MANAGING CONTEXT DATA IN PERSONAL SMART SPACES

Nicolas Liampotis, Nikos Kalatzis, Ioanna Roussaki, Efstathios Sykas and Miltiades Anagnostou

School of Electrical & Computer Engineering, National Technical University of Athens, Athens, Greece

Keywords: Personal Smart Space, Context management, History of context, Context refinement, Community-

awareness.

Abstract: Current research on pervasive computing is opening the way for convergence between mobile

telecommunications and the Future Internet. This paper introduces a novel approach to this convergence in

the form of the self-improving Personal Smart Spaces (PSSs). One of the core aspects that need to be

supported within PSSs is the management of the user related context information. This paper elaborates on

the mechanisms that have been designed and implemented in order to address the advanced requirements of

PSSs regarding the establishment of a robust distributed context management framework.

1 INTRODUCTION

Pervasive computing (Schmidt, Spiekermann,

Gershman, and Michahelles, 2006) is a next

generation system paradigm that aims to assist users

in their everyday tasks in a seamless unobtrusive

manner. It assumes that users are surrounded by

numerous communication and computing devices of

various features, which interoperate and are capable

of capturing and processing information regarding

users, their behaviour and their environments. In this

framework, there have been various research

initiatives aiming towards the design and realisation

of smart spaces (Singh, Bhargava and Kain, 2006),

where various automation facilities support the

users. However, these are fixed spaces that provide

pervasive features and adapt to the user needs in a

static and geographically limited environment.

Nevertheless, when dealing with mobile users,

different and more challenging problems are

introduced and need to be resolved (Hansmann,

Merk, Nicklous and Stober, 2003; Huang and

Mangs, 2008). In this case, the users require the

same pervasive services wherever they are and

whatever devices they carry along.

Personal Smart Spaces (PSSs) (Roussaki et al.,

2009) aim to couple the facilities offered by next

generation mobile communications with the features

provided by the static smart spaces to support a more

ubiquitous and personalised smart space that is able

to follow the user wherever he/she goes. A PSS

provides to its owner multimodal intelligent

interfaces, via which he/she is able to access and

configure the various services and resources that are

available locally and remotely, even when limited or

even no network connectivity is available. PSSs are

able to discover other PSSs and interact with them in

order to create a richer and more flexible

environment for their owners. Each PSS consists of

multiple devices, both mobile and fixed, owned by a

single user. PSSs constantly monitor their owner’s

behaviour & environment and they exploit learning

techniques to further optimise the pervasive

experience perceived by their owners.

In PSSs, network operators, individual sensors,

sensor networks or even web and computing

resource managers capture valuable information

(e.g. device location/status, user profiles, movement

patterns, user activities, network performance) and

exploit them to provide enhanced services to users.

These are called context-aware services (Anhalt et

al., 2001; Loke, 2006) and are based on the

exploitation of context data such as the above.

The PSS Context Management (CM) framework

acts as an intermediate layer between PSS/3

party

context-aware services and the sources of context

information. The components comprising the CM

architecture are: Context Broker, Context

Refinement, Context History Management, Context

DB Management, and Context Source Management.

A description for each component is provided in

subsequent sections of this paper. Apart from the

components above, the CM architecture comprises a

Database Management System (DBMS) which

enables access to the actual context repositories, i.e.

the Context Database and the History of Context

288

Kalatzis N., Liampotis N., Roussaki I., Sykas E. and Anagnostou M..

MANAGING CONTEXT DATA IN PERSONAL SMART SPACES.

DOI: 10.5220/0003374302880291

In Proceedings of the 1st International Conference on Pervasive and Embedded Computing and Communication Systems (PECCS-2011), pages

288-291

ISBN: 978-989-8425-48-5

 2011 SCITEPRESS (Science and Technology Publications, Lda.)

Database (HoC DB). The former database is used for

current context information, while the latter stores

past data.

This paper briefly presents the functionality

offered by the components above. More specifically,

it is structured as follows. In Section 2, the features

implemented by the components that are responsible

for sensing, maintaining, managing and distributing

context information are descibed. In Section 3, the

historic context data management approach, as well

as the context refinement mechanisms are shortly

presented. Section 4 elaborates on the requirements

that need to be addressed to extend the PSS CM

framework so that community-related features are

supported. Finally, in Section 5, the paper

conclusions are drawn.

2 CONTEXT SENSING,

MANAGEMENT AND

DISTRIBUTION

In this section, the functionality of the CM

components, which are responsible for sensing,

managing and distributing context information, is

briefly presented.

2.1 Context Broker

Access to context information is managed by the

Context Broker. This component provides a query

interface for retrieving, updating, adding or

removing context data. Context queries can be

classified as follows: (i) CtxIdentifier-based, where

Context consumers can specify context IDs in their

context queries in order to identify the data items

they are interested in; (ii) location-based, which is

actually a special use case for ID-based queries and

allows for discovering new context information

based on location hierarchies modelled by context

entities and their associations; (iii) temporal-based,

which provides context consumers with access to

past values of context data. The actual processing of

such data is managed by the ContextHistoryMgmt

component which is detailed in Subsection 3.1. It

should be noted that context retrieval queries can be

performed either synchronously or asynchronously.

