Information Technology Reference
In-Depth Information
the law. This law can regulate the agents' behaviors on one side. On the other
side, it can insure other agents' action style. This guarantees the realization of
agents' behaviors. Decker proposed a dynamic coordination algorithm in
distributed sensor network (Decker, 1995). The agent can revise the sensor
domain according to the unified standard which can automatically reach the
coordination. This dynamic reconstruction domain can realize the system load
balance in the whole system. Its performance is better than static sub area
algorithm. Especially, this approach can decrease the fluctuation of performance
which is more adapted to dynamic real condition.
In the multi agent planning, agents can regulate the pertinence of goals of to
realize the autonomic behaviors coordination. German Wei
proposes approach
to form action and goal's correlation value for specific problem through the
distributed learning. The other autonomic coordination approach is Markov
process proposed by Kosoresow. Markov process is a quick probability
coordination approach when the agent's goal and preference is consistent. If
Markov process is applied as the agent's reasoning mechanism, it can analyze the
astringency of multi agent interaction and average astringency time. When the
agent's goal and preference is not consistent, it can check the inconsistence at
specific time point and solve the conflict by submitting it to higher coordination
protocol. In order to describe a group of agent actions, we need a common
intention to connect the member's behavior. Jennings adopts common duty
concept to emphasize the function of intention as the action controller. It
prescribes how the agents to action in the cooperative problem solution. This
common duty can provide function instruction for architecture design, provide
standard for inspecting problem solution and exception dispose. They realized
this common duty in GRATE system.
14.5.2
Contract net protocol
In 1980, Smith proposed the contract net protocol in distributed problem solution
(Smith, 1980). This protocol was widely applied in the multi agent coordination.
The communication among agents has the same message format. The real
contract net system provides a contract protocol based the contract net protocol.
It prescribes the task assignment and the roles of agents. Figure 14.10 shows the
architecture of node in contract net protocol.
The locale database includes the knowledge base related to the node which is
used to negotiate the information in current state and problem solution process.
