Page 134 -

AGENT-ORIENTED METHODS AND METHOD ENGINEERING 119
AGENT-ORIENTED METHODOLOGIES

A Genealogy of Agent-Oriented Methodologies

The development of AO methodologies has taken many routes. Some methodologists have based
their methodological approach on Artificial Intelligence (AI) or Knowledge Representation ideas;
others have commenced with basic definitions of objects and then asked what modifications are
necessary to support agents; others have commenced with an established OO methodology and
asked how agent support can be grafted on.
Figure 8.1 graphically depicts some of these linkages and influences. OO methodologies such
as the Rational Unified Process (RUP) (Kruchten, 1999), the Object Modeling Technique (OMT)
(Rumbaugh et al., 1991), and Fusion (Coleman et al., 1994) have all been used by various AO
methodology groups as the basis for agent-oriented extensions. RUP has formed the basis for
Adelfe (Bernon et al., 2002) and also for MESSAGE (Caire et al., 2001), which, in turn, is the
basis for INGENIAS (Pavón, Gomez-Sanz, and Fuentes, 2005), and, more recently, RUP has been
a useful input to the Radical Agent-oriented Process (RAP) (Wagner and Taveter, 2005), a direct
descendant of Agent-Object-Relationship (AOR) (Wagner, 2003). OMT is said to have directly
influenced MAS-CommonKADS (Iglesias et al., 1996, 1998), which merges these OO ideas with
concepts from AI and Knowledge Engineering, as well as the AAII approach (Kinny, Georgeff,
and Rao, 1996), which, in turn, is said to have been a major influence on MaSE (DeLoach, 1999;
Wood and DeLoach, 2000). Fusion has strongly influenced Gaia, which, in turn, has influenced
SODA (Omicini, 2000). ROADMAP (Juan, Pearce, and Sterling, 2002) is an extension of Gaian
ideas. Prometheus (Padgham and Winikoff, 2002a, 2002b) is a fully AO methodology but states
that one should use UML-style diagrams when appropriate rather than “reinvent the wheel.” All of
these AO methodologies are “standalone”—effectively “one size fits all”—approaches, although
some mergers are currently being initiated (e.g., between ROADMAP and Prometheus). Care
must, however, be taken when merging methodologies, since this can lead to an excessively large
(sometimes implicit) and partially redundant metamodel.
Other methodologies in Figure 8.1 do not acknowledge any influence from any OO approach—
although clearly some have had an implicit influence. Tropos is said to be based on i* (Yu, 1995)
and has a distinct strength in early requirements modeling. Its use of the i* modeling language gives
it a different look and feel from those that use Agent UML (AUML); (Odell, Van Dyke Parunak,
and Bauer, 2000) as a notation. It also means that the non-OO mindset permits users of Tropos to
take a unique approach to the modeling of agents in the methodological context.
There is no obvious, explicit evidence of an OO influence in the published versions of Cassiopeia
(Collinot and Drogoul, 1998; Collinot, Drogoul, and Benhamou, 1996), the earlier versions of
2
PASSI (Burrafato and Cossentino, 2002; Cossentino and Potts, 2002), and the work of Kendall,
Malkoun, and Jiang (1996). CAMLE (Shan and Zhu, 2004) does, however, draw some parallels,
particularly between a CAMLE caste and an OO class and with respect to UML’s composition
and aggregation relationships.

Synopsis of Specific Agent-Oriented Methodologies

As noted above, many individualistic methodologies have been formulated and published. Here,
we briefly review a small selection. Each description below emphasizes the agent-oriented aspects
of that methodology, which are needed to go beyond the basic object-oriented concepts that many
of them utilize.