D8 - The AmbieSense Multi-Agent Framework, Architecture Report

Till Christopher Lech (Ed.), Marius Mikalsen, Leendert Wienhofen, Robert 
Engels, Stuart Watt, Ralf Bierig, Murat Yakici, Hans Inge Myrhaug, Anders 
Kofod-Pedersen, Frank Sembowski, Martí Alemany

This document describes a scalable, reusable infrastructure for the 
context-aware delivery of content - the AmbieSense Multi-Agent System (A-
MAS). The purpose of this document is to communicate both the background 
and motivation for using agent-based technology in AmbieSense, and to 
provide guidelines for the further design and implementation of the A-MAS. 
The implemented multi-agent system will be provided and documented in 
AmbieSense Deliverable 9, "Intelligent, personalised agents for mobile use" 
(to be published in April 2004).  
 
There are many views and definitions of intelligent agents. However, in 
AmbieSense, we take two particular perspectives:  
 
	a) The lower-level perspective - multi-agent systems as platform 
	   for protocol and communication 
	b) The metaphor of agents and human-like behaviour. 
 
With regards to a), the architecture uses JADE/LEAP as the agent 
development environment of choice for AmbieSense, which will provide a 
proven, well-documented platform for the intelligent agents in AmbieSense. 
Concerning b), the high level, intelligent functionality is provided by a 
number of enabling state-of-the-art technologies such as Information 
Retrieval mechanisms, Case-Based reasoning, personalisation modules and the 
AmbieSense content base components. The main objective for the design of 
the A-MAS was to use a multi-agent system as a platform for the combination 
of enabling state-of the-art technologies in order to satisfy the needs of 
mobile users in a distributed setting. 
 
Agent properties such as autonomy, reactivity, goal-orientedness and 
communicativeness, together with the fact that agents are well-suited for 
distributed systems make the agent programming paradigm a natural choice 
for involvement in certain tasks in AmbieSense. Among these agent-supported 
tasks are detection and maintenance of user contexts as well as the 
retrieval of content, where intelligent agents have proven to be useful in 
a number of systems. The AmbieSense agents can in principle be deployed and 
run on all three main components of any AmbieSense solution in the future: 
 
	* The wireless Context Tag 
	* The mobile computers 
	* The content service provision platform 
 
Thus, by designing a scalable agent system that is possible to use in 
distributed system settings, the A-MAS will have a value of its own, making 
it possible to exploit the multi-agent system in other context-aware 
systems after the project's end. 
 
In addition to being scalable and reusable, the A-MAS shall contribute to 
agent standardisation by reusing the JADE/LEAP platform, the latter of 
which was developed in the EU IST project IST-1999-1021. The process of 
selecting this framework, along with a survey of currently available agent 
development platforms is also described in this document.