Thank you very much for your attention ! Imad Afyouni supervised - - PowerPoint PPT Presentation

RESEARCH CONTEXT METHODOLOGY & RESULTS DISCUSSION

CONTEXT-AWARE PROCESSING OF CONTINUOUS

LOCATION-DEPENDENT QUERIES IN INDOOR ENVIRONMENTS

PHC PAVLE SAVIC 2012 - UNIVERSITY OF NIS - SERBIA

Imad AFYOUNI

Naval Academy Research Institute Department of Computer Science

8 May 2012 IMAD AFYOUNI (IRENAV) RESEARCH SEMINAR - NIS - PHC 2012 8 MAY 2012 1 / 22

RESEARCH CONTEXT METHODOLOGY & RESULTS DISCUSSION

TALK OUTLINE

1 RESEARCH CONTEXT

Location- and context-aware services and queries in indoor environments Research challenges

2 METHODOLOGY & RESULTS

Preliminary recommendations A hierarchical and context-dependent indoor data model Continuous query processing architecture Query language for navigation-related queries in indoor environments Continuous processing of location-dependent queries

3 DISCUSSION

Implementation Experience Achievements and Status of Research Activities Research perspectives

IMAD AFYOUNI (IRENAV) RESEARCH SEMINAR - NIS - PHC 2012 8 MAY 2012 2 / 22

RESEARCH CONTEXT METHODOLOGY & RESULTS DISCUSSION

TALK OUTLINE

1 RESEARCH CONTEXT

Location- and context-aware services and queries in indoor environments Research challenges

2 METHODOLOGY & RESULTS

Preliminary recommendations A hierarchical and context-dependent indoor data model Continuous query processing architecture Query language for navigation-related queries in indoor environments Continuous processing of location-dependent queries

3 DISCUSSION

Implementation Experience Achievements and Status of Research Activities Research perspectives

IMAD AFYOUNI (IRENAV) RESEARCH SEMINAR - NIS - PHC 2012 8 MAY 2012 3 / 22

RESEARCH CONTEXT METHODOLOGY & RESULTS DISCUSSION LOCATION- AND CONTEXT-AWARE SERVICES AND QUERIES IN INDOOR ENVIRONMENTS

CONTEXT CLASSIFICATION

Modelling the notion of context returns to the following questions : Who is the user ? Where is he/she ? What resources are nearby ? What is his/her purpose ?

IMAD AFYOUNI (IRENAV) RESEARCH SEMINAR - NIS - PHC 2012 8 MAY 2012 4 / 22

RESEARCH CONTEXT METHODOLOGY & RESULTS DISCUSSION LOCATION- AND CONTEXT-AWARE SERVICES AND QUERIES IN INDOOR ENVIRONMENTS

CONTEXT CLASSIFICATION

Modelling the notion of context returns to the following questions : Who is the user ? Where is he/she ? What resources are nearby ? What is his/her purpose ?

User-centred context Context of execution Environmental context Temperature Preferences Time Location Infrastructure System Capabilities User interface Light User profile Context of use

IMAD AFYOUNI (IRENAV) RESEARCH SEMINAR - NIS - PHC 2012 8 MAY 2012 4 / 22

RESEARCH CONTEXT METHODOLOGY & RESULTS DISCUSSION LOCATION- AND CONTEXT-AWARE SERVICES AND QUERIES IN INDOOR ENVIRONMENTS

LOCATION- AND CONTEXT-AWARE SERVICES AND

QUERIES IN INDOOR ENVIRONMENTS

Context-aware indoor navigation services exploit context dimensions in

• rder to

⋆ anticipate user’s needs ⋆ customize the user’s experience

Many components that contribute to the design of a context-aware system should be reflected by the spatial model underneath

IMAD AFYOUNI (IRENAV) RESEARCH SEMINAR - NIS - PHC 2012 8 MAY 2012 5 / 22

RESEARCH CONTEXT METHODOLOGY & RESULTS DISCUSSION LOCATION- AND CONTEXT-AWARE SERVICES AND QUERIES IN INDOOR ENVIRONMENTS

