Emergency Fire Incident Training in Virtual Worlds

1
Emergency Fire Incident Training in Virtual Worlds

• Dugdale, J. Pavard, B., Palamin N., El Jed M.
• Cognitive Engineering Research Group
• Computer Science Institute (IRIT)
• Toulouse, France.

2
The South of Paris Ambulance Service
3
NetCrise
4
SMUR-SCRIPT Mobile technology for ground zero
medics
5
Project 15-18


Palaiseau
VAL d YERRES
Viry Chatillon
EVRY
ARPAJON
CENTRE 15/18
ETAMPES
6
SWARM based Model
7
(No Transcript)
8
Project SoCute (distributed VR)

• Aim of the project
• Use distributed VR to overcome the social and
  cultural limitations of distributed AI approach
• Design a training tools for emergency management
• Use this tool to design new procedures or test
  new communication technologies (www.irit.fr/GRIC/3
  D)
• Recurrent problems
• Coordination between teams
• Style of management
• Perturbations in information
• Flow
• Interoperability

9
Plan

• Situated vs AI simulations (towards cultural
  technologies)
• Methodology
• Challenges for the design of situated
  collaborative spaces
• Cognitive emotional architecture
• Running emergency scenarios
• Future work

10
1. Situated vs AI simulations (towards cultural
technologies)
The AI approach (Gratch Marsella,
2002) (Univ. Southern Calif.)
The SoCute approach A Social
Culturally Situated approach (GRIC)
11
2. Methodology

• Methodology based on extensive FIELD ANALYSIS
• From video analysis, we list the mains MANAGEMENT
  COMMUNICATION ERRORS in order to assess their
  consequences as well as their nature.
• From this field analysis, we design a generic VR
  environment architecture that will allow trainees
  to face the same kinds of difficulties.
• We emphasize verbal and non verbal communications
  as well as interactions through artifacts in
  order to reproduce the same situation in virtual
  world
• With professionals, we design training scenarios
• We will then compare collaboration strategies in
  real and virtual world (normal and degraded
  situations)
• The scenarios focuses on the training for
  management (strategic tactic)

12
Example 1 Air traffic crash (40 people
injured, 5 casualties, 200 people involved)

• Coordination and cooperation problems between
  local medical teams, firemen and medics
• Agent roles change over time
• Bad anticipation of the victim flow due to
  information management problems

13
Example 2 Hotel fire (10 people caught in a
fire in the upper hotel levels)

• Firemen may be strongly disturbed by the amount
  of external requests
• Bad style of management
• Difficulties to keep a bird view of the
  situation

14
3. Challenges for the design of professional
virtual collaborative spaces

• Management of verbal non verbal communication
  between characters
• Characters with emotional behaviour
• Mixing intentional and automatic behaviour
• Interfaces that supports non verbal
  communications like deictics

15
4. Cognitive Emotional architecture

• situation (importance, desirability)
• conversation (colleagues, civilians, officials)
• solutions

Context (cognitive, social, etc.)
Environmental parameters
EMOTIONAL MODEL
Emotions (4)
(Noise, explosion, etc.)

• Emotion
• Personality
• Mood

Avatar internal state (cognitive emotional)
16

• Emotional model for the production of gestures
  and facial expressions

Users speech
Context
Context
FACIAL EXPRESSIONS
EMOTIONAL MODEL
Emotions (4)
GLOBAL
BEHAVIOUR
BODY EXPRESSIONS
Avatar internal state (cognitive emotional)
Context
Users actions
17

• General architecture for the emotional model

18
Role of gestures in communications

Full gestual expression
Without deictics
Uncorrelated gestures
19
How the context can modify automatic behavior
Ex. FIREMEN CONTEXT
Each training scenario is associated to several
contexts which are user dependant (depending of
the agent role)
20
Context and character animation

• Example 2 Context and communication.

