A Semantic Model for Proactive Home Care Systems

Transcription

A Semantic Model for Proactive Home Care Systems
A Semantic Model for Proactive
Home Care Systems
Alencar Machado
Leandro Krug Wives
José Palazzo Moreira de Oliveira
Introduction
• Ambient Intelligence will enable environments to support
people
– Smart Home (CHAN et al., 2008)
• Related to an increase in Home Automation on user’s living environment
– Ambient Assisted Living (JARA; ZAMORA; SKARMETA, 2011)
• Related to an increase in the Range of services to support independent
living
• This research seeks to identify how:
– Systems in smart environment can manage and adapt the
environment to user’s needs
– Users needs are associated the situations that involve the user in their
environment of living
– Unwanted Situations
• System actions are necessary to mitigate or eliminate them
2
Complex Events Processing
3
Background
• Context-awareness
– Reasoning about context
• Detect facts of interest
Sensor
isTypeOf
Current_Situation
Patient
Sensor
isSituationOf
useDevice
4
Background
• Context-awareness
– Reasoning about context
• Detect CURRENT SITUATION
» Rules: Semantic Web Rule Language (SWRL)
• Antecedent=> Consequent
Patient(John) ∧ Sensor(Sensor_Heartbeat1) ∧
Current_Situation(Emergency) ∧ useDevice(John, Sensor_Heartbeat1) ∧
hasValue (Sensor_Heartbeat1, colectledValue) ∧ swrlb:greaterThanOrEqual(collectedValue,130)
→ isSituationOf (John, Emergency)
Sensor
isTypeOf
Current_Situation
Patient
Sensor
isSituationOf
useDevice
5
Background
• Context-awareness
– Reasoning about context
• Detect CURRENT SITUATION
» Rules: Semantic Web Rule Language (SWRL)
• Antecedent=> Consequent
Patient(John) ∧ Sensor(Sensor_Heartbeat1) ∧
Current_Situation(Emergency) ∧ useDevice(John, Sensor_Heartbeat1) ∧
hasValue (Sensor_Heartbeat1, colectledValue) ∧ swrlb:greaterThanOrEqual(collectedValue,130)
→ isSituationOf (John, Emergency)
The sentences are always true (100%) in all logical interpretations.
Sensor
isTypeOf
Current_Situation
Patient
isSituationOf
(100%)
Sensor
useDevice
(100%)
6
Background
• Context-awareness
– Reasoning about context (uncertainty)
• To detect Future Situation is necessary reasoning over uncertainty
– Prediction with probability
– Probabilistic Ontology
» Multi Entity Bayesain Network (MEBN)
Sensor
isTypeOf
Current_Situation
Sensor
Patient
isSituationOf
(90%)
useDevice
(70%)
7
Conceptual Model
Selected ontologies to generate the conceptual model for AAL
Feature
Basic
Capacity
Preference
Experiences
Activity
Role
Health
Location
Device
Service (OWL-S)
Structure
Event
Situation
Actions
Uncertainty
Pi
Go
*
**
**
**
Sk
Do
User Model
***
**
***
***
***
***
***
*
*
Ab
Re
mI
*
*
•
*
*
•
*
*
*
*
*
***
*
*
(Hervás et al., 2010)
(Bonino et al., 2008),
• Abdulrazak et al., (2010),
• DomoML
*
***
Proactive Model
• PiVOn
• Golemati et al.,(2007)
• Skillen et al.,(2012)
• Dogont
*
*
*
*
*
Be
•
***
*
Physical Environment Model
*
*
*
***
*
*
***
*
*
Do
(Sommaruga et al., 2011)
• Reinisch et al. (2011),
• mIO! Poveda et al., (2010)
• BeAware
•
(Baumgartner et al., 2010)
*
*
*
8
Conceptual Model
Selected ontologies to generate the conceptual model for AAL
Feature
Basic
Capacity
Preference
Experiences
Activity
Role
Health
Location
Device
Service (OWL-S)
Structure
Event
Situation
Actions
Uncertainty
Pi
Go
*
**
**
**
Sk
Do
User Model
***
**
***
***
***
***
***
*
*
Ab
Re
mI
*
*
•
*
*
•
*
*
*
*
*
***
*
*
(Hervás et al., 2010)
(Bonino et al., 2008),
• Abdulrazak et al., (2010),
• DomoML
*
***
Proactive Model
• PiVOn
• Golemati et al.,(2007)
• Skillen et al.,(2012)
• Dogont
*
*
*
*
*
Be
•
***
*
Physical Environment Model
*
*
*
***
*
*
***
*
*
Do
(Sommaruga et al., 2011)
• Reinisch et al. (2011),
• mIO! Poveda et al., (2010)
• BeAware
•
(Baumgartner et al., 2010)
*
*
*
9
Conceptual Model
• The conceptual model was developed using
the network ontology methodology
• Three Domain
– User domain
– Physical Environment Domain
– Proactive Domain (here only presentation it)
1 Introdução – 2 Fundamentação– 3 Abordagem – 4 Modelo de Contexto – 5 Arquitetura – 6 Estudo de Caso – 7 Conclusões
10
Proactive Domain
REGARDING AN
USER
ACTION
executedBy
SERVICE
REGARDING A
DEVICE
included
subConceptOf
Object Property
concept
Ontology of
another domain
hasPreconditionSituation
hasService
isComposedBy
AUTOMATED
ACTION
REGARDING AN
ORGANIZATION
hasReactiveAction
FUNCTIONALITY
hasProactiveAction
