1. Introduction

The emerging mobile Internet makes it possible for users to access a myriad of services and applications while on the move. At the same time, it imposes constraints that require higher levels of automation than on the fixed Internet (e.g. limited input/output functionality of mobile devices, time-critical, goal-driven tasks, and scenarios where users are subject to many more distractions). Context awareness offers the prospect of (semi)automatically matching users with content/services that are relevant to their locations, activities and a myriad of other contextual attributes. The number of possible applications seems limitless (e.g. context-aware message filtering, context-aware travel planning, context-aware restaurant concierge, context-aware notification agents, etc.). Today however information about a user’s context is distributed across a number of heterogeneous resources. His location might be stored in the Home Location Register of his mobile phone operator, his activities in his calendar, the list of his colleagues in a company database, etc. To make matters worse, personal/contextual information is generally represented in an ad hoc fashion, making it prohibitively expensive to develop and maintain applications that could actually make sense out of it. The same is true for Web services, which all rely on idiosyncratic ways of representing information and interacting with the user. This situation is not specific to civilian scenarios. When it comes to developing context-aware military applications, relevant information services and contextual resources are similarly spread across a broad range of heterogeneous systems, making it nearly impossible to dynamically identify and re-use them in support of new applications.

The objective of this project is to demonstrate and evaluate the use of Semantic Web technologies in support of mobile context-aware scenarios with a particular emphasis on the development and evaluation of automated discovery and access of personal resources (e.g. calendar, location tracking, organizational databases, etc.) and Web services. Technology evaluation and transfer is being conducted in the context of both civilian and DoD scenarios. Civilian evaluation is taking advantage of Carnegie Mellon University’s campus-wide wireless LAN (which includes location tracking functionality) and revolves around a growing collection of task-specific, context-aware agents aimed at enhancing everyday campus life. DoD evaluation and technology transfer revolves around a series of increasingly more sophisticated DAML Experiments and builds on initial work carried out jointly by BBN and CMU in FY02 where CMU integrated and demonstrated a context-aware notification agent in the context of the Semantic Operational Net Assessment Tool (SONAT).

2. Work to be Conducted in FY2003

Work in FY2003 will focus on 3 related areas:

Semantic eWallet: Automated discovery and access of personal/contextual resources
Semantic Web Services: Automated discovery and access of Web Services
Contribution to the DAML Experiment – context-aware notification agent

This is further detailed below.

2.1 Semantic eWallet: Automated Discovery and Access of Personal/Contextual Resources

Different people have different sets of personal/contextual resources available. Some may be using a calendar, others may not; some may have access to location tracking functionality, others may not. Organizational information (e.g. who reports to whom; who is responsible for what) or social relationships (e.g. who your friends/colleagues/family members are) may or may not be available and can be stored in a number of different resources, etc. In fact, resources available for a given user may even change over time (e.g. a user moves outside of the area covered by some location tracking functionality) and some of the same information may be accessible through different resources (e.g. a user’s location may be available through location tracking functionality or by checking his or her calendar). We define a Semantic eWallet as a semantic directory (similar to a DAML-S directory) of personal/contextual resources available for a given user. Semantic e-Wallets enable developers to build context-aware applications/agents that can automatically discover and access relevant personal/contextual resources available for a given user, thereby eliminating the prohibitive development and maintenance costs otherwise associated with context-aware applications aimed at serving an heterogeneous population of users – e.g. not everyone is using Outlook Calendar.

This task will demonstrate and evaluate the use of a Semantic eWallet in support of several context-aware agents. This will include the automated discovery and access of different sets of personal/contextual resources by context-aware agents. Our objective is to enable a user to pull a new context-aware agent into his/her personal environment and have the agent automatically customize itself by discovering and accessing relevant personal/contextual resources through the user’s Semantic eWallet. We plan to use DAML-S in this task and expect in the process to propose modifications/refinements of the language through collaboration with the DAML-S coalition.

Metrics:

Ø Number of personal resources referenced in the user’s eWallet

Ø Expressiveness:

o Size of contextual/personal resource ontology and variety of personal/contextual resources that can be accommodated

o Variety of context-aware agents that can be supported

2.2 Semantic Web Services: Automated discovery and access of Web Services

The prototype semantic web environment for mobile, context-aware services developed by our group in FY02 relied on SOAP and an extension of UDDI for Web services – over 20 web services were developed. In FY2003, we plan to refine a number of these services (e.g. restaurant services, weather service, etc.) using DAML-S, extend the matchmaking functionality developed in FY02 to take full advantage of DAML-S, and make available to the DAML user community a number of our services in the form of DAML-S use cases. This work will also include close interactions with the DAML-S coalition in the form of possible suggestions for improvement/refinement of the language.

Metrics:

Ø Number and variety of Web services developed

Ø Expressiveness supported by our matchmaking functionality

2.3 Contribution to the DAML Experiment

In FY02, we demonstrated two variations of a context-aware SONAT notification agent capable of taking into account information about the user’s profile, his/her message forwarding preferences and calendar activities, where all this information was represented in DAML+OIL. We plan to extend this functionality in support of more realistic (and more sophisticated) scenarios to be identified jointly with BBN and prospective DoD users. We expect that this will include a demonstration of our semantic eWallet.

Metrics:

Ø Number and variety of contextual attributes and preferences that can be accommodated

Ø Additional metrics to be defined jointly with BBN and prospective DoD users

3. Next Logical Steps (Beyond FY03)

Work in FY04 and beyond will proceed along the following three dimensions:

Semantic eWallet Access Control: Extending DAML-S to capture a rich set of access control constraints (e.g. what personal resources should be visible/accessible to whom and under which conditions). This work will include contributions to DAML-S and the work of the DAML-Security working group.
Integration of AI planning and automatic Web Service/Personal Resource Discovery and Access: Here, our group will demonstrate and evaluate (semi)automatic planning functionality that leverages Web Service and Personal Resource discovery and access.

As in FY03, this work will be demonstrated and validated in the context of both civilian and DoD scenarios, the latter through collaboration with BBN and prospective DoD users in the context of increasingly more sophisticated variations of the DAML Experiment. Opportunities for technology transfer beyond the DoD are also being pursued with several companies.