This dissertation provides a systematic study of critical dependability challenges in the context of ubiquitous computing, with particular consideration of the needs of the end-user, which to this point in time has not been investigated deeply in the research community. In doing so, we focus on two fundamental areas of ubiquitous computing: human-computer interaction and context-aware computing.
Dependability in the area of human-computer interaction so far mainly focused on the design and implementation of user interfaces. However, with the emergence of highly interactive ubiquitous computing environments, the user's dependence on interface devices has been increasing significantly even in everyday life situations. This leads to a new, user-centric type of dependability threat: given that a user requires certain user interface devices to interact with or to control a surrounding ubiquitous computing environment on a daily basis, it has to be investigated how he or she can cope with the unanticipated unavailability of these devices. It is therefore important to ensure the physical availability of user interface devices in order to preserve the accessibility of personal data and device functionality required as part of the human-computer dialogue.
A further fundamental challenge is dependable context- and location-aware computing, as the manifold mobile or portable devices and smart objects found in ubiquitous computing environments are particularly subject to frequent and spontaneous changes in their surrounding context. Mobile devices not only have to cope with dynamic changes in locally available resources and in the physical properties of the user's environment, but they often also have to take into account unanticipated alterations of user-specific parameters such as the user's location, activity, personal preferences, and intentions. By enabling mobile devices to retrieve context autonomously in the respective location where it matters and using it as implicit input, context-aware computing provides means for more user-friendly, unobtrusive computing services. However, from the viewpoint of dependability, a crucial challenge is to provide mobile devices with reliable means of context awareness, and -- in addition -- to enable applications to use redundant context for the implementation of fault-tolerance mechanisms. Here an important task is to enable mobile devices and applications to tolerate and adapt at runtime to disturbances that are liable to occur in highly dynamic ubiquitous computing environments, such as the temporary or permanent unavailability of resources, the absence of network connectivity, or the unavailability of remote infrastructure-based services.
The main contributions of this dissertation are threefold. Firstly, we present concepts that increase the accessibility of personalized device functionality and improve the availability of personal user data. For that, we developed and prototypically implemented a system that exploits the diversity of user interface devices by means of a redundant input/output diversification. We further motivate the concept of instant personalization and temporary ownership as a method of freeing the user from the dependence on individual personalized devices while preserving the advantages of customized device functionality and access to personal user data.
Secondly, we describe concepts for improving the dependability of user-centric systems in ubiquitous computing environments by exploiting redundancy and diversity of resources, with a focus on location awareness. For that we have investigated and developed two concepts that enable fault-tolerant computing based on localized cooperation and resource sharing: (1) fault-tolerant services based on cooperating smart everyday objects, and (2) super-distribution of smart entities. The first approach enables mobile devices to exploit the multitude and diversity of volatile resources, which are found in the local computing environment of these devices, for the realization of fault-tolerant services. The idea of the second approach is to distribute computerized entities in a highly redundant way over object surfaces and physical environments, and to use the resulting physical infrastructure as a substrate for the realization of self-contained, fault-tolerant location-aware services and applications. As a further elaboration of the concept of redundancy, we present two systems that exploit diversity of sensor technologies and sensing capabilities by applying sensor data fusion techniques: (1) a lightweight and extensible system for the self-positioning of mobile devices based on an open sensor-fusion architecture, and (2) a solar-cell based positioning system that uses locally available context knowledge for real-time self-calibration and service optimization to improve the quality and robustness of the positioning procedure.
Last but not least, we identify and discuss dependability challenges of human-centered ubiquitous computing systems from a broader social and ethical perspective, including general aspects of reliability, control, social compatibility, and user acceptance.
Jürgen Bohn
HCI RFID distributed computing fault tolerance mobile computing positioning smart objects