Research on Human Behaviour Simulation in the Built Environment

1
Research onHuman Behaviour Simulation inthe
Built Environment

• www.ddss.arch.tue.nl
• Bauke de Vries

2
Programme

• Who am I / Where do I come from
• Research programma Design and Decision Support
  System
• PhD and Graduation projects on Behaviour
  Simulation in the Built Environment

3
Who am I

• Name Bauke de Vries
• Age 49
• Occupation Married, 2 children
• Education Msc in Architecture, Building and
  Planning, PhD at Eindhoven University of
  Technology (TU/e)
• Profession Professor at TU/e in Design Systems

4
Where am I from

• Eindhoven University of Technology, founded by
  Philips 50 year ago
• Faculty Architecture, Building and Planning
• -1000 Bachelor students
• - 500 Master students
• - 50 PhD students

5
DDSS
Research ProgrammeDesign and Decison Support
Systems
Planning
Design
ICT
Artificial Intelligence
6
EU and PhD projects

• Building Management Simulation Centre
• Decision Support System for Building
  Refurbishment
• Measuring User Satisfaction through Virtual
  Environments
• Using a Virtual Environment for Understanding
  Real-World Travel Behavior
• A Learning Based Transportation Oriented
  Simulation Systems
• Human Behavior Simulation in the Built Environment

7
Projects on Behaviour Simulation

• VR experiments (Amy Tan)
• Space utilisation (Vincent Tabak)
• Pedestrian behavior (Jan Dijkstra)
• AMANDA framework (Joran Jessurun)
• Evacuation and smoke simulation (Martin Klein)
• Safety assesment (Ruben Steins)

8
The Reliability and Validity of Interactive
Virtual Reality Computer Experiments SPIN System
9
SPIN Demo

• (start CD-ROM)

10
The Reliability and Validity of Interactive
Virtual Reality Computer Experiments Conclusions

• The structural dimensions (number of stops,
  number of activities) were better measured by
  SPIN.
• The PAPI questionnaire yielded better responses
  for durations (of shopping activity, services
  activity, out-of-home leisure activity, travel
  between activities, whole schedule).
• Route choice data indicated that SPIN was not
  able to measure this dimension better than PAPI.

11
User Simulation of Space Utilisation

• Existing models focus on evacuation behaviour
• Aim Analyze the performance of a design through
  user behaviour simulation

12
System overview

• The User Simulation of Space Utilisation (USSU)
  system.
• Important aspect interaction between persons.

13
System overview

• Input
• The organisation Roles, activities, persons
  (FTE)
• The design of the building in which the
  organisation is (or will be) housed the spatial
  conditions.

14
System overview

• Output
• Movement pattern for each member of the
  organisation. From this performance indicators
  can be deduced, like
• Average/maximum walking distance/time per
  individual.
• Number of persons per space in time.
• Usage of facilities.

15
Skeleton activities
The core activities for a certain period (a
workday). Activities depend on the
organisational workflow.Some activities require
interaction between employees.
Time Activity Resources
0900-1000 Research Office space X
1000-1100 Get coaching Office space Y
1100-1600 Research Office space X
1600-1700 Attend presentation Meeting room Z
1700-1800 Research Office space X
16
Intermediate activities

• Activities adjust/complement the skeleton
  activities.Categories of activities
• Physiologic getting a drink, having lunch,
  going to toilet.
• Social having a chat with colleague.

Time Activity Resources
0900-0915 Research Office space X
0915-0920 Get a drink Coffee corner
0920-1000 Research Office space X
1000-1100 Get coaching Meeting room Z
1100-1215 Research Office space X
1215-1315 Lunch Canteen
1315-1400 Research Office space X
--- --- --
Time Activity Resources
0900-1000 Research Office space X
1000-1100 Get coaching Office space Y
1100-1600 Research Office space X
1600-1700 Attend presentation Meeting room Z
1700-1800 Research Office space X
17
Intermediate activities

• S-curve method to predict the intermediate
  activities.
• Shape of curve influenced by
• Time pressure.
• History of executed activities.
• Skeleton activity (task).