LOCATION- AND CONTEXT-AWARE SERVICES AND

QUERIES IN INDOOR ENVIRONMENTS

Context-aware indoor navigation services exploit context dimensions in

• rder to

⋆ anticipate user’s needs ⋆ customize the user’s experience

Many components that contribute to the design of a context-aware system should be reflected by the spatial model underneath Location-dependent queries (LDQ) comprise requests triggered by the user to “pull” some location-dependent information about objects of interest

⋆ Position queries ⋆ Navigation queries ⋆ Range queries ⋆ K nearest neighbour (kNN) queries

IMAD AFYOUNI (IRENAV) RESEARCH SEMINAR - NIS - PHC 2012 8 MAY 2012 5 / 22

RESEARCH CONTEXT METHODOLOGY & RESULTS DISCUSSION RESEARCH CHALLENGES

RESEARCH CHALLENGES

Developing a context-dependent indoor data model that

⋆ represents the features that populate the environment along with their dynamic properties ⋆ supports a large spectrum of services and queries (at different levels of abstraction)

IMAD AFYOUNI (IRENAV) RESEARCH SEMINAR - NIS - PHC 2012 8 MAY 2012 6 / 22

RESEARCH CONTEXT METHODOLOGY & RESULTS DISCUSSION RESEARCH CHALLENGES

RESEARCH CHALLENGES

Developing a context-dependent indoor data model that

⋆ represents the features that populate the environment along with their dynamic properties ⋆ supports a large spectrum of services and queries (at different levels of abstraction)

Designing a continuous query processing architecture for LDQs in indoor environments Introducing a query language to improve expressiveness of navigation-related queries Developing algorithms to process continuous navigation, range, and nearest neighbour queries in indoor environments

IMAD AFYOUNI (IRENAV) RESEARCH SEMINAR - NIS - PHC 2012 8 MAY 2012 6 / 22

RESEARCH CONTEXT METHODOLOGY & RESULTS DISCUSSION

TALK OUTLINE

1 RESEARCH CONTEXT

Location- and context-aware services and queries in indoor environments Research challenges

2 METHODOLOGY & RESULTS

Preliminary recommendations A hierarchical and context-dependent indoor data model Continuous query processing architecture Query language for navigation-related queries in indoor environments Continuous processing of location-dependent queries

3 DISCUSSION

Implementation Experience Achievements and Status of Research Activities Research perspectives

IMAD AFYOUNI (IRENAV) RESEARCH SEMINAR - NIS - PHC 2012 8 MAY 2012 7 / 22

RESEARCH CONTEXT METHODOLOGY & RESULTS DISCUSSION PRELIMINARY RECOMMENDATIONS

PRELIMINARY RECOMMENDATIONS

SERVICE-ORIENTED REQUIREMENTS

Localisation : either geometric or symbolic format, should support hybrid positioning techniques Context-aware Adaptive navigation : multi-criteria path selection Location-aware communication : quality and format of exchanged data Activity-oriented interaction

IMAD AFYOUNI (IRENAV) RESEARCH SEMINAR - NIS - PHC 2012 8 MAY 2012 8 / 22

RESEARCH CONTEXT METHODOLOGY & RESULTS DISCUSSION PRELIMINARY RECOMMENDATIONS

PRELIMINARY RECOMMENDATIONS

SERVICE-ORIENTED REQUIREMENTS

Localisation : either geometric or symbolic format, should support hybrid positioning techniques Context-aware Adaptive navigation : multi-criteria path selection Location-aware communication : quality and format of exchanged data Activity-oriented interaction

EFFICIENCY-RELATED REQUIREMENTS

Modelling effort : the cost and complexity of building the model Flexibility : may a simple unmixed spatial model deal with a very diverse intelligent navigation tasks efficiently ? Performance and scalability : real time constraints

