a passive influence model for adapting environments based on semantic preferences morelab –...

A Passive Influence Model for Adapting A Passive Influence Model for Adapting Environments based on Semantic Environments based on Semantic PreferencesPreferencesMoreLab – Mobility Research LabMoreLab – Mobility Research Labhttp://www.morelab.deusto.eshttp://www.morelab.deusto.es

Juan Ignacio Vazquez, Diego López de Ipiña and IñigoSedanoJuan Ignacio Vazquez, Diego López de Ipiña and IñigoSedano

CTSB 2006 - International Workshop on Combining Theory and Systems Building in

Pervasive Computing(at Pervasive 2006)

Dublin, Ireland, 7 May 2006

2University of Deusto – MoreLab

A Passive Influence Model for Adapting Environments

based on Semantic Preferences

MotivationMotivation

There is an increasing concern about empowering ubiquitous computing environments with more intelligent capabilities: To reduce user interaction and annoyance about operating the

environment Reactive spaces: the Ambient Intelligence vision

Hypothesis: Context-awareness seems to be at the heart of this problem. Automatic environment adaptation to user’s needs would reduce

user interaction. Some form of reasoning would boost up intelligence.

SOAM: Smart Object Awareness and Adaptation Model




Context informationContext information

Any environment exposes several kinds of information that can be measured through existing devices: temperature, TV channel, ambient music style and volume, …

E = {i1, i2, i3, …, in} …adopting concrete values at the moment of time t:

i1(t) = v1; i2(t) = v2 ; i3(t) = v3 ;…; in(t) = vn

E(t) = {i1(t) , i2(t) , i3(t) , …, in(t) } = {v1, v2, v3, …, vn} In SOAM, the evolution of one concrete kind of

environment information depends on its current value, and the influences (constraints) exerted on it:

i (t+1) = f ( i (t) , C (i , t) ) = f ( v , C (i, t) ) For example:

C (temperature, t) = {>23ºC, <27ºC, =25ºC}




Context awareness & adaptationContext awareness & adaptation

So, the state of the environment at time t+1…

E(t+1) = { i1(t+1) , i2(t+1) , i3(t+1) , …, in(t+1) }= {f ( v1 , C(i1, t) ) , f ( v2 , C(i2, t) ) ,…, f ( vn , C(in, t) ) }= F ( E(t), C(E, t) )

…depends on the current state of the environment and all the influences exerted on it at that moment of time.

Thus, constraints drive environmental adaptation. In SOAM:

Environment information (context information) is provided by devices Constraints are processed by devices Constraints can be expressed at a higher abstraction level in terms of

conditional rules: Adaptation Profiles An entity is context-aware when its behaviour is driven by Adaptation

Profiles activated under concrete conditions bounded to context information.




Passive InfluencePassive Influence

Within this model, there are no explicit commands or invocations to perform an action on a device like traditional mechanisms (RPC, RMI, SOAP, IIOP, …) Active Influence.

One entity can influence others in the environment by: Disseminating context information Disseminating adaptation profiles (desires, suggestions, not orders)

It is what we call Passive Influence (Vazquez EUSAI 2004)

Example: As the user enters a hotel room, his PDA disseminates its profile, and the

room automatically adapts without explicit intervention (temperature, TV channel configuration).

As the users gets into a rented car, his PDA disseminates its body proportions and suitable adjustments are carried out automatically.




Coping with conflictCoping with conflict

Passive mechanisms perform quite well in conflictive scenarios since there are not conflicting orders, but available information: Two people with disjoint temperature preferences the

temperature control system decides to set a mid-point between preferred ranges

The environment decides how to adapt and resolve the conflict, since it holds the information to take a proper decision.

Strategies: An intermediate value The first wins The last wins Depending on privileges or other attributes …




Issues raisedIssues raised

What information to spread and to whom? In which way can elements be influenced? Can reasoning be applied to generate new information? How long does the influence persists? What about data representation, protocols, …?