18
System design
19
Scheduler

• After drawing the skeleton activities scheduler
  is activated.Consists of 9 AI (Artificial
  Intelligence) modules.Responsible for (among
  others)
• Scheduling skeleton activities
  (SkeletonScheduler)
• Scheduling intermediate activities
  (IntermediateScheduler)
• Repairing schedules (OverlapRemover
  GapRemover)
• Determining interaction between activities
  (InteractionScheduler)
• Finding combinations of activities
  (CombinationFinder)
• Finding an appropriate location (ResourceFinder)

20
Experiment
Capture the real space utilisation Using RFID
to capture the real space utilisation. Merge
spaces into zones.
21
Pedestrian Behaviour

• Shopping environment populated with agents
  representing pedestrians
• Agents
• are supposed to carry out a set of activities Ai
• have different motivational states
• move across the network
• have perceptual fields that may vary according
  agents awareness threshold and the signalling
  intensity of a store

Context
22
Basic Equation
Behavioural Aspects
23
Data Collection
24
Data Collection
25
Estimation Results

• Basic equation is estimated for fixed distances
• The dichotomous response variable
• is the awareness of a store category within the
  perceptual field
• Explanatory variables are
• store category
• motivation for visiting the city centre

26
AMANDA framework

• Extension of pedestrian/user behaviour models
  with destination and route choice, and activity
  scheduling
• Domain pedestrian behaviour in a public space
  (e.g. shopping environment), user movement in a
  building (e.g. office building)

27
Agent Architecture
28
Environment

• Pedestrians move in a built and/or urban
  environment
• Pedestrians are represented by agents
• A hybrid (grid and polygon) based model is used
  to simulate their behaviour across the network
• Each cell in the grid can be considered as an
  information container object it has information
  about which agents and polygons occupy it.

Context
29
Simulation of Individual Behaviour
Action Selection strategy, goals, planning
Steering path determination
Pedestrian Movement
Context
30
AMANDA demo

• (start AMANDA test application)

31
Evacuation and smoke simulation

• Simple evacuation behaviour shortest route to
  exit
• CAD vendor independent IFC based
• Using existing smoke simulation CFAST
• No interaction between evacuation and smoke
  simulation

32
Evacuation simulation (AMANDA)
Occupants data
Building model (IFC)
Source file (XML)
Results
Designer
User Interface
Fire data
Smoke simulation (CFAST)
33
Testcase Vertigo building

• Model created with Autodeks/Revit and exported to
  IFC
• 9-th floor
• 26 rooms
• 2 exits

34
IFC input
Evacuation simulation
IFC
Smoke simulation
35
Evacution Simulation AMANDA
36
Smoke simulation CFAST

• Consolidated Model of Fire Growth and Smoke
• National Institute of Standards and Technology
  (NIST)
• Import/export facilities
• Max 30 spaces, 50 openings

37

• File Input
• 3D geometry
• Openings
• Simulation time, output interval
• User interface input
• Fire specification

38
Linking results Evacuation and Smoke simulation

• Required egress time lt available egress time
• Simulation results
• Evacuation ? Space location for each person at
  any time
• Smoke ? (harmful) conditions in each space at any
  time.

39
Test results

• Total and everage evacuation time
• Numbers per exit
• Per agent
• Distance covered
• Spaces crossed (!)
• Walking speed
• Per space
• Space utilisation

40
Safety assesment

• The main purpose of the Dutch Working Conditions
  Act (WCA) is to ensure three things
• Safety no acute dangers for people at work
• Health no long term or chronic physical health
  risks
• Wellbeing no psychological problem caused by
  working conditions

41
Compliance checking

• Soft coded regulations
• Each firm must have a policy stating in what way
  the personal privacy of individuals is
  guaranteed.
• Hard coded regulations
• For seated work a free space is present beneath
  the working surface of at least 70 cm in height
  and 60 centimeters in depth and width. For
  office-work the minimal depth for legs and feet
  is 65 and 80 centimeters respectively.

42
Example Soft coded Privacy factors (self defined)

• In offices that are shared by many people, the
  chance of privacy problems is higher.
• Rooms with high ceilings have more sound
  resonance, which means more inconvenience, which
  results in less privacy
• Rooms adjacent to busy corridors suffer from
  higher sound levels, resulting in more
  inconvenience
• Rooms next to windows give a higher feeling of
  privacy, since people can lose themselves in
  the view

43
Method Fuzzy logic (1)

• Input Membership function amountOfPeople
• Input Membership function
• officeHeight

44
Method Fuzzy logic (2)

• Output membership function
• privacyProblem

45
WCA system
46
Input data

• IFC file created with Autodesk/Revit Building
  geometry
• Organisational data generated with USSU
  Acitivity and location for each person at any
  time
• Building physics data generated with ecoTect

47
Output data
48
Thanks !