of Gustave Le Bon (1897) the definition of crowd dynamics
is still controversial (considering both ordinary and emergency situations)
due to the lack of standard guidance for data collection, ethical and practical restrictions
and the variability of the phenomenon.
“A crowd can be defined as a gathering of people standing in close proximity to observe a specific event, !
who feel united by a common social identity and who are able to act in a socially coherent way, despite being
strangers in an ambiguous or unfamiliar situation.” (Challenger et al., 2009 – UK Cabinet Office)!
INTERDISCIPLINARY NEEDS
What is a Crowd?
6. According to the Le Bon’s approach,
as anonymous members of a crowd, people lose
their sense of self-awareness and personal
identity, with antisocial and irrational behaviors.
More recent approaches proposed a social
normative definition of crowd behavior, even in
emergency situations.
• Contagion-Transformation Theory
(Mass Panic Approach)
• Elaborated Social Identity Model
• Emergent Norm Theory
• Affiliative Approach
INTERDISCIPLINARY NEEDS
Theoretical Framework
7. 13.05.2013
We propose to analytically study pedestrian crowd dynamics, focusing on the
empirical investigation of proxemic behavior, grouping and the level of density in the environment.
This is finally aimed at supporting the validation of computer-based modelling and simulations.
“Crowds can be defined as complex systems that comprise many interacting parts with the ability to generate
emergent collective behavior through self-organization and self-driving feedback loops” (Meyers, 2009). !
INTERDISCIPLINARY NEEDS
What is a Crowd?
8. 13.05.2013
In analogy with territorial behavior in animals, E.T. Hall (1966) introduced the term proxemics
for the study of human spatial behavior during social interaction.
Proxemics is a type of nonverbal communication, based on the definition of
four interaction zones: intimate, personal, social and public distances.
!!
INTERDISCIPLINARY NEEDS
Proxemic Theory
intimate distance!
personal distance!
social distance! public distance!
0.45 m! 1.20 m! 3.60 m! 6 m!
9. 13.05.2013
Personal space is the area immediately surrounding individuals, into which strangers
cannot intrude without arousing discomfort. The condition of spatial restriction
in high density situations is linked with the negative stress response of crowding.
INTERDISCIPLINARY NEEDS
Personal Space
10. 13.05.2013
In static situations, the functional space shared by group members (F-formation)
assumes different configurations to facilitate social interaction. Depending on the level of density,
the proxemics behavior of walking groups generates typical patterns of spatial arrangement
(line abreast, V-like, river-like pattern).
INTERDISCIPLINARY NEEDS
Group Proxemic Behavior
11. The proposed methodological approach can be represented as a virtuous cycle:
synthesis (modelling and simulation) and analysis (interpretation of results and comparison
with descriptive sets of empirical metrics and parameters for sake of model validation).
12. • Interdisciplinary Needs
! Empirical Case Studies:
* The Admission Test of the University of Milano Bicocca
* The Vittorio Emanuele II gallery
* The Impact of Turning paths and Grouping
* Pedestrian Personal Space
• Modeling and Simulations
Empirical Studies and Computational Results of a
Proxemic-based Model of Pedestrian Crowd Dynamics
February 6th 2014
OUTLINE
13. Data Collection
Admission Test of the faculty of Psychology
of the University of Milano-Bicocca (Milan, Italy),
1st September 2011 (7:30 am – 10:00 am).
Video footages from a zenith point of view and
three different locations, avoiding images
distortion, trajectories occlusion and to not
influence subjects’ behavior.
Data Analysis
Comparison among on site data collection and
manual people counting from video-footages
(4% and 10% over estimation errors about total
number of pedestrians and groups)
EMPIRICAL CASE STUDIES
In Vivo Observation
14. Results
• No. 1897 people were manually counted
• cyclical up-down peak levels
• level of service A (5.09 ped/min/m): free flows
in situations of low density
• groups: 66% of the total flow
• groups walked line-abreast or V-like pattern
Sample: 50 singles, 50 couples, 17 triples
• groups walked 9% slower than singles
• no gender difference in walking speed
• no spatial layout difference in walking speed
EMPIRICAL CASE STUDIES
In Vivo Observation
15. Data Collection
Vittorio Emanuele II gallery (Milan, Italy),
24th November 2012 (2:50 pm - 4:10 pm).