Background researchBackground research

Some Ubicomp architectures apply web technologies for communication & information exchange: UPnP – Universal Plug and Play WSAmI: Web Services for Ambient Inteligence

Semantic Web technologies (RDF, OWL) are used to represent information concepts and relationships, and generating new knowledge from existing one (reasoning).

SW is completely integrated into the web model. The Web of Knowledge is the next generation web. The application of SW to UC scenarios has been pioneered in

initiatives as Task Computing and CoBrA/SOUPA.




The Pervasive Semantic WebThe Pervasive Semantic Web

Our goal is to create Pervasive Semantic Web environments with devices forming a location-constrained web of knowledge, where they exchange semantic information and perform reasoning, in order to react accordingly to every situation.

Context Information is provided by existing devices in the form of RDF using appropriate domain ontologies.

Context-Awareness & Adaptation is performed by devices periodically retrieving every others’ provided context information, reasoning upon it and trying to honour active constraints.

Sensors & Effectors

Wireless Sensor Networks

Appliances

User devices

Location-constrainedWeb of Knowledge

Pervasive Semantic Web

10

University of Deusto – MoreLab



SOAM BasicsSOAM Basics

SOAM: Smart Objects Awareness and Adaptation Model

A Pervasive Semantic Web model for automatic environment adaptation based on user preferences. Further or manual control is always permitted

UPnP devices augmented to exchange semantic information.

Basically: We represent the state of the environment as an RDF Graph

(triples) We represent user preferences as a desired state of the

environment (equivalent to RDF triples) under certain conditions (conditional rules)

A special entity, called Orchestrator, tries to change the environment state based on devices’ declared capabilities.

11




SOAM ArchitectureSOAM Architecture

Messages: Adaptation Profiles Capabilities Context Information Constraints

Entities Adaptation User-Agent Smobjects Orchestrator

Phases:1. Discovery2. Adaptation-Profiles Injection3. Capabilities retrieval4. Context Information Retrieval5. Reasoning (DL, domain rules)6. Constraints injection

Environment

SmobjectAdaptation User-Agent

Orchestrator

Smobject

Smobject

Smobject

APCa

CoI

Smobject

Messages

AP: Adaptation ProfileCa: CapabilitiesI: Context InformationCo: Constraints

12




ExamplesExamples

Example 1: AP: “I would like to listen classical music when I am working with my laptop” HiFi system’s capabilities: “I can operate on the music domain” Laptop’s capabilities: “I can perceive who is working with me” Orchestrator:

Retrieves Context Information (from Laptop) If preconditions are true

Send to the HiFi system: “play classical music”

Example 2: AP: “If I am at location x of type HotelRoom, I would like x’s temperature to be

24ºC” Room: I can perceive my type Location Control System: I can perceive whether the guest is in. Temperature control system’s capabilities: I can perceive the temperature of the

present location, I can operate the temperature of the present location Orchestrator:

Retrieves Context Information (from the Room and the Location control system) If preconditions are true

Send to the TempControlSystem: “set temperature to 24ºC”

13




ResultsResults

Smobjects HW: ARM7-based 55 MHz with μClinux and a Java VM platform, simulating devices (Hi-Fi, TV, lights, temperature control, microphone, …). Small form factor: 7cm x 2cm.

Orchestrator: Pentium-M 1.86 GHz, Jena2 API Recently migrated to a distributed choreograhic

model without central orchestration

14




Conclusions and ContactConclusions and Contact

Theory vs Implementation: Semantic Web was a fine selection Platform limitations: no orchestrator at every device, power

consumption

MoreLab – Mobility Research Lab at the University of Deusto: http://www.morelab.deusto.es

My email: [email protected]

15




Examples – Adaptation ProfileExamples – Adaptation Profile

<adaptationProfile id="urn:uuid:prof2" expires="PT2M"><variable id="x"/><variable id="y"/><precondition id="" predicate="http://www.awareit.com/onto/task#isDoing" subject="urn:uuid:Bob">

