Ambient Intelligence Through Agile Agents - Ambient Intelligence for Scientific Discovery - page 291

oriented applications, there is a need to support the federation of AMS and DF
services. This federated service is realized by a third System Agent, entitled
the Super DF. The Super DF provides a common point of interaction between
the various AMS and DF agents. Within the configuration of individual APs,
developers are able to specify the agent identifier of a Super DF agent that they
would like the AMS and DF agents to connect to. Upon start-up, this agent
identifier is passed to the AMS and DF agents who register themselves with
the specified Super DF. Once registered, the AMS and DF agents are able to
perform global (at least within the set of APs that are connected to the Super
DF) white and yellow pages searches. Ongoing research is concerned with the
development of a more robust infrastructure that exhibits autonomic properties
such as self-organization and self-healing, and which contains no single point of
failure.

2.3 The Development Methodology

Methodological support for the fabrication of agent-oriented applications using
Agent Factory is provided through a UML-based software engineering process
that supports the design, implementation, testing, and deployment phases of the
software engineering lifecycle. A diagrammatical overview of this process, details
of which can be found in [5], is presented in Fig. 1. In this diagram, it can be
seen that the design stage of the process is focused around the development of
five models:

- The System Behavior Model (SBM) identifies the main roles that agents will
play within the system, and associates those roles with the key system behav-
iors. Visually, this model is formalized using a customized UML Use Case
Diagrams where actors are stereotyped as roles, and use-cases are stereo-
typed as system behaviors. Some typical system behaviors that might apply
to a mobile computing system include: user movement updates, hotspot ac-
tivation, and map generation.
- The Interaction Model (IM) expands on the SBM through the modeling of
the interactions that occur within each of the system behaviors. Visually,
this model is formalized using a customized UML Collaboration Diagram.
Specifically, objects are stereotyped as roles, and messages are restricted to
valid FIPA ACL performatives via the “fipa-acl” stereotype.
- The Activity Model (ActM) complements the IM, in that it expands on the
SBM through the modeling of the set of activities that occur within each of
the system behaviors. Visually, this model is formalized using a customized
UML Activity Diagram where swimlanes are employed to represent roles.
- The Protocol Model (PM) represents a demarcation point within the method-
ology, in that, it represents the transition point where the focus turns from
understanding system behaviors to the formalization of those behaviors as
a set of protocols and agent-classes. Specifically, the PM is derived from the
IM with agent interactions formalized as protocols. Visually, these protocols
are represented using Agent UML Sequence Diagrams.