SFB 627: Nexus
home uni uni suche search sitemap sitemap kontakt contact
unilogo University of Stuttgart
SFB 627: Nexus

Scientific Challenges

Deutsche Version
 
The development of global world models not only comprises research questions concerning the modeling and federated management of models, but also concerning the communication, the integration of sensor data and the presentation of model information. With respect to the acceptability of the approach, convincing security and privacy concepts as well as the consideration of social aspects are essential. Therefore we see research challenges in the following areas:

 Modeling and Extensibility Concepts
 Federated Model Management
 Integration of Temporal Aspects
 Generic Integration of Sensor Data
 Concepts of Consistency
 Model-based Communication
 Security Concepts
 Automatic Acquisition of Model Data
 Model Presentation and Interaction
 Security, Acceptability, Business Models
 Quality of Context Information
 Higher Level Context
 Homogenized Query Processing
 Application Support

Modeling and Extensibility Concepts


For applications to be able to run in a common technical and semantical environment instead of isolated from each other, an extensible context model is required, which is able to represent different kinds of context in a uniform way, especially geographical context (streets, buildings, rooms, etc.), dynamic context (digital documents or virtual information towers) as well as technical context, which similarly represents the infrastructure available (e.g. the extent and bandwidth of wireless networks).

Federated Model Management


The federated model management shall anytime provide a uniform view on a variety of context model to applications. This forms a World Wide Space, which consists of many heterogeneous model parts offered by different providers in an open, largely distributed system. This requires a flexible and scalable architecture, in which a large number of different stationary and mobile system components cooperate, whereas the system's configuration can change anytime in a dynamic way.

Integration of Temporal Aspects


When introducing temporal aspects, also past states of the world model are incorporated, i.e., updating the model does not override old values, but keeps them with a time stamp. Managing data histories of mobile objects is especially challenging, because their spatial relation changes over the time. The collected data histories can be used for predictions and in recognition of situations.

Generic Integration of Sensor Data


In the future, selective sensor integration processes have to consider the particular context of an application. Therefore cognitive behavior patterns are to be developed, which permit it to agents, to combine the sensors, which are needed for a task, and to fuse data of sensors of several agents selectively. The context and its inconsistencies are to be determined by behavioral models, the sensor integrations situation shall be described and appropriate real time based actions shall be triggered to reduce the inconsistencies.

Concepts of Consistency


At first only accurate data has been used concerning questions of consistency. In a further step context quality was in the centre of the research efforts. The introduction of the so called demoted information requires at the same time the definition of a quality calculus for the determination of the demotion of context data as well as the deduction of higher level context, which is distinguished by characteristics of demotion. The definition of such calculations is done with the help of the installation of the cross section project Q in the current promotion period. The work concerning basic consistency concepts was finished in 2006 and documented by a subproject-spreading consistency report. Those concepts serve as a guideline for all projects.

Model-based Communication


The Nexus system is open for a broad spectrum of end systems, communication networks, and usage scenarios. In such a heterogeneous environment, efficient communication – in particular the transmission of model information – requires cross-layer and cross-application adaptation control mechanisms. To optimize adaptation decisions, context information like currently available communication resources or the current user situation can be utilized.
Context information from the world model can also be used to distributed information selectively among users with a certain context, e.g. users at a certain location or in a certain situation. Special overlay network structures and context-based routing algorithms are required for efficient delivery of such contextcast messages.

Security Concepts


Users, operators, and service providers of context-aware systems have different security requirements. Maintaining control over their own data is a major requirement of users, whereas operators and content providers need to ensure that their content and services are used by authorized users only and that it is possible to measure and account for their utilization. Therefore, a prerequisite for the acceptability of context-aware services is a security platform that takes into account these requirements as well as further legal and regulatory conditions.

Automatic Acquisition of Model Data


Detailed spatial world models require the automatic acquisition of model data as well as the automatic generation and update of the spatial models. Existing methods for model data collection are to be augmented by sensor fusion techniques and additional information sources as for example natural language text modules. Basic prerequisites for the evaluation of context quality are the acquisition, storage and processing of quality information. For this purpose, quality data has to be integrated into the world model management in order to make it available to the applications.

Model Presentation and Interaction


In the area of model presentation and interaction efficient methods are developed to provide an intuitive access to information of the augmented world model for users and applications, especially for projects of the collaborative research centre. Therefore, the challenge is to determine an adequate visualization technique, based on context information (position, direction, client device, meta data, data type, etc.), and adapt it correspondingly. The integration of context-aware interaction methods allows controlling the visualizations based on context information, as e.g. position and direction.

Security, Acceptability, Business Models


Security, acceptability and basic economic questions vastly dominate large parts of the research in the Centre of Excellence. Goal of the second funding period is, to meet the different security requirements of the different stakeholders and users of context-aware applications in a well balanced way. Therefore in an interdisciplinary research cooperation an all comprising, multilateral and economically sustainable security architecture has to be developed, in which the different bounding conditions of the system as well as the security requirements of the participants have to be identified, integrated and balanced.

Quality of Context Information


The quality of context information may highly differ. The reasons are the heterogeneity of context information and the variety of approaches for context acquisition. Processing and federating context information of different providers therefore requires a reference model for assessing context quality. This model has to incorporate the various abstraction levels of context information (sensor data, observable context, and situation) as well as the different aspects of context quality (degradation, consistency, trust).

Higher Level Context


The environment model administers observable context information that is gained for example from sensors. It was however shown that many context-referred applications are in need of context information containing higher level context data (situations), which must be derived from observable context data.
Information that is called higher level context is derived from existing observable context data. As the conduction of the situation can be very complex and can also depend on the context, it is reasonable to develop general and re-usable methods and concepts. Those must be developed on application level first and have to be integrated into the environment model data administration later.

Homogenized Query Processing


To cope with highly dynamic data, beyond simple request-response patterns, additional methods of interaction like continuous queries, queries on data streams or notifications of the incidence of certain states or situations are required. Therefore it has to be investigated, how different access patters and methods of interaction can be flexibly implemented by a uniform query interface. Domain- and application-specific functionality (e.g. computations for navigation) shall be integrated in the common query processing. As a foundation, a concept for homogenized query processing is required, which brings together different operator-based query processing techniques.

Application Support


New software engineering methodologies are required for the development of complex context-aware systems. Context-aware applications have to continuously readjust to the physical word and therefore need to be highly adaptive. Complicating adding to that sensors can only fuzzily or faulty perceive the physical world resulting from measurement errors. Therefore applications have to constantly operate on fuzzy or faulty data. Existing applications models are not designed for such assumptions. Further challenges arise if context aware applications are not merely executed isolated but rather execute integrated in already existing process oriented environments.