Information Technology Reference
In-Depth Information
3.3.4.
Analysis of agents
This can start at the end of functional analysis of the service analysis phase. The
models of agents, behaviors, communication links and deployments of agents are
identified here for personalization needs. They are structured into four subphases:
- The
analysis of agent models
identifies the types of agents necessary for the
realization of functional needs. The agent models can be identified following the
rules below:
- rule 1: an agent model by the functionality expected from the service. For
example, the functionality “itinerary search” will be translated by an agent model of
“itinerary search”,
- rule 2: an agent model by type of interaction platform. For example, an agent
model for a PDA, another model for a PC, etc.,
- rule 3: an agent model by external resource with which the system must
interact. For example, an agent model for itinerary search on transport operator
servers, another for the search of restaurants on web services, etc. In this stage of
analysis of agent models, the diagrams for use and deployment cases of the analysis
phase are taken up again from the angle of personalization. Only the aspects that are
judged as being part of personalization are considered;
- The
identification of communication links
describes relations between the
different agents and their beliefs regarding the abilities of various individuals. Here,
relationships of the interactions of each agent are modeled. The modeling of
communication links can be done via UML extensions put forward by [ODE 99]
which are used in
Agent-Oriented Design Process with UML
.
-
Behavior analysis
,
for each agent to adequately attain its objective. This
analysis determines the granularity necessary for breaking down behaviors. This
will therefore depend on the intuition and know-how of the developer. The main
rule to apply consists of breaking down a behavior for as long as the task associated
with this behavior can be associated with another behavior.
-
Deployment information
describes the different physical locations where the
agents go to be executed. This description can concern information of static location
(the agent is located at a same place and never changes place) as well as information
of dynamic location (the agent can dynamically change place according to the tasks
it wishes to accomplish). The deployment information is modeled via deployment
diagrams.