Video footages from a zenith point of view
(balcony of the gallery) to avoid images distortion,
trajectories occlusion and to not influence
subjects’ behavior.
Data Analysis
Alphanumeric grid superimposed on images
to manually perform data analysis
EMPIRICAL CASE STUDIES
In Vivo Observation
16. Results
• No. 7773 people were manually counted
• cyclical up-down peak levels
• level of service B (7.78 ped/min/m): irregular
flows in situations of low-medium density
• groups: 84% of the total flow
• groups walked with line-abreast or V-like pattern
Sample: 30 singles, 15 couples, 10 triples
and 8 groups of four members
• the trajectories of singles were 4% longer
than group members
• groups walked 37% slower than singles
• couples walked 41% less disperse (35 cm)
than larger groups (centroid method)
EMPIRICAL CASE STUDIES
In Vivo Observation
FlowRate(pedestrian/minute/meter)!
17. Investigating the impact of turning paths (0°, 45°,
60° and 90° degrees) and grouping on normal
crowd egress flows, tested in laboratory setting
(12th April 2012, Monash University, Melbourne).
Hypothesis
In high-density situations, the flow rate and
walking speed is negatively affected by the
increase in turning angle
Sample
No. 68 subjects, spontaneously organized into
groups. We focused on 15 singles and 4 couples,
1 triple and 1 group of four members.
EMPIRICAL CASE STUDIES
In Vitro Experiment
18. Results
• level of service of E (74.18 ped/min/m):
irregular flow in condition of high density
• the angle path with 60° is 12% less effective
compared to the 0°-45° angle path degrees
• the walking speed of group members was
11% lower than the one of single pedestrian
• the angle path with 60° has a negative impact
on the walking speed of singles and groups
compared to the 0°-45° angle path degrees
EMPIRICAL CASE STUDIES
In Vitro Experiment
19. Data Collection
Measuring the size of the front zone of personal
space in static and motion situations (8th June,
2013, The University of Tokyo, Japan).
Hypothesis
Depending on walking speed, the front zone
of pedestrian personal space is larger than the
one in static situations.
Sample
20 male subjects, aged from 18 to 25 years old
(17 Japanese, 2 Vietnamese and 1 Chinese).
EMPIRICAL CASE STUDIES
In Vivo Observation
20. stop-distance procedure!
approach-distance procedure!
locomotion-distance procedure!
Procedure
Participants were randomly coupled and asked to
stop the approach of the each other when they felt
uncomfortable about spatial nearness.
Experimental procedures: A stop-distance,
B approach-distance, C locomotion-distance.
To test the impact of speed (low: 0.93 m/s,
medium: 1.23 m/s, high: 1.46 m/s), participants
were asked to walk following footmarks drawn
on the floor and to synchronize their gate to
metronome background sounds.
EMPIRICAL CASE STUDIES
In Vitro Experiment
30 cm!
30 cm!
20 cm!
20 cm!
stop-distance procedure- low speed related!
21. Results
A. the size of personal space in static situations
is affected by the difference in walking speed
B. the size of personal space moving towards
a stationary person is not affected by speed
C. the size of pedestrian personal space moving
towards an oncoming pedestrian is affected by
speed and it is larger than the one in static
situations (procedure A, low speed related)
* Cultural differences (Hayduk, 1983)
EMPIRICAL CASE STUDIES
In Vitro Experiment
59 cm 72 cm
22. • Interdisciplinary Needs
• Empirical Case Studies
! Modeling and Simulation:
* MAKKSim Simulation Platform
* Simulation Campaign Execution
Empirical Studies and Computational Results of a
Proxemic-based Model of Pedestrian Crowd Dynamics
February 6th 2014
OUTLINE
23. Simulation Approaches
Pedestrian as particles (Helbing, 2001), cells
(Nishinari et al., 2004) and agents (Ferber, 1999).
