Information Technology Reference
In-Depth Information
The interaction between agents has two kind of relationship: negative and
positive relationship. Negative relationship causes the conflict. The resolution of
conflicts constructs coordination. Positive relationship represent the plan of
multi-agent has overlap or some agent has the excluded ability. Every agent can
get help through the cooperation.
Before mid of 1980s, research on coordination and cooperation in distributed
intelligence mainly focus on the help each other to realize the goal without
conflicts. This kind of research is used in distributed problem solution. In the mid
of 1980s, Rosenschein did a deep research in his doctor thesis on multi agent
interaction when agents had the goals conflicts. He applied the game theory to
build the static model for rational agent interaction (Rosenschein,1986).It then
become the theory foundation for the multi agent coordination and cooperation.
After then, many researchers used the game theory to formalize the multi agent
cooperation. All of these researches are to build model to coordinate the actions
or through the cooperation to realize the goal with the inconsistent goals
condition. Some of research considered time preference. Some of research is
open environment oriented.
MIT's research adopted the meta-communication to coordinate multi agent
computing under the foundation of intensifying FA/C and PGP approaches.
MacIntosh applied the heuristic approach to introduce the cooperation to machine
theorem proof. Sycara researched the negotiation based the labor and capital
problem (Sycara, 1996). The approach applied the heuristic and constraint
satisfaction technology to solve the distributed search problems. It adopted the
asynchrony trace to resume the inconsistent search policies. The disadvantage is
this approach need an arbitrage machine to solve the conflicts. Conry research
multi steps negotiation under the multi goal and resources. Hewitt proposed the
distributed artificial intelligence approach which challenges Rosenschein's static
interaction model. The real world is open and dynamic. The coordination and
cooperation is also open and dynamic. Computing bionomics argue that the
agents need not have the strong reasoning ability in the open and dynamic
environment. They can gradually coordinate the relationship with the
environment through the consecutive interaction to make the whole system
having the evolution ability. That is similar to ecosystem. In BDI model, it
emphasizes the agent intention, desire and intention's rational balance in
interaction process.
Shohan proposed the artificial agent society need a law to regulate the agents'
actions. Every agent must follow the law and believe that other agents also abide