hasPreconditionSituation
HUMAN
ACTION
generate
generate
TASK
hasHumanAction
hasTask
INTERNAL EVENT
EXTERNAL EVENT
USER
EVENT
influence
start
end
PR-OWL
OWL-Time
CURRENT
SITUATION
PREDICTIVE
SITUATION
willBeSituationOf
HISTORY
SITUATION
SITUATION
hasSituation
isSituationOf
11
Proactive Domain
REGARDING AN
USER
ACTION
executedBy
SERVICE
REGARDING A
DEVICE
included
subConceptOf
Object Property
concept
Ontology of
another domain
hasPreconditionSituation
hasService
isComposedBy
AUTOMATED
ACTION
REGARDING AN
ORGANIZATION
hasReactiveAction
FUNCTIONALITY
hasProactiveAction
hasPreconditionSituation
HUMAN
ACTION
generate
generate
TASK
hasHumanAction
hasTask
INTERNAL EVENT
EXTERNAL EVENT
USER
EVENT
influence
start
end
PR-OWL
OWL-Time
CURRENT
SITUATION
PREDICTIVE
SITUATION
willBeSituationOf
HISTORY
SITUATION
SITUATION
hasSituation
isSituationOf
12
Proactive Domain
TASK
hasTask(us,tar)
USER
runningTask(tas)
AUTOMATED
ACTION
automatedAction
Perfomed(ps)
EVENT
influence(ps, t)
hasReactiveAction(ps,aa)
OWL-TIME
influence(ps, tPrev)
PREDICTIVE
SITUATION
willBeSituationOf(ps, t, us)
Probabilistic Ontology to the Proactive Domain
13
Proactive Domain
The question that the probabilistic ontology
must answer is:
• What is the probability of a Predictive
Situation at a particular time to involve the
User in his living environment?
14
Proactive Domain
What is the probability of a Predictive Situation at a particular time to
involve the User in his living environment?
Theory MEBN: Reference Model for Predictive Situation-awareness
15
Proactive Domain
What is the probability of a Predictive Situation at a particular time to
involve the User in his living environment?
runningTask(tar) - hasTask(us, tar)
State: True, False
Theory MEBN: Reference Model for Predictive Situation-awareness
16
Proactive Domain
What is the probability of a Predictive Situation at a particular time to
involve the User in his living environment?
AutomatedAction(sp)
State: Automated Actions
Theory MEBN: Reference Model for Predictive Situation-awareness
17
Proactive Domain
What is the probability of a Predictive Situation at a particular time to
involve the User in his living environment?
influence(sp, t)
State: Events
if any sp have ( automatedActionPerformed = action1 ) [
if any tar have ( runningTask = false )
[ event 1= 0.3, event 2 = 0.7] else
[event 1 = 0.6, event 2 = 0.4 ]
] else[ event 1 = 0.5, event 2 = 0.5 ]
Theory MEBN: Reference Model for Predictive Situation-awareness
18
Proactive Domain
What is the probability of a Predictive Situation at a particular time to
involve the User in his living environment?
willBeSituationOf(sp, t, us)
State: True, False
if any ps have ( influence = event1)
[true = 0.99, false = 0.01]
else [ true = 0.5, false = 0.5]
Theory MEBN: Reference Model for Predictive Situation-awareness
19
Case Study
• Applied in two scenarios
– Profile:
• 75 years
• Diabetes, hypertension and dementia light
– Scenario 1:
• Assistance in the activity taking medications
– Scenario 2:
• Assistance in cooking activity controlling a smart stove
• Profile:
– State of senile
20
Case Study
What's new? (The same Model)
• can generate differents bayesian networks
• The structure of the network is generated in time
execution and depends only on the
instances of ontology
• Support time steps (Dynamic Bayesian)
(a) activity taking medications
(b) cooking activity
21
Work in progress
• Approach to support the development of
systems to AAL with reactive, proactive and
extensible features.
– Inside it:
• Complex Events Processing
• Reasoning over uncertainty
– Detected and Prediction Activities and forgetfulness by the
user
• Tool to modeling applications (make decision) for AAL
– Preliminary studies with ADOxx plataform
22
Thanks for your attention!
Obrigado!
[email protected]
[email protected]
Extensibility
Development Pervasive
Applications
Making Decision in Real Time
Interest in
develop
application
a) Current
Situation
Model
(1)
b) Predictive
Situation
Model
Situation of Interest
c) Reactive
Action Plan
(2)
Methodologies for develop
Systems
Developed
Application
e) Context of
Interest
(3)
f) Implement
application
(4)
d) Proactive
Action Plan
Action Plan
ONTOLOGY
Conceptual Model
Future Work: Use a platform (eg ADOxx) to build a modeling tool to support the generation of
Pervasive applications in assisted living environments
24
Middleware
25