IMAD AFYOUNI (IRENAV) RESEARCH SEMINAR - NIS - PHC 2012 8 MAY 2012 8 / 22

RESEARCH CONTEXT METHODOLOGY & RESULTS DISCUSSION A HIERARCHICAL AND CONTEXT-DEPENDENT INDOOR DATA MODEL

A HIERARCHICAL AND CONTEXT-DEPENDENT INDOOR

DATA MODEL

Room06 Balcony MeetingRoom02 MeetingRoom03 Room07 Room05 Room04 Room10 Breakroom Room11 Room09 Room08 Room12 HW05 Bathroom Bathroom Bathroom Bathroom Bathroom Bathroom Bathroom Balcony Balcony Room02 Room03 Room13 Room14 Room15 MeetingRoom01 Room01 Room17 Room16 HW01 HW08 Balcony Balcony e1 e2 e3 e4 e5 e6 e7 e8 e9 F l

• r

E x i t 1 e10 e11 e12 e13 e14 e15 Balcony Balcony HW02 HW03 HW04 Balcony HW06 HW07 Balcony FloorExit02

FloorExit01 e9 e6 e4 e1 e5 e2 e3 e7 e8 e10 e11 e12 e13 e14 FloorExit02

hierarchical link (parent-child) horizontal link (directly accessible)

Stair01 HW03 HW02 MR01 HW01 R01 BT01 BT02 R17 R16 R15 HW04 R02 R03 BT01 R13 R14 Stair02 IMAD AFYOUNI (IRENAV) RESEARCH SEMINAR - NIS - PHC 2012 8 MAY 2012 9 / 22

RESEARCH CONTEXT METHODOLOGY & RESULTS DISCUSSION A HIERARCHICAL AND CONTEXT-DEPENDENT INDOOR DATA MODEL

A HIERARCHICAL AND CONTEXT-DEPENDENT INDOOR

DATA MODEL SPATIAL COMPONENT

The spatial component S =

i=1...|S| Si is made of a set of layers hierarchically

• rganised and representing the indoor space, and thus defining the multi-granular

spatial structure of the model.

FEATURE COMPONENT

The feature component F = (P, OOI) encompasses the features (i.e., persons and

• bjects of interest) located in the environment.

ACTION COMPONENT

The action component A represents actions that are either predefined and triggered automatically by the system in form of informative, context-aware messages, or generated by a given feature acting in the environment.

IMAD AFYOUNI (IRENAV) RESEARCH SEMINAR - NIS - PHC 2012 8 MAY 2012 10 / 22

RESEARCH CONTEXT METHODOLOGY & RESULTS DISCUSSION CONTINUOUS QUERY PROCESSING ARCHITECTURE

CONTINUOUS QUERY PROCESSING ARCHITECTURE

User

System interface Generating an SQL- like query Parsing the query Rewriting and analysing the query Is route computa- tion required ?

Route manager

• Refine candidate routes based
• n updated edge weights
• Maintain the set of relevant

routes up-to-date Generating standard queries

Query execution engine

• Update queries with the loca-

tions of relevant objects and with the new set of relevant routes, if needed

• Execute standard queries
• Correlate results of standard

queries

• Present the answer

Distribution manager

Graph & context data Graph & context data Graph & context data Graph & context data

Cancel ? stop

Phase 1

SQL-like query

Phase 2 Phase 3 Phase 4 Phase 5

yes no

Phase 7 Phase 6

yes no

IMAD AFYOUNI (IRENAV) RESEARCH SEMINAR - NIS - PHC 2012 8 MAY 2012 11 / 22

RESEARCH CONTEXT METHODOLOGY & RESULTS DISCUSSION CONTINUOUS QUERY PROCESSING ARCHITECTURE

CONTINUOUS QUERY PROCESSING ARCHITECTURE

ROUTE MANAGER

In charge of determining the best candidate routes for navigation-related queries based on context data (e.g., information about user profiles, descriptions of objects of interest, etc.)

