empirical investication on pedestrian crowd dynamics and grouping
TRANSCRIPT
Empirical Investigation on Pedestrian Crowd Dynamics and Grouping
Andrea Gorrini1, Stefania Bandini2, Giuseppe Vizzari2
1 Information Society Ph.D. Program, Department of Sociology and Social Research University of Milano-Bicocca, Italy
2 Complex Systems and Artificial Intelligence Research Center (CSAI)University of Milano-Bicocca, Italy
TGF 2013 - Jülich Supercomputing Centre - Sept. 25-27, 2013
Outline
• Why observing groups? Motivations and fundamental research question
• Available literature, data and observations on group influence on pedestrian behaviour
• An observation and analysis of group influence in Vittorio Emanuele II Gallery in Milano
• Towards an interdisciplinary research on pedestrian and crowd analysis and synthesis
• Conclusions and discussion
TGF 2013 - Jülich Supercomputing Centre - Sept. 25-27, 2013
TGF 2013 - Jülich Supercomputing Centre - Sept. 25-27, 2013
Crowds of pedestrians as complex systems
• Overall system behaviour depends on individuals’ decisions and actions…• … that are generally influenced
by a large number of factors• … intertwined in an often
unpredictable way• Mixed and conflicting mechanisms
• Competition for the shared space…
• … but also cooperation (non written social norms) to prevent stall situations
• Imitation...• ... but also natural tendency to
stay at a distance (proxemics)• Emergent phenomena • …
TGF 2013 - Jülich Supercomputing Centre - Sept. 25-27, 2013
Impact of groups in pedestrian and crowd dynamics
• Current approaches to modeling and simulation generally consider every pedestrian as a individual with almost no relationships• Considering only his/her own
goals• Considering other pedestrians
as moving obstacles or movement opportunities
• Nonetheless, in several situations pedestrians are bound by relationships influencing their movement• Generally speaking, a crowd is
made up of groups of pedestrians...
• What do we miss by neglecting this aspect of pedestrian behaviour?
TGF 2013 - Jülich Supercomputing Centre - Sept. 25-27, 2013
Groups in the literature - Observations
• At least two studies report observations about groups
• Willis A, Gjersoe N, Havard C, Kerridge J, Kukla R, 2004, "Human movement behaviour in urban spaces: implications for the design and modelling of effective pedestrian environments" Environment and Planning B: Planning and Design 31(6) 805 – 828
• Michael Schultz, Christian Schulz, and Hartmut Fricke. “Passenger Dynamics at Airport Terminal Environment”, Pedestrian and Evacuation Dynamics 2008, Springer-Verlag, 2010
• Observations carried out in low density conditions
• Groups of small size were most frequently observed
• Little data about dynamic group proxemic behaviour
TGF 2013 - Jülich Supercomputing Centre - Sept. 25-27, 2013
Admission test University of Milano-Bicocca
• Admission test of the Faculty of Psychology at the University of Milano-Bicocca - September 1, 2011
• Counting activity supported by video footages of the event
• About two thousand students attended the test• About 34% individuals, 50%
couples, 13% triples and 3% groups of 4 members (!)
• Statistically validated relationship between group size and velocity• Additional quantitative analyses
about the arrival and entrance process, LOS
• Qualitative analysis of group shapes and related phenomena
• More details in PED and ACRI (C&CA) 2012 papers
TGF 2013 - Jülich Supercomputing Centre - Sept. 25-27, 2013
TGF 2013 - Jülich Supercomputing Centre - Sept. 25-27, 2013
Scenario Analysis The Vittorio Emanuele II gallery is a popular commercial-touristic walkway situated in the Milan city centre; it represents an optimal scenario for video recording pedestrian flows
TGF 2013 - Jülich Supercomputing Centre - Sept. 25-27, 2013
TGF 2013 - Jülich Supercomputing Centre - Sept. 25-27, 2013
Data CollectionThe survey was performed the 24th of November 2012 from 2:50 pm to 4:10 pm. Thanks to the official authorization of the Municipality of Milan, the video footages
were performed from the balcony of the gallery; the staff was composed of four
observers; the equipment consisted of two video-camera with tripods
TGF 2013 - Jülich Supercomputing Centre - Sept. 25-27, 2013
TGF 2013 - Jülich Supercomputing Centre - Sept. 25-27, 2013
Data Analysis The survey was aimed at collecting data about: level of density and walkway level of
service; presence of groups; group size and proxemics spatial patterns; trajectories;
walking speed; group proxemics dispersion
TGF 2013 - Jülich Supercomputing Centre - Sept. 25-27, 2013
TGF 2013 - Jülich Supercomputing Centre - Sept. 25-27, 2013
Data Reduction and Analysis A selected portion of the gallery was considered for the analyses (12.8m – 12.8m). A grid of square cells (designed by using Adobe Photoshop CS5) was used in order
to discretize the environment in cells and to perform data analyses
TGF 2013 - Jülich Supercomputing Centre - Sept. 25-27, 2013
