Multi-agent Cooperation and Stability - Belief-based Stability in Coalition Formation with Uncertainty - page 2

that are able to achieve some goals with only minimal human super-
vision.
There may be multiple agents residing in a certain environment.
This means that the agents must have a model of the world that they
are situated in. For instance, a sensor network can be described as
a multi-agent system, where each sensor should have a model of the
sensing area and its other neighbour sensors.
In many applications, the computation in a multi-agent system
is typically distributed in nature as there are no central-controlling
agents. To accomplish their tasks in such systems, each agent needs
to independently sense and perceive the environment where it is situ-
ated in, update its internal state accordingly (represented as its beliefs
regarding the environment), and then act on the environment accord-
ing to its perception and internal state so that its long-term goals can
be achieved. Note that these beliefs are private to respective agents.
Generally, one agent cannot know about the beliefs of another agent
unless they communicate with each other.
Agents are not omniscient. They cannot be assumed to know about
every environmental variable. Rather, their knowledge (or more accu-
rately, beliefs) regarding the world depends on their past perceptions,
and how those perceptions are interpreted. In any case, both the per-
ceptions and the beliefs are limited in scope and they are fallible, as
there can be inaccuracies in the perceptions, or even in the agent's
deduction logics.
Intelligent agents have social ability . They are able to communicate
and cooperate with both humans and other agents in order to achieve
tasks that cannot be done alone. Traditionally, there are two major
categories of agent cooperation. The first one is the purely cooperative
systems, in which two or more agents work as a team and coordinate
their activity in order to achieve some common goals. In such systems,
the global outcome is the only concern.
The other category is the semi-competitive systems, in which the
individual agents cooperate only if it is beneficial for them to do so.