Computational Model Validation
Computer-based simulations allow to study complex
social systems, envisioning those phenomena that
are difficult to be directly observed in real case
scenarios: what-if scenarios.
Models have to be validated comparing results with
benchmark scenarios: fundamental diagram
(density, walking speed), trajectories and space
occupation, representation of crowd phenomena
(e.g., lane formation). Data related to group
phenomena are scarce.
MODELING and SIMULATIONS
Models Validation
24. Simulation platform MAKKSim
The computational model represents a crowd
as a system of reactive autonomous agents that act
and interact in a shared environment, achieving some
individual or collective goals
Computational Model
The environment is discretized into squared
cells (spatial markers, floor field approach).
The agents are driven by the defined utility function
and proxemic behavioral rules:
• avoid physical contact
• maintain spatial cohesion among members
MODELING and SIMULATIONS
MAKKSim Platform
25. Simulation campaign execution
Validation of the group cohesion mechanism
introduced in the model for representing group
proxemic behavior (LOS B and LOS D)
Results
• the spatial dispersion among group members in
situation of irregular flow is consistent with
reference to the tested scenarios
• quite similar outcomes on trajectories and
walking speed, compared to the data collected
at the gallery (LOS B)
MODELING and SIMULATIONS
Simulation Campaign Execution
LOS B
LOS D
26. FINAL REMARKS
Thesis Work Flow
Innovative Contributions
• systematic review of the theoretical framework
about crowd dynamics and proxemics behavior
in pedestrian dynamic
• interdisciplinary methodology for the study
of pedestrian crowd dynamics
• empirical investigation of the impact of
proxemics, grouping and density (flow rate,
trajectories, speed, personal space, spatial
layout, group arrangement and dispersion)
27. Empirical Studies
• grouping and crowding
• human-animal comparative studies of crowd
evacuation dynamics
Modeling
• personal space, mean density map and crowding
• heterogeneous walking speed
• cognitive agents (tactical level, way finding)
• automated techniques for data collection
FINAL REMARKS
Future Works
28. Empirical Studies and Computational Results of a
Proxemic-based Model of Pedestrian Crowd Dynamics
" Gorrini, A., Shimura S., Bandini, S., Ohtsuka, K. and Nishinari, K. (2014). An Experimental Investigation of
Pedestrian Personal Space: Towards Modeling and Simulations of Pedestrian Crowd Dynamics. Transportation Research
Board 93rd Annual Meeting, Washington DC, US (accepted). !
" Gorrini, A., Bandini, S., Sarvi, M. (2014). Groups Dynamics in Pedestrian Crowds: Proxemic Behavior Estimations.
Transportation Research Board 93rd Annual Meeting, Washington DC, US (accepted). !
!
" Bandini, S., Gorrini, A., Vizzari, G. (2013). Towards an Integrated Approach to Crowd Analysis and Crowd Synthesis:
a Case Study and First Results. Journal of Pattern Recognition Letter - http://dx.doi.org/10.1016/j.patrec.2013.10.003. !
!
" Gorrini, A., Bandini, S., Sarvi, M., Dias, C., Shiwakoti, N. (2013). An empirical study of crowd and pedestrian
dynamics: the impact of different angle paths and grouping. Transportation Research Board, 92nd Annual Meeting,
Washington DC, US, p.42.!
!
" Bandini, S., Gorrini, A., Manenti, L., Vizzari, G. (2012). Crowd and Pedestrian Dynamics: Empirical Investigation and
Simulation. 8th International Conference on Methods and Techniques in Behavioral research - Proceedings of Measuring
Behavior 2012, Utrecht, Netherlands, 308-311. !
!
" Federici, M.L., Gorrini, A., Manenti, L., Vizzari, G. (2012). An innovative scenario for pedestrian data collection:!
the observation of an admission test at the Uni- versity of Milano-Bicocca, 6th International Conference on Pedestrian!
and Evacuation Dynamics - PED 2012, Zurich, Switzerland (in press). !
29. Thank You
Information Society Ph.D. Program
Department of Sociology and Social Research
February 6th 2014
Andrea Gorrini
Ph.D. candidate of Information Society
University of Milano-Bicocca