In this chapter two original and influential visions of the future relationship between computers and people have been discussed, namely Weiser’s Ubiquitous Computing vision [Weiser,91] and Norman’s concept of information appliances and invisible computing [Norman,98]. To illustrate the developments since these statements the vision of the disappearing computer as formulated by the European research council has been presented [Wejchert,00]. This document also outlines which research should be carried out to reach the goals stated in the vision. Common to all statements is the observation that context: the real world around is essential to such systems.
A range of characterisations and definitions for context and context-aware systems has been surveyed and analysed. It can be observed that context-awareness and context-aware system evolved from location-awareness by generalisation. Further concepts that constitute the environment and that can be measured are included in the understanding of context. Exemplarily, several projects that reflect current trends in ubiquitous computing research and in the area of context-awareness have been presented. It can be observed that even if the notion of context is widened in most research groups, the systems that have actually been implemented rely mostly on location. Location as a prime context is very well understood [Leonhardt,98] and context-acquisition devices are available off-the-shelf, at least for outdoor use. Furthermore, the value of location as context is obvious. The value of other context
information, especially about the environment, is often not clear and measuring them often requires specific hardware.
The research surveyed lead to the following observations. Based on these observations issues to take on in the course of the research were identified:
• Most of the work on context acquisition is centred on location or shows an opportunistic sensor selection. In contrast in chapter 3 context acquisition and perception using a variety of sensing technologies is assessed systematically with respect to their potential use in ubiquitous computing systems.
• As seen from examples above context is not isolated, it is linked to entities. This fact is used to create a model where context acquisition and context use is related to artefacts. By structuring context in this way the model offers an effective means to reduce complexity. This bottom-up model is strongly connected to the approach of prototyping as shown in chapter 4.
• Related research suggests software architectures and libraries to ease the use of context when developing applications. However as there is no standard sensing infrastructure or common hardware most approaches excluded hardware or only include very specific solutions. To advance this matter a context acquisition platform consisting of hardware, software, and communication was developed and is presented in chapter 5. This is extended to a more general physical rapid prototyping system for ubiquitous computing. Together with the tools a methodology was developed, too.
• In the literature so far the distribution of context assumes an active management of such information, e.g. based on subscription models. In chapter 6 an alternative approach is suggested where context is around and available depending only based on spatial and temporal relationships.
• The very basic idea of invisible computing, interaction with a system that is not regarded as operating a computer, questions many concepts in traditional human computer interaction. An alternative implicit human computer
interaction model is presented in chapter 7. Also the notion of invisibility is investigated further.
• The research methodology that can be observed in the work reviewed in this chapter varies to a great extent. In many cases the reported evaluation is opportunistic. In chapter 8 a systematic overview of evaluation techniques and their applicability is provided.
Acquiring Context using Sensors
In this chapter mechanisms to gain context information by the use of sensors are introduced. The reasoning behind this is that context-aware applications rely on the availability of information about the situation they are used in. The ultimate goal is to make available to the system, a representation of the world around that is close to the perception of the user. In this chapter it is assessed which steps are needed to provide basic perception for context-aware systems. This approach looks systematically on how to narrow the gap between the user’s and the system’s perception of the real world in certain situations.
As outlined earlier context sensing received little attention in ubiquitous computing research so far. In many cases context is solely based on location. For location different means of sensing and interpretation are well established. However the physical world offers a much richer environment. With respect to situation and context the contribution of further sensors, especially monitoring the physical world, is little understood.
Sensing and sensor technologies are widely used in robotics, automation, and engineering. In these environments sensors proofed to be an essential source of information to create useful systems. In general in such systems the task is well defined and specific requirements (e.g. regarding accuracy and update rate) can be determined. In contrast in Ubiquitous Computing environments sensors are often used
to monitor unstructured phenomena and new methods and techniques are required [Estrin,02].
In the following the concepts of cognition, perception, recognition, and abstraction are discussed based on the hypothesis, that context is related to parameters that can be observed when a certain situation occurs. Starting from these observations various types of sensors and algorithms are introduced and assessed for their utility for building context-aware systems. In particular sensing technologies and perception algorithms are evaluated with respect to the requirements implied by Ubiquitous Computing environments. Furthermore a layered recognition architecture is introduced that offers interfaces on various levels and supports distribution.