21
5. Running emergency scenarios
22
6. Future work

• Adopting an ethnomethodological approach to
  validate our virtual environment
• Introduce interfaces for the asesssment of new
  communication technologies (ANOTO pen LAN on
  the disaster area)
• Link with distributed agent modelling
• SoCute Social Cultural TEchnologies

23
Intégration modèle agents distribués et réalité
virtuelle
24

• www.irit.fr/GRIC/3D

25
(No Transcript)
26
5. Conclusion Demo

• Exemples
• expression faciale et gestuelle.
• Geste de déictique en conversation.
• Gestion du regard en conversation.
• Intervention Feu Hôtel.
• Modification du comportement intentionnel

27
Jade plateforme
28
1. Introduction

• Aim
• Developing an emotional model that can be
  integrated into characters in an interactive
  virtual reality environment.
• Approach
• Based on the mixed agent paradigm where agents
  who have an important role are animated by humans
  according to their own perceptions, objectives,
  knowledge, history and culture.

29
1. Introduction

• Advantages
• Avoiding to model a complete cognitive system,
• Simplicity of implementation,
• Speed of execution.
• Problems
• How to represent the emotional dimension on a
  virtual actor (emotions, personality, mood)?
• How to perform non intentional actions on a
  virtual actor (gestures, behaviour)?

30
2. Emotional model

• Emotion
• Emotion is a reaction to events deemed relevant
  to the needs, goals, or concerns of an
  individual. Kleinginna Kleinginna,1981
• The role of emotions in virtual reality training
  systems?

Increasing the credibility of the agent
Influence on decision-making
Engaging the user
Learning and training
31

• Our model of emotions

32

• Emotions affect the avatar state in three main
  ways

Affective Reaction
Behavioural Reaction
Cognitive Reaction
Modification of perception (e.g restriction of
the field of view)
Facial Expression

• Modification of intentional actions (e.g walk)
• Gestures
• Reflexes

33

• Personality

• Five-Factor Model (FFM) Costa McCrae, 1992
• One of the most widely accepted models of
  personality.
• The five dimensions providing a description of
  personality
• Openess  curious, broad interests, creative,
  original, imaginative.
• Consciencious  organised, reliable,
  hard-working, self-disciplined.
• Extravert  sociable, active, talkative,
  optimistic.
• Agreeable  good-natured, trusting, helpful,
  forgiving
• Neurotic  worries, nervous, emotional,
  insecure, inadequate.

34
3. Emotional model and behavioural engine
Context Task Behavioural Mode
Action Proximity None
Context Task Behavioural Mode Stand Proximity
None
35
3. Emotional model and behavioural engine
Context Communication Behavioural Mode
Listening Modality Verbal Proximity Mehdi
Context Communication Behavioural Mode
Talking Modality Verbal Proximity Nico
36

• Mood

• Acts as a filter for the emergence of emotions
  and influences the interpretation of a situation
  .
• Mood can be seen as a background affective
  filter through which both internal and external
  events are appraised.

1
-1
Good
Bad
37
The new emotional state is computed by
considering the potential emotional state and by
moderating it considering the mood.
According to the emotional impact of the event,
we determine the potential emotional state.

• Process of computing emotional
• state

Rules and prescribed procedures defined from the
scenario
Context Explosion in the hotel
Perception of events
According to the personality, we define the
sensitivity of the avatar to emotions
An emotion of FEAR will appear
38
The new emotional state is computed by
considering the potential emotional state and by
moderating it considering the mood.

• Process of computing emotional
• state

Moderation factor
Emotion will be accentuated
39

• Process of computing emotional
• state

Updating the mood considering the new emotional
state.
According to the emotional state and the MEH
matrix, we evaluate the impact of each emotion on
the mood
The emotion of Anger tends to set the mood to
hanger
The mood will be set to the mean value of hanger
and hfear
40

• Process of computing emotional
• state

Process of decay of emotions.
The emotional decay vector defines how long the
intensity of an emotion will be maintained before
decreasing.
0.2 µ1
0.8 µ2