Community Computing - Advanced Design Approaches to Emerging Software Systems - page 214

that research about system engineering is
required to manage massive connections
among powerful devices in a community
computing environment (Microsoft, 2005).
Microsoft is systematically trying to sup-
port its community computing by devel-
oping various tools which enable people
to interact with one another for gathering
and exchanging services. In its research,
a community is a sort of group of devices
sharing one another's information and ca-
pacity to generate communication between
people. It seems that it concentrates on sys-
tematic support to communities.

(Cockburn, 1996) was proposed. In this
model, cooperation is represented as a
recipe or a set of tasks. The generation of
this recipe is an iterative process requiring
interaction with other agents on their own
schedules related to other projects. When a
recipe is completed, it is sent to all partici-
pating agents, and the project commences.
Once committed, each participating agent
receives the final recipe and is committed
to the relevant time interval in the recipe.
The CM (Cooperation Management) is a
component that is responsible for all tasks,
commitments, and cooperation. Among the
sub-components of CM, the PPC (Prepare
Project Commitments) component decides
a preferred set A of activities with which
goal g can be accomplished and subse-
quently generates a dependency graph of
the activities using the critical path method
and domain knowledge. Using this depen-
dency graph, PPC determines the partici-
pant agents and then sends a list of agents
to the GMR (Generate and Modify project
Recipe). The goal of a component GMR
is to design a recipe R , where the recipe
R is interactively designed by generating
and sending the proposed recipe to agents
interested in participation. A recipe R con-
sists of a task A , willing participants ca-
pable of performing that task, a priority p ,
and a deadline T for that task. The partici-
pants accept, adapt, or reject the proposed
recipe. In this model, however, the method
used to make an agent cooperate with other
members is specified, but how a recipe R it-
self is generated using a dependency-graph
of activities is not specified. In addition,
a means to describe cooperation among
agents is not represented in that model.

• Community Computing in this chapter:

To design and develop a cooperative multi-
agent system conveniently, we propose
community computing as a new computing
paradigm. Other research has not seriously
concentrated on cooperation, so it did not
provide a way to design and execute a co-
operative group, such as a community and
cooperation in that group, immediately.
Towards this goal, in our community com-
puting approach, we provide the abstrac-
tion models to design them intuitively and
develop a community computing system
quickly. Our community computing is
more useful especially when a number
of existing agents exist, and a developer
needs to design cooperation among those
agents and execute it immediately.

Existing Cooperation Models

To propose the static community situation based
cooperation model which was described in section
3.2.1, we surveyed existing cooperation models
for multi-agent systems. In this section, we briefly
introduce some works as follows:

• AGDRSCOM (Hua, 2003): It is an agent

cooperation model in which member
agents are able to adjust their own coop-
erative tasks according to the changes of

• The cooperation model for ARCHON

(Brazier, 1997): In 1997, a refined for-
mal cooperation model for ARCHON