<objectResource ref="http://www.awareit.com/onto/task#WorkingWithLaptop"/></precondition><precondition id="" predicate="http://www.awareit.com/onto/location#isLocatedIn" subject="urn:uuid:Bob">

<objectVariable ref="x"/></precondition><precondition id="" predicate="http://www.awareit.com/onto/sound#hasSound" subject="x">

<objectVariable ref="y"/></precondition><postcondition id="" predicate="http://www.awareit.com/onto/sound#volume" subject="y">

<objectLiteral datatype="http://www.w3.org/2001/XMLSchema#int">3</objectLiteral></postcondition><postcondition id="" predicate="http://www.awareit.com/onto/sound#style" subject="y">

<objectResource ref="http://www.awareit.com/onto/sound#ClassicalMusic"/></postcondition>

</adaptationProfile>

16




Examples – CapabilitiesExamples – Capabilities

<capabilitiesCollection owner="urn:uuid:hifi1" xmlns="http://www.awareit.com/soam" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://www.awareit.com/soam soamdt.xsd"><perceptionCapability id="urn:uuid:hifi1_pcap1">

<subject resource="urn:uuid:hifi1"/><predicate resource="http://www.awareit.com/onto/sound#hasSound"/>

</perceptionCapability><perceptionCapability id="urn:uuid:hifi1_pcap2">

<subject resource="urn:uuid:hifi1"/><predicate resource="http://www.w3.org/1999/02/22-rdf-syntax-ns#type"/>

</perceptionCapability><perceptionCapability id="urn:uuid:hifi1_pcap3">

<subject resource="urn:uuid:hifi1_sound"/><predicate resource="http://www.awareit.com/onto/sound#volume"/><predicate resource="http://www.awareit.com/onto/sound#style"/>

</perceptionCapability><operationCapability id="urn:uuid:hifi1_ocap1">

<subject resource="urn:uuid:hifi1"/><predicate resource="http://www.awareit.com/onto/sound#hasSound"/>

</operationCapability><operationCapability id="urn:uuid:hifi1_ocap2">

<subject resource="urn:uuid:hifi1_sound"/><predicate resource="http://www.awareit.com/onto/sound#volume"/><predicate resource="http://www.awareit.com/onto/sound#style"/>

</operationCapability></capabilitiesCollection>

17




Examples – Context InformationExamples – Context Information

<rdf:RDF xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#" xmlns="http://www.awareit.com/onto/examples/example1#"> <rdf:Description rdf:about="urn:uuid:hifi1"

xmlns:sound="http://www.awareit.com/onto/sound#"> <rdf:type rdf:resource="http://www.awareit.com/onto/sound#SoundSystem"/> <sound:hasSound> <sound:Sound rdf:about="urn:uuid:hifi1_sound"> <sound:volume

rdf:datatype="http://www.w3.org/2001/XMLSchema#int">8</sound:volume> <sound:style

rdf:resource="http://www.awareit.com/onto/sound#RockMusic" /> </sound:Sound> </sound:hasSound> </rdf:Description></rdf:RDF>

18




Examples - ConstraintsExamples - Constraints

<constraint id="urn:uuid:xxx" requester="urn:uuid:Bob" subject="urn:uuid:hifi1_sound" predicate="http://www.awareit.com/onto/sound#volume">

<objectLiteral datatype="http://www.w3.org/2001/XMLSchema#int">3</objectLiteral>

</constraint>

<constraint id="urn:uuid:xxx" requester="urn:uuid:Bob" subject="urn:uuid:hifi1_sound" predicate="http://www.awareit.com/onto/sound#style">

<objectResource ref="http://www.awareit.com/onto/sound#ClassicalMusic"/>

</constraint>

a passive influence model for adapting environments based on semantic preferences morelab –...

Documents