QUERY EXECUTION ENGINE

1

Update simple queries with locations of relevant objects and with the new set of relevant routes, if needed,

2

Execute standard queries,

3

Join the results to standard queries,

4

Present the answer to the user

DISTRIBUTION MANAGER

Keeping track of the relevant servers needed to execute a given continuous query Computing the answer in a distributed environment

IMAD AFYOUNI (IRENAV) RESEARCH SEMINAR - NIS - PHC 2012 8 MAY 2012 12 / 22

RESEARCH CONTEXT METHODOLOGY & RESULTS DISCUSSION QUERY LANGUAGE FOR NAVIGATION-RELATED QUERIES IN INDOOR ENVIRONMENTS

QUERY LANGUAGE FOR NAVIGATION-RELATED QUERIES

IN INDOOR ENVIRONMENTS

General query structure Query → (Standard-query | Navigation-query) Navigation-query → select (Attr-Projections | ‘*’) from All- routes-expression (‘,’Class-names)* (with Stop-vertices) ? (where Conds) ? (optimization-criteria) ? All-routes-expression → All-routes ‘(’Loc-Ref ‘,’ Loc-Target ‘)’ Loc-Ref → Object-id (‘.’‘coord’) ? | gr ‘(’Map-id ‘,’Object-id ‘)’ | gr-map ‘(’ Map-id ’,’ Gr-id ‘)’| Vertex-id Loc-Target → Class-name | Object-id | Vertex-id ‘.’‘coord’| gr ‘(’Map-id ‘,’Class-name ‘)’ Location-dependent conditions LDQ-Cond → inside ‘(’ Args-Inside ‘)’ | nearest ‘(’Args-Nearest ‘)’| ... Args-Inside → Radius ‘,’ Loc-Ref ‘,’ Loc-Target Args-Nearest → K ‘,’ Loc-Ref ‘,’ Loc-Target Example of a navigation query : Find the shortest route from person ‘userID1’ to person ‘userID2’, showing the results at the room level : SELECT gr(‘room-level’, RO.id) FROM Person AS P1, Person AS P2 All-routes(gr(‘micro’, P1), gr(‘micro’, P2)) AS RO WHERE P1.id = ‘userID1’ AND P2.id = ‘userID2’ MINIMIZE length(RO) Example of a range query : Retrieve all the communicating entities in the vicinity (at a distance smaller than 100 meters) of a user identified by ‘userID’ and with a communication range of at least 200 meters : SELECT CO.id FROM Object AS CO WHERE inside(100 meters, gr(‘micro’,‘userID’), CO) AND CO.Communicate = true AND CO.commRange >= 200 IMAD AFYOUNI (IRENAV) RESEARCH SEMINAR - NIS - PHC 2012 8 MAY 2012 13 / 22

RESEARCH CONTEXT METHODOLOGY & RESULTS DISCUSSION QUERY LANGUAGE FOR NAVIGATION-RELATED QUERIES IN INDOOR ENVIRONMENTS

QUERY LANGUAGE FOR NAVIGATION-RELATED QUERIES

IN INDOOR ENVIRONMENTS LOCATION-DEPENDENT QUERIES

All-routes (Loc-Ref, Loc-Target) : returns a set of tuples representing the valid routes between the current locations of the reference and target objects Two kinds of location-dependent conditions can be expressed in the WHERE clause : inside(Radius, Loc-Ref, Loc-Target) and nearest(K, Loc-Ref, Loc-Target)

LOCATION GRANULES

The gr operator allows to formulate queries using the location terminology required by the user (e.g., vertices at the fine-grained level, rooms, floors, buildings, etc.) It can be referenced in the SELECT clause, the FROM clause and/or the WHERE clause of a query, for visualisation and/or processing of constraints or routes

IMAD AFYOUNI (IRENAV) RESEARCH SEMINAR - NIS - PHC 2012 8 MAY 2012 14 / 22

RESEARCH CONTEXT METHODOLOGY & RESULTS DISCUSSION CONTINUOUS PROCESSING OF LOCATION-DEPENDENT QUERIES

PROCESSING OF CONTINUOUS NAVIGATION QUERIES

STEP 1 → STEP 3 : HIERARCHICAL PATH SEARCH

1

Find the optimal path within the initial granule until reaching the nearest exit

2

Search at the abstract level for the optimal path from the exit of the initial granule to the granule containing the target object

3

Find the optimal path within the last granule to the target object starting from the corresponding entrance of the granule

