Context Information

Communication technology and digital technology in general, is getting more and more integrated. Because of the continuous improvements of different embedded actuators and sensors, these technologies gain access to many different areas of everyday life. Today, we are already using many different digital devices, which are able to collect data about their users, or about their environment. Examples for such devices can be found in many different domains, like: Thermomethers, which collect temperature information and control air conditions, cellular phones which collect data about the users communication, and which offer different services, like sending an SMS(=Short Message Service). Stereo, Video and TV devices have the possibility to analyze the users personal program favors and react on a special user.

The problem is, that these digital devices are isolated from each other and are not able to gain access to data another device has collected. It would be convenient, if an organizer (like Palm, iPAQ), could access the context information from all embedded devices around. Embedded context aware devices could change automatically their own configuration, depending on the context they are situated.

For this purpose, it is necessary to rely on a general framework, which is able to deliver a suitable and flexible world model, describing objects in the real world, with their context information. It should be easy to enable an existing device (both hardware and software devices), to offer its own context information to other devices. For this purpose it is also necessary to specify which data is part of context information and how context information and context sensitive applications are defined. A powerful and flexible world model should be able to adapt every aspect of a changing world, in order to provide an effective bridge for human computer interaction.

This paper addresses the research issues associated with the vision of enabling context aware computing. The structure of the paper is as follows. Section 2 presents the use of context information world models and presents already existing world models with their definitions. Section 3 introduces our thoughts about a new context information world model, combined with dynamic context rules, which are able to model dynamic relations between objects.

World Model and Context Framework Middleware

The definition of a world model helps to understand which objects are relevant for describing certain situations. A situation or an environment is also described by the relation between the different objects, which are involved. The context information world model defines every information a context sensitive application is able to retrieve. The world model consists of semantic notations. When new information occurs, this semantic notations are used to map the information into the existing world model. The framework middleware is responsible for the gathering, mapping, representation and transport of relevant data into a world model. So the implemented middleware manipulates the data in the world model at runtime. This basic relation is shown in Fig. 1. Sensors gather information from the real world into the context framework. On the other side, the context framework is able to change the real world objects with actuators. Dynamic context rules are able to express the relation between the different context objects. The context world model is just a small view on the real world which is not able to present all objects. Additionally there exists an interface, which is used by context sensitive applications to access the information about the world, modeled by the world model.

Fig.1. Shows the relation between the context world model and the context framework middleware.

Open World vs. Closed World Assumption

When modeling an environment there are basically two different approaches that heavily influence the implementation of context information middleware:

1.        Closed World Assumption: Data and Services are completely known at design time. Therefore it is easy to define a suitable semantic model in order to handle these information. It is not possible to add any unknown element in a closed world assumption. Moreover parts of information are explicitly excluded, in order to relieve clear reasoning.

2.        Open World Assumption: Data and services are not known at design time. Therefore it is necessary to derive all the information from the semantic description of the data collected. This issue is hard to realize and a major problem in AI-research.

Demonstration Material

    Emergency scenario model 2002

    Tangible context aware artifacts 2003

Related research projects

Hewlett-Packard Project CoolTown

Xerox PARC ubiquitous computing environment

IR related links