among multiple agents means managing their activities. There have been many efforts and
opinions regarding how intelligent software agents should organize themselves
into teams. The joint intentions theory
was proposed in 17, 18. The theory describes
agents which desire to be a part of team need to coordinate their local
intention to the team agents. It focuses on formation of a team when each agent
has committed to some goal and has received a coordination as well from other
According to 3, 4,
an agent needs to perform a set of basic actions to realize its goal in a
planning system. This approach defines an action which is modeled using a set
of preconditions and post-condition. If the state of the world fulfils its
precondition, an action may be executed, and the changes of state is expressed
by post-condition. An agent planner must find an order of actions such that
each action in this order must be executed and that the state of the system
obtained satisfies the goals.
can be realized by indirect or direct interaction among agents. Indirect interaction has proven
to be suitable for coordination because it decouples the identity of the
interaction actors as well as the space of interaction and the time
They have also figured out the need for a clear concept of handling concerns
apart from the core agent functionalities.
coordination has been a dominant subject in agents-based software, and
precisely in distributed artificial intelligence. The coordination can be
viewed from two major point of views, negotiation and cooperation. The
negotiation is capable of achieving accords in the agents, when each one
preserves its own interest, leading to a situation that profits all 2. On the other hand, cooperation is one kind of
coordination which consists in numerous software agents with a non-antagonist communications
(when an agent is performing actions, it does not harm the other agents) in
order to accomplish shared goals.
Our proposed research
will be merging concepts from the field of normative multi-agent systems, where
human social concepts such as norms will be explicitly represented and reasoned
about by agents, and multi-agent probabilistic planning, which seeks to find optimal
strategies for choosing an action in any given state.