An Agent Framework for Ubiquitous Services

   page       BibTeX_logo.png   
Conor Muldoon
School of Computer Science and Informatics
University College Dublin
Dublin, Ireland
February 2008

This thesis introduces a framework for the rapid prototyping of ambient applications. A solution is proposed that draws upon distributed artificial intelligence and pervasive com- puting to create autonomous self-managing context-sensitive applications. The framework comprises a common core of agent-based functionality that software engineers use for the rapid prototyping of ubiquitous applications.

Intelligent agents encapsulate a number of features that make them an attractive and viable option for realizing ambient applications. Their autonomous nature, ability to react to external events, as well as an inherent capability to be proactive in fulfilling their ob jectives make them apposite for operating in complex and dynamic environments. The ability to migrate further enhances an agent?s capabilities to adapt and respond to unexpected events. This flexibility makes agents particularly adept at managing the complexity of emergent behaviour synonymous with ambient systems.

The thesis details a minimised footprint platform for multi-agent systems, an extensible infrastructure for context-sensitive graphical user interfaces, and a service language that facilitates agent-based service interaction and outsourcing. Two demonstrator applications are described along with some development support tools.

When developing agents for embedded devices, it is desirable that effective use be made of the limited computational infrastructure available. A rational decision making infras- tructure is proposed that enables agents to reason about their finite resources. Resource management is viewed from both an individual and a group perspective. The framework is evaluated through a number of experiments.

The thesis includes a detailed literary survey.