IMAD AFYOUNI (IRENAV) RESEARCH SEMINAR - NIS - PHC 2012 8 MAY 2012 15 / 22

RESEARCH CONTEXT METHODOLOGY & RESULTS DISCUSSION CONTINUOUS PROCESSING OF LOCATION-DEPENDENT QUERIES

PROCESSING OF CONTINUOUS NAVIGATION QUERIES

STEP 1 → STEP 3 : HIERARCHICAL PATH SEARCH

1

Find the optimal path within the initial granule until reaching the nearest exit

2

Search at the abstract level for the optimal path from the exit of the initial granule to the granule containing the target object

3

Find the optimal path within the last granule to the target object starting from the corresponding entrance of the granule

STEP 4 : CONTINUOUS PROCESSING OF THE QUERY

4

Step 4 starts a continuous path search by taking into account updated locations

• f reference and target objects (considering moving targets)

1

Transform an initial search tree rooted by the previous vstart to an updated tree rooted by the current vstart

2

The algorithm continues either by expanding new sub-trees from the leaves towards the target and/or by removing sub-trees that are no longer needed

IMAD AFYOUNI (IRENAV) RESEARCH SEMINAR - NIS - PHC 2012 8 MAY 2012 15 / 22

RESEARCH CONTEXT METHODOLOGY & RESULTS DISCUSSION CONTINUOUS PROCESSING OF LOCATION-DEPENDENT QUERIES

PROCESSING OF CONTINUOUS RANGE QUERIES

1ST ITERATION : HIERARCHICAL NETWORK EXPANSION

1

Perform a hierarchical network expansion in all directions around the reference

• bject

2

Keep all visited nodes along with pointers to their parent nodes as well as weights to the source node

IMAD AFYOUNI (IRENAV) RESEARCH SEMINAR - NIS - PHC 2012 8 MAY 2012 16 / 22

RESEARCH CONTEXT METHODOLOGY & RESULTS DISCUSSION CONTINUOUS PROCESSING OF LOCATION-DEPENDENT QUERIES

PROCESSING OF CONTINUOUS RANGE QUERIES

1ST ITERATION : HIERARCHICAL NETWORK EXPANSION

1

Perform a hierarchical network expansion in all directions around the reference

• bject

2

Keep all visited nodes along with pointers to their parent nodes as well as weights to the source node

CONTINUOUS PROCESSING OF THE QUERY

3

Update the set of parent nodes when changing the root of the sub-tree (i.e., when the reference object moves)

4

Boundary nodes are checked to decide, for each of them, whether to further expand this node or to perform a reverse search towards the source to remove nodes that are not relevant any more

IMAD AFYOUNI (IRENAV) RESEARCH SEMINAR - NIS - PHC 2012 8 MAY 2012 16 / 22

RESEARCH CONTEXT METHODOLOGY & RESULTS DISCUSSION

TALK OUTLINE

1 RESEARCH CONTEXT

Location- and context-aware services and queries in indoor environments Research challenges

2 METHODOLOGY & RESULTS

Preliminary recommendations A hierarchical and context-dependent indoor data model Continuous query processing architecture Query language for navigation-related queries in indoor environments Continuous processing of location-dependent queries

3 DISCUSSION

Implementation Experience Achievements and Status of Research Activities Research perspectives

IMAD AFYOUNI (IRENAV) RESEARCH SEMINAR - NIS - PHC 2012 8 MAY 2012 17 / 22

RESEARCH CONTEXT METHODOLOGY & RESULTS DISCUSSION IMPLEMENTATION EXPERIENCE

IMPLEMENTATION EXPERIENCE

IMPLEMENTATION REQUIREMENTS

An extensible DBMS that supports ⋆ developing network-based data models ⋆ new algebra (i.e., specific data types and operations) ⋆ implementing algorithms for LDQ processing over moving objects Continuous query processing architecture : could integrate a simulator of moving

• bjects

A STAND-ALONE PLATFORM ?

