Context-aware rule-based data distribution algorithms and methods for pervasive computing

title: Context-aware rule-based data distribution algorithms and methods for pervasive computing
author: Bart P.I. van der Poel
published in: August 2002
appeared as: Master of Science thesis
Delft University of Technology
pages 116
PDF (1.145 KB)


The research field of pervasive computing is concerned with computing environments of diverse, possibly mobile computing units connected over wireless networks. The introduction of mobility in networking poses new challenges for the middleware with regard to accessibility and usability of data. To address these challenges this research focuses on context-aware data distribution algorithms and methods. Context-awareness provides environment dependent adaptation regarding relevance, timeliness and fidelity of data to the distribution.

In this thesis algorithms are introduced for client profiling and data selection. The client-profiling algorithm combines user-defined rules with contextual information to set up the selection rules. The data selection algorithm applies these rules to incoming data. Instead of Boolean decision the data selection algorithm maps the list of interested clients to priorities. Sending data to the clients is performed in order of priority.

For the implementation of the algorithms an expert system is embedded in the data distribution agent. This makes the subscription language for the client profiling highly expressive since conditions can be defined in conjunctive, disjunctive and negative forms. The subscription language is extended to support location aware conditions.

A collaborative system, called DISCIPLE, applying the proposed context-aware data distribution algorithms and methods demonstrates how the results of the research can be used in real applications. FLATSCAPE is a military collaborative application running on top of DISCIPLE that is developed for operational planning by commanders.

Grouping users with equal profiles is applied to improve scalability. Experimental results show that the agent performs well for large number of data and users and consequently scalability of the agent is satisfactory.

blue line
University logo