TGF 2013 - Jülich Supercomputing Centre - Sept. 25-27, 2013
1. Results – Fruin’s LOS The bidirectional pedestrian flows were measured every minute; the total flow was composed of 7773 people; the level of density was low; the average walkway
level of service corresponded to the B level (7.78 ped/min/m) that correspond to an irregular flow in low density situations
TGF 2013 - Jülich Supercomputing Centre - Sept. 25-27, 2013
TGF 2013 - Jülich Supercomputing Centre - Sept. 25-27, 2013
2. Results – Presence of Groups The video was sampled considering one minute every five: a subset of 15 minutes
was extracted (1645 people - 21.16% of the total); the presence of groups (84.19%), their size and proxemics arrangements were detected
Patterns Couples Triples ≥ 4 Members
Lane 94.43% 31.91% 29.61%
River 5.57% 9.57% 3.12%
V-like - 58.51% 10.39%
Rhombus - - 7.79%
Two Dyads - - 32.47%
Triad + Single - - 10.39%
Single + Triad - - 6.23%
TGF 2013 - Jülich Supercomputing Centre - Sept. 25-27, 2013
TGF 2013 - Jülich Supercomputing Centre - Sept. 25-27, 2013
3. Results – Walking Path and Speed A sample of 30 singles, 15 couples, 10 triples and 8 groups of four member was
considered related to the B level of service. Data analysis was aimed at checking any differences in walking path and speed between singles and group members
Singles Couples Triples 4 Members
Walking Speed 1.22 m/sec (sd 0.16)
0.92 m/sec(sd 0.18)
0.73 m/sec(sd 0.10)
0.65 m/sec(sd 0.04)
Path 13.96 m(sd 1.11)
13.39 m(sd 0.38)
13.45 m(sd 0.27)
13.16 m(sd 0.43)
TGF 2013 - Jülich Supercomputing Centre - Sept. 25-27, 2013
TGF 2013 - Jülich Supercomputing Centre - Sept. 25-27, 2013
Identification of groups (and elderlies)
TGF 2013 - Jülich Supercomputing Centre - Sept. 25-27, 2013
3.1 Results – Trajectories and Path Within B level of service the differences in walking path between singles-couples and
singles-triples were significant (t-test, p < 0.05); no significant difference in path
between couple-triples (p > 0.05). The path of singles is 4.48% longer the the average
path of group members
Singles Couples Triples 4 Members
TGF 2013 - Jülich Supercomputing Centre - Sept. 25-27, 2013
TGF 2013 - Jülich Supercomputing Centre - Sept. 25-27, 2013
3.2 Results – Walking Speed Within B level of service the differences in walking speed between singles-couples,
singles-triples, singles-4members, couples-triples and triples-4 members were
significant (t-test, p < 0.01). The average walking speed of group members is 37.21%
lower than the one of singles
TGF 2013 - Jülich Supercomputing Centre - Sept. 25-27, 2013
TGF 2013 - Jülich Supercomputing Centre - Sept. 25-27, 2013
4.1 Results – Proxemics Dispersion Within B level of service situations the differences in proxemics group dispersion
(distance from the centroid of the group) between couples-triples and couples-4
members was significant (t-test, p < 0.01). The average spatial dispersion of triples and
groups of 4 members is 40.97% higher than the one of couples
Couples Triples 4 Members
DistanceCentroid
0.58 m(sd 0,22)
0.76 m(sd 0,11)
0.67(sd 0.12)
TGF 2013 - Jülich Supercomputing Centre - Sept. 25-27, 2013
TGF 2013 - Jülich Supercomputing Centre - Sept. 25-27, 2013
4.2 Results – Proxemics Dispersion The normalized position of each pedestrian with respect to the centroid and the
movement direction show that couples tend to walk with a line-abreast layout and a
distance of 0.4 m; triples tend to walk with a line-abreast pattern of 0.6 m (from left to
right members); no particular pattern regarding groups of 4 members can be detected
TGF 2013 - Jülich Supercomputing Centre - Sept. 25-27, 2013
Couples Triples
Groups of 4
TGF 2013 - Jülich Supercomputing Centre - Sept. 25-27, 2013
Towards integrated analysis and synthesis?
TGF 2013 - Jülich Supercomputing Centre - Sept. 25-27, 2013
Model and simulator
Target system
Simulation results
Empirical data
Simulation campaignexecution
Modellingand design of
a simulator
Analysis ofresults and
interpretation
Analysis of thedynamics of target
system
Synthesis
Analysis
(i) Formalisation ofphenomenologies
(ii) Metrics, indicators,techniques
(i) Motivations and goalsfor model innovation(ii) Data for calibration
and validation
TGF 2013 - Jülich Supercomputing Centre - Sept. 25-27, 2013
Towards integrated analysis and synthesis?
TGF 2013 - Jülich Supercomputing Centre - Sept. 25-27, 2013
TGF 2013 - Jülich Supercomputing Centre - Sept. 25-27, 2013
Conclusions and discussion
• Groups are relevant and significant
• Models for the simulation of crowds of pedestrians focused on collective events should consider them
• More observations, experiments and simulations are necessary to improve our understanding of the phenomenon
• More tight collaboration between researchers working on synthesis and analysis of crowds is promising and possibly beneficial for both
Grazie dell’attenzione! Thanks for your attention! Giuseppe Vizzari