Ruaidhrí Power, Declan O’Sullivan, Owen Conlan, David Lewis, Vincent Wade

Resolving Queries in a Heterogeneous Context Rich Environment

Knowledge and Data Engineering Group
Department of Computer Science
Trinity College Dublin

The vision of ubiquitous computing[1] is that of many computing devices interacting in a natural way with humans in the real world. The software running on these devices is currently limited in how it can provide for users’ needs by the amount and quality of the information it can retrieve about the environment in which it is operating. This context information must be provided in a form each software component can understand, which is a difficult problem given the wide array of heterogeneous information sources involved in any ubiquitous computing scenario such as computers, embedded sensors and information appliances.

One of the driving forces behind the design for a context system proposed in this paper is to minimise the effort required to make a piece of software context aware. These software components will come in many forms, from the e-mail clients and office tools that are prevalent today, to tiny embedded operating systems with minimal processing power, to massive mainframe or cluster computers running large databases. For the purpose of this paper, any of these software components are referred to as applications. While this term may bring to mind today's software which is not context aware, throughout this paper it refers to any software component which wishes to make use of context information. These applications need to have easy access to more information about the environment in which they are operating, rather than being limited to the explicit input or information from hardwired data sources provided to current applications.

In this architecture the ‘context service’ is the service provided to applications to make context information available to them. One role of a context service is to take queries from a context-aware client and to resolve those queries by acting as a mediator between the client and other information sources that the service has access to. As well as acting as consumers of context information (by executing queries), applications can also act as producers of context information by providing their context service with a description of the information they have available. If an application produces context information, a context service can advertise that information available to it to other context services.

This design uses an ontology-driven approach to bridge the heterogeneity of context information sources in ubiquitous computing systems. Ontologies are a technique for formally representing domain knowledge in an application independent way. Ontologies feature heavily in the Semantic Web initiative[2], which aims to provide ways of defining information so that it can be understood and processed by computers more easily. Examples of ontology languages are W3C's OWL1, the Web ontology language and DARPA's DAML2.

In summary, this paper proposes an ontology-driven context system for heterogeneous context-rich environments that aims to minimise the effort required to make an application context-aware.

The paper is laid out as follows: Section 2 describes the state of the art with regard to integration of heterogeneous context information. Section 3 describes the example context scenario covered in this paper, and section 4 describes the process of designing a context-aware application. Section 5 describes how a context-aware device would operate, describing the interactions between applications and the context service. Section 6 describes the internal structure of a context service, which consists of query interface and analysis (section 6.1), query decomposition (section 6.2) and query routing (section 6.3). Section 7 describes our experimental work to date, and section 8 concludes the paper.

An initial experiment to verify the approach presented in this paper is underway, and is due for completion in September 2004. The full version of this paper will provide the results of this experiment.