A system architecture for context-aware mobile computing

TLDR: This thesis explores how to support and also exploit the dynamic configurations and social settings characteristic of mobile systems and advances the following goals: enabling seamless interaction across devices; creating physical spaces that are responsive to users; and and building applications that are aware of the context of their use.

Abstract: Computer applications traditionally expect a static execution environment. However, this precondition is generally not possible for mobile systems, where the world around an application is constantly changing. This thesis explores how to support and also exploit the dynamic configurations and social settings characteristic of mobile systems. More specifically, it advances the following goals: (1) enabling seamless interaction across devices; (2) creating physical spaces that are responsive to users; and (3) and building applications that are aware of the context of their use. Examples of these goals are: continuing in your office a program started at home; using a PDA to control someone else's windowing UI; automatically canceling phone forwarding upon return to your office; having an airport overhead-display highlight the flight information viewers are likely to be interested in; easily locating and using the nearest printer or fax machine; and automatically turning off a PDA's audible e-mail notification when in a meeting. The contribution of this thesis is an architecture to support context-aware computing; that is, application adaptation triggered by such things as the location of use, the collection of nearby people, the presence of accessible devices and other kinds of objects, as well as changes to all these things over time. Three key issues are addressed: (1) the information needs of applications, (2) where applications get various pieces of information and (3) how information can be efficiently distributed. A dynamic environment communication model is introduced as a general mechanism for quickly and efficiently learning about changes occurring in the environment in a fault tolerant manner. For purposes of scalability, multiple dynamic environment servers store user, device, and, for each geographic region, context information. In order to efficiently disseminate information from these components to applications, a dynamic collection of multicast groups is employed. The thesis also describes a demonstration system based on the Xerox PARCTAB, a wireless palmtop computer.