Postgresql (Potential extensions : PgRouting, TelegraphCQ) Operators and location-dependent constraints are implemented as pl/pgsql functions applied on user-defined types A potential use of TelegraphCQ extension is expected for expressing continuous queries over spatial streams

IMAD AFYOUNI (IRENAV) RESEARCH SEMINAR - NIS - PHC 2012 8 MAY 2012 18 / 22

RESEARCH CONTEXT METHODOLOGY & RESULTS DISCUSSION ACHIEVEMENTS AND STATUS OF RESEARCH ACTIVITIES

ACHIEVEMENTS AND STATUS OF RESEARCH ACTIVITIES

CONTRIBUTIONS OF THE THESIS

A hierarchical and context-dependent indoor data model A generic architecture to process continuous location-dependent queries in indoor environments A continuous indoor-based query language for location-dependent queries Hierarchical and incremental algorithms applied to both static and moving objects are proposed for continuous navigation and range queries

IMAD AFYOUNI (IRENAV) RESEARCH SEMINAR - NIS - PHC 2012 8 MAY 2012 19 / 22

RESEARCH CONTEXT METHODOLOGY & RESULTS DISCUSSION ACHIEVEMENTS AND STATUS OF RESEARCH ACTIVITIES

ACHIEVEMENTS AND STATUS OF RESEARCH ACTIVITIES

CONTRIBUTIONS OF THE THESIS

A hierarchical and context-dependent indoor data model A generic architecture to process continuous location-dependent queries in indoor environments A continuous indoor-based query language for location-dependent queries Hierarchical and incremental algorithms applied to both static and moving objects are proposed for continuous navigation and range queries

COLLABORATIONS

A fruitful collaboration with Dr. Ilarri from University of Zaragoza, Spain, within the context of the MOVE COST Action IC0903 Participating in the PHC Pavle Savic 2012 : CG&GIS Lab from University of Nis - Naval Academy Research Institute (IRENAV)

IMAD AFYOUNI (IRENAV) RESEARCH SEMINAR - NIS - PHC 2012 8 MAY 2012 19 / 22

RESEARCH CONTEXT METHODOLOGY & RESULTS DISCUSSION RESEARCH PERSPECTIVES

RESEARCH PERSPECTIVES

CURRENT AND FURTHER WORKS

Analysing the efficiency and scalability of the solutions proposed to deal with continuous location-dependent queries Developing a complete context model by integrating user activities as well as user-generated contents into query processing Generalising the hierarchical data model to higher levels of abstraction (floor and building levels)

IMAD AFYOUNI (IRENAV) RESEARCH SEMINAR - NIS - PHC 2012 8 MAY 2012 20 / 22

RESEARCH CONTEXT METHODOLOGY & RESULTS DISCUSSION RESEARCH PERSPECTIVES

PUBLICATIONS

• I. Afyouni, S. Ilarri, C. Ray, and C. Claramunt.

Context-aware modelling of continuous location-dependent queries in indoor environments. Journal of Ambient Intelligence and Smart Environments (JAISE, Impact factor : 1.5), accepted, pages 1–22, 2012.

• I. Afyouni, C. Ray, and C. Claramunt.

Spatial models for indoor and context-aware navigation systems : A survey. Journal of Spatial Information Science (JOSIS), accepted, pages 1–43, 2012.

• I. Afyouni, C. Ray, S. Ilarri, and C. Claramunt.

Un modèle de données pour les requêtes sensibles au contexte dans les environnements “indoor”. In Actes des Huitièmes Journées Francophones Mobilité et Ubiquité (UBIMOB), pages 1–11. Cépaduès, 2012.

• I. Afyouni, C. Ray, and C. Claramunt.

A fine-grained context-dependent model for indoor spaces. In Proceedings of the 2nd ACM SIGSPATIAL International Workshop on Indoor Spatial Awareness, pages 33–38. ACM, 2010. IMAD AFYOUNI (IRENAV) RESEARCH SEMINAR - NIS - PHC 2012 8 MAY 2012 21 / 22

RESEARCH CONTEXT METHODOLOGY & RESULTS DISCUSSION RESEARCH PERSPECTIVES

Thank you very much for your attention ! Imad Afyouni

supervised by :

• Pr. Christophe Claramunt

&

• Dr. Cyril Ray

Naval Academy Research Institute, France imad.afyouni@ecole-navale.fr

IMAD AFYOUNI (IRENAV) RESEARCH SEMINAR - NIS - PHC 2012 8 MAY 2012 22 / 22