Distributed Intelligence - Advanced Artificial Intelligence - page 524

524 Advanced Artificial Intelligence
14.5 Coordination and Cooperation

14.5.1 Introduction

In computer sciences, one of the most challenge goals is to build a computer
system which can cooperate with each other (Kraus,1997). Computer system
becomes more and more complex. To integrate the intelligent agents is more
challenge. The cooperation between agents is the key point for cooperative work.
Besides, the multi-agent cooperation is an important concept for distinguishing
multi-agent system with other distributed computing, object-oriented system and
expert system(Doran, 1997). Coordination and cooperation is the core problem of
multi-agent research since based the intelligent agent. Coordinating the
knowledge, desire, intention, plan and action to implement the cooperation is the
main goal of multi-agent system.
Coordination is a kind of character that multi-agents interact to finish a group
of activities. Coordination is the adaptation to the environment. Coordination is
to change the intention of agents. The reason for multi agent to coordinate is that
the agents have intention. Cooperation is special example of multi-agent
coordination. Multi-agent research is based on the human being society. In
human being society, the communication is indispensable. The communication is
between the conflict and non-conflict. At the same time, in open, dynamic multi-
agent environment, agents with different goals and resources must coordinate
with each other. For example, there will generate dead lock when resource
conflict without coordination. In other side, if one agent can not finish the goal
independently, it needs the cooperation with other agents.
In multi-agent system, cooperation can not only improve the ability of single
agent but the performance of multi-agent system and increase the agent's ability
to solve the problem which makes the system more agile. Through the
cooperation, multi-agents can solve more real problems and extend application.
Although for the single agent, it only cares its own requirement and goal. Its
design and implementation can be independent to other agents. However, in the
multi-agent system, agents are not existent alone. The agents' actions must meet
the society rules although they are complied agents. The relationship between the
multi agent systems makes the interaction and cooperation has great

conditionality to agent's implementation and design.
In current phase, the research on multi agent cooperation has two species. The
one is borrowing the approach from the research on multi entities' behaviors
(Kraus, 1997).The other is from the view of intention, goal and plan to research