Geoscience Reference
In-Depth Information
3- Multi agent systems and generalization
Multi agent system (MAS) is a concept of artificial intelligence. The sys-
tem contains agents which represent different objects. One important char-
acteristic of agent is its autonomy. An agent can choose its own behