Research‎ > ‎

Challenges

Today's Information Access

With the advent of powerful Internet-connected objects such as smart phones, personal digital assistants and tablet computers, an ever increasing number of European citizens have constant access to the wealth of information stored on the millions of servers connected via the Internet. At the present time, the availability of such connected objects is causing a drastic paradigm shift in the way people deal with information. Instead of collecting – and printing – potentially relevant information in advance using a personal computer that is only available at particular locations, they now access more and more information on-demand and on-the-go.
 
Yet, despite this significant change in behavior, the technical means to access information have only changed marginally. In most cases, information is accessed via the web which requires users to memorize long URLs, click through sequences of web pages or browse through irrelevant search results. Alternatively, if they are frequently accessing the same service, they may install an app or application that provides more convenient access. However, such an installation requires advance planning and does not provide suitable support for services that are primarily useful in a particular environment. Moreover, even if they are using a local proxy, the utilization of a more complex service, for example, to book a train ticket, requires users to specify numerous inputs such as destination, time, etc. using miniaturized and often, inadequate peripherals. As a consequence, the state-of-the-art puts a natural limit on the complexity of the software and thus, on the level of support, that can be gained from existing services.
 
 
On-demand access through mobile browser or app
 

The Ubiquitous Computing Vision

In contrast to this, ubiquitous computing envisions services which provide seamless and distraction-free support for simple and complex everyday tasks of their users. In order to realize this vision, the set of services that is available and the services themselves must be adapted to the user’s situation, behavior and to varying user intents. Thereby, this adaptation must be performed autonomously in order to ensure that it does not conflict with the goal of providing a distraction-free user experience. This, in turn, requires services to gather a broad range of characteristics of the user's context at runtime. Examples for these characteristics include the user's location, activity, plans and goals.
 
 
Internet-connected objects such as smart mobile phones and personal digital assistants provide a promising basis for determining user context in an automated manner on a large scale. The reasons for this are manifold. First and foremost, many Internet-connected objects are self-contained and do not require additional infrastructure support, but existing cellular and wireless local area networks can provide the backbone for object-to-object interaction if needed. Secondly, though these objects are often resource-constrained, newer generations are designed to support more complex tasks such as displaying a high-resolution movie. As a consequence, the objects are often not utilized to their fullest capacity, leaving enough resources to perform context recognition. Thirdly, with a variety of on-board sensor, many Internet-connected objects have access to both physical and virtual data sources which allows multi-modal context recognition with high precision. Lastly, since the objects are often carried by and owned by a single user continuously, the object’s context is tightly correlated to the user's context and the recognition alone does not invade privacy.

Beyond Isolated Scenarios

In the past, these characteristics have contributed to the development of a number of context recognition systems for Internet-connected objects. The recognition methods applied by existing systems are usually fine-tuned for specific requirements in order to provide reasonably accurate results while requiring limited resources. Although, these methods are suitable for accurately detecting desired characteristics, they cover only a narrow set that can be detected by one object alone. Moreover, due to the resource-constrained nature of many Internet-connected objects, developers have usually concentrated on providing solutions for a concrete service.
The vision of ubiquitous computing, however, extends beyond the boundaries of a single service as it envisions seamless support for everyday tasks. As a consequence, achieving the overall vision of ubiquitous computing raises a number of novel research challenges which include:
  • the development of concepts to support the automated recognition of a broad range of context information types to support a variety of application scenarios in a generic fashion,
  • the development of context recognition methods that are able to cope to the limited resource availability and energy-constraints of resource-poor connected objects,
  • the development of novel data acquisition and distribution protocols to share context information in order to increase the recognition accuracy without endangering privacy,
  • the definition of an interoperable data representation model for context information and associated query models to support object to object communication,
  • the design of a scalable data infrastructure to share and aggregate possibly frequently changing context information gathered by a large number of connected objects,
  • the development of tools to reduce the required amount of manual configuration of policies and the mechanisms to validate them in order to protect the privacy of users,
  • the design of new context-based human computer interaction techniques that are able to incorporate user goals and intents.

Continue reading the scientific and technical objectives.