The September 22nd RESCUE drill: What have we learned?

1
The September 22nd RESCUE drill What have we
learned?

• Miruna Petrescu-Prahova
• Vidhya Balasubramanian
• Chris Davison

2
Background

• A safety drill was organized on September 22nd at
  2pm as part of Responsphere
• Complete evacuation of the CalIT2 Building
• Real fire alarm
• Only the Building Manager knew, apart from the
  RESCUE drill team
• Aim to collect data about evacuation behavior

3
Data collection

• Questionnaire
• The questionnaire was administered to evacuees
  upon their return to the building by a team of
  sociologists located the main entrances of the
  building
• Cookies and punch were offered as incentives for
  participation
• N 42 completed questionnaires
• Indoor cameras
• Outdoor cameras

4
Results questions 1 2

• 1. When you first became aware of the alarm, what
  floor were you on?
• First floor 4 persons, 9.5
• Third floor 15 persons, 35.7
• Fourth floor 23 persons, 54.8
• 2. Which of the following best describes what you
  were doing when you first became aware of the
  alarm?
• Working at desk 33 persons, 78.5
• Engaged in small group meeting (4 or fewer
  persons) 6 persons, 14.3
• Engaging in non-work activities (e.g., coffee
  break) 1 persons, 2.4
• Changing locations within building 1 persons,
  2.4
• Entering/exiting building 1 persons, 2.4

5
Results question 3
Question - Between the time you first became aware of the alarm and the time you began to exit the building.... Percent Yes
Did you notice others leaving the building? 76
Did you retrieve personal belongings? 69
Did you save files or shutdown equipment? 40
Did you tell others to leave the building? 37.5
Were you told by other evacuees to leave the building? 34
6
Results question 3
Question - Between the time you first became aware of the alarm and the time you began to exit the building.... Percent Yes
Did you finish work or other tasks in progress? 25
Did you provide exit route information to others? 22.5
c. Were you told by emergency personnel to leave the building? 17
d. Did you ask other evacuees about exit route? 12
e. Did you ask emergency personnel about exit route? 7
7
Results questions 4 5

• In selecting your exit route, which of the
  following affected your decision?
• A. Knowledge of exit locations 36 persons, 86
• B. Observation of other evacuees 11 persons,
  26
• C. Advice from other evacuees 2 persons, 5
• D. Directions from emergency personnel 2
  persons, 5
• When exiting the building, which of the following
  did you use?
• Stairwell 1 10 persons, 24
• Stairwell 2 21 persons, 50
• Stairwell 3 11 persons, 26

8
Results - question 6

• Didnt work so well, people were confused

1st Location 2nd Location (if applicable) 3rd Location (if applicable)
Location name/description _____________ _______________ _______________
Reason(s) for selecting (circle)
A. Safe distance from building A A A
B. Followed friend/coworker B B B
C. Followed crowd C C C
D. Directed by emergency personnel D D D
9
Results - question 7

• When you first became aware of the alarm, what
  did you believe was the reason for the building
  evacuation?
• Fire 5 persons, 12
• False alarm 17 persons, 40
• Safety drill 19 persons, 45 (doesnt add to
  100 because of missing data)
• Note these results may be biased as a result of
  the retrospective nature of the data
• Good news people on the 4th floor were not more
  likely to believe it was a drill!
• (Chi-square 3.2, df 4, p-value0.52)

10
Results - question 8

• Using the map below, please trace the path you
  took when evacuating the building. Begin with
  your initial location (when you first became
  aware of the alarm), and end with the exit taken
  to leave the building. If you are unsure of your
  exact path, please draw a path which reflects
  your best estimate.
• This is where we need help!

11
(No Transcript)
12
(No Transcript)
13
(No Transcript)
14
(No Transcript)
15
Issues

• Synchronization between camera feeds
• Mismatch between different frames in different
  cameras
• Software to start cameras remotely and start
  saving instantaneously
• Algorithms for video based tracking should
  consider
• Shadows
• Synchronization issues

16
Conclusions

• We obtained valuable information about what
  people do between hearing an alarm and evacuating
• Behavior and belief heterogeneity among actors
  was observed, should be reflected in the model
• There is some inaccuracy regarding peoples
  reports of what they did