Apart from context query handling, the Context

Broker supports a hybrid distributed-centralised

management of context data. The hybrid approach

was chosen over a centralised or full distributed one

in order to address scalability issues in terms of

processing, storage, remote communication and,

power consumption. More specifically, in this

approach, each PSS device contributes to collecting

context information and is assigned with a different

role based on its capabilities. Although every PSS

member hosts a context database, context

information is not replicated in all of them. There is,

however, a single device (master), typically one with

server capabilities, i.e. high processing power and

storage capacity, whose database contains every

piece of context information and is thus able to

handle all context queries. In addition, the master

device is responsible for maintaining historic context

data in order to support inference of current and

future context. Regarding the other (slave) devices

in the PSS, context queries cannot always he

handled locally and must therefore be forwarded to

the master device for further processing. If the

master device becomes unavailable, a slave one may

be elected in replacement.

2.2 Context DB Management

The PSS CM framework maintains two databases

for storing context information; one for current

values and the other for historical data. The scope of

the Context DB Management (CDBM) component is

to serve as an intermediate layer between the context

query interface provided by the Context Broker and

the underlying DBMS that actually controls the

storage, management and retrieval of data in these

two databases. Therefore, this intermediate layer is

responsible for translating context queries submitted

by platform and 3

party context-aware services via

the Context Broker component into standard SQL

queries that can be executed in the framework’s

DBMS. More specifically, the CDBM component

utilises Hibernate (Hibernate, 2010), an object-

relational mapping library, in order to map context

model objects, i.e. entities, attributes & associations,

to records in a traditional relational database.

2.3 Context Source Management

The Context Source Management (CSM) component

serves as an interface between a diversity of context

sources and the Context Database. These sources

include sensors, as well as, context-providing

services that are able to feed the CM framework

with context information following a source

registration process. Furthermore, the CSM is

responsible for monitoring the quality characteristics

of available context sources in order to be able to

select the most appropriate source in the presence of

multiple sources for the same type of context

information. This decision is based on the QoC

requirements posed by context consumers. Such

MANAGING CONTEXT DATA IN PERSONAL SMART SPACES

289

requirements may also lead to the re-configuration

of registered context sources.

3 CONTEXT HISTORY AND

REFINEMENT

The functionality of the CM components that are

responsible for managing historic context data, as

well as, refining context information is described in

the subsections that follow.

3.1 Context History Management

The main functionality of the Context History

Management component is to collect and manage

History of Context (HoC) data (Kalatzis et al.,

2008). Maintaining these data is of great importance

to the CM, as it enables the system to provide old

data that are no longer stored in the Context DB.

This is even more critical for other components of

the PSS architecture, as it supports the provision of

the necessary training data sets to the components

that facilitate the self-improving and pro-active

behaviour. As already stated, a PSS consists of

various devices, each of which hosts a CM system.

However, the HoC database is maintained only on

the master device of the PSS, which has rich storage

and processing resources. It should be pointed out

that the CM does not automatically monitor all types

of context information and store the respective past

values in the HoC database. Instead, the

services/components interested and/or the PSS

owner or a system administrator have to explicitly

specify the context types to be maintained in the

HoC database.

There are several components of the PSS

framework that make use of the stored historic data.

Such a component facilitates predictions of the

user’s next action regarding the usage of pervasive

services. Context history data sets are also used for

creating profiles of user’s preferences, for providing

recommendations regarding service usage and are

also used by context reasoning algorithms.

Retrieving historic data for a specific context

attribute is a straightforward process provided by

various methods of the Context History

Management interface. Those methods implement

standard SQL queries based on the attribute

CtxIdentifier in question and criteria related with the

timestamps or/and the values of the records.

However, there are cases where context history

consumers require data sets containing sequences of

context data escorted by other context values

occurred at the same time. For example, a sample

data set containing the values for a sequence of a

context attribute A escorted by the values of context

attributes B and C will form tuples, as follows:

<CtxAttrA=val1, CtxAttrB=val2, CtxAttrC=val3>

<CtxAttrA=val4, CtxAttrB=val5, CtxAttrC=val6>

<CtxAttrA=val1, CtxAttrB=val2, CtxAttrC=val8>

In order for the HoC component to support these

types of queries, an additional registration process is

required prior to the storage of the records. The

Context History Management provides methods

allowing the history consumer to specify the primary

attribute identifier to be monitored along with the

identifiers of the escorting attributes. After this

registration process, every update of the primary

attribute will trigger the storage of each of the

escorting attributes in the HoC DB along with the

primary attribute.

3.2 Context Refinement

Context refinement implies the processing of raw

context data captured by sensors and other sources,

in order to extract high level context information.

The overall context refinement process consists of

two main steps. The system initially extracts context

inference rules based on context history data. This

process takes place on the master device. The

inference rules are then distributed to all PSS

devices capable of performing context refinement.

When a context consumer submits a request for

context information that is neither available by

sensors, nor lies in the context database, the

refinement process is triggered automatically in an

attempt to provide the necessary results. The actual

inference takes place based on the distributed rules

that are used to process raw sensor data. Thus, even

the resource constrained devices are capable of

performing context refinement locally without

having to contact another system. The overall design

of the context refinement mechanisms and the

implemented inference algorithms are extensively

described in (Frank et al., 2009).

4 EXTENDING THE CM IN

SUPPORT OF COMMUNITIES

The described CM approach fulfils all the necessary

requirements in order to support context awareness

in mobile and fixed PSSs. However, as the focus of

PSS is the individual rather than the communities of

users, an important part of human behaviour is

neglected: socialising. Social computing, on the

PECCS 2011 - International Conference on Pervasive and Embedded Computing and Communication Systems

290

other hand, has enjoyed meteoric success in bringing

people together online. The time is ripe for these two

paradigms to converge. But in order to extend the

PSS beyond the individual to dynamic communities

of users, the context management system described

in the previous sections needs to address additional

requirements, thus introducing several research

challenges. The new requirements that need to be

met include the following:

 Community context modelling, management

and inference

 Shared community context and inherited

context from hierarchical communities

 Dynamic community creation and membership

based on context similarities

 History of community context management and

exploitation

 Community enhanced locating of individuals

with user involvement in location tagging

All requirements above introduce new

promissing research fields that are currently being

investigated by the authors.

5 CONCLUSIONS

Personal Smart Spaces (PSSs) is a notion introduced

to bridge the gap between isolated fixed smart

spaces and eliminate the islands of pervasiveness

that may exist, where users have no access to

pervasive services while being disconnected from

fixed smart spaces. The delivery of pervasive

services in PSSs relies on the implementation of a

context management system that enables third party

and PSS services to utilise context data and interact

with the user at the appropriate time and in the

appropriate manner and get personalised and

adapted to the user's requirements and context. The

context management system collects & stores these

context data and maintains the history of context

information. It also provides context data to other

PSS components in support of the learning,

reasoning and self-improvement features of PSSs.

The CM framework presented in this paper is

extendible to address the needs not only of

individual users, but also of dynamic communities

of users. This extension introduces several research

challenges and new concepts, such as community

context, community-enhanced context handling and

inference, etc. The implementation of the PSS

context management system framework started in

the beginning of 2009. The first prototype has been

delivered in September 2009 and since then various

versions were released. The final and complete

version of the PSS framework was released on

September 2010

(http://psmartspace.sourceforge.net/).

ACKNOWLEDGEMENTS

The research leading to these results has received

funding from the European Community's Seventh

Framework Programme [FP7/2007-2013] under

grant agreement n° 215098 of the PERSIST

(PERsonal Self-Improving SmarT spaces)

Collaborative Project, as well as under grant

agreement n° 257493 of the SOCIETIES (Self

Orchestrating CommunIty ambiEnT IntelligEnce

Spaces) Collaborative Project.

REFERENCES

Anhalt, J., Smailagic, A., Siewiorek, D., Gemperle, F.,

Salber, D., Weber, S., Beck, J., Jennings, J., 2001.

Toward Context-Aware Computing: Experiences and

Lessons, IEEE Intelligent Systems, 16(3), 38-46.

Frank, K., Robertson, P., McBurney, S., Kalatzis, N.,

Roussaki, I., Marengo, M., 2009. A Hybrid Preference

Learning and Context Refinement Architecture,

PERSIST Workshop on Intelligent Pervasive

Environments (AISB 2009), Edinburgh, Scotland.

Hansmann, U., Merk, L., Nicklous, M. S., Stober, T.,

2003. Pervasive Computing: The Mobile World, New

York, USA: Springer Professional Computing.

Hibernate, 2010. Retrieved November 11, 2010, from

http://www.hibernate.org/

Huang, S., Mangs, J., 2008. Pervasive Computing:

Migrating to Mobile Devices: A Case Study. In 2

Annual IEEE Systems Conference. Montreal, Quebec,

Canada.

Kalatzis, N., Roussaki, I., Liampotis, N., Strimpakou, M.,

Pils, C., 2008. User-centric inference based on history

of context data in pervasive environments, 3

ACM

International Workshop on Services Integration in

Pervasive Environments (SIPE’08), Sorrento, Italy.

Loke, S. (2006). Context-Aware Pervasive Systems:

Architectures for a New Breed of Applications.

Florida: Auerbach Publications.

Roussaki, I., et al. (2009). Persist Deliverable D2.5:

Revised Architecture Design, Technical report,

PERSIST project (FP7-ICT-2007-1).

Schmidt, A., Spiekermann, S., Gershman, A., Michahelles,

F. (2006). Real-World Challenges of Pervasive

Computing. IEEE Pervasive Computing, 5(3), 91-93.

Singh, R., Bhargava, P., Kain, S.(2006). State of the art

smart spaces: application models and software

infrastructure. ACM Ubiquity, 7(37), 2-9.

MANAGING CONTEXT DATA IN PERSONAL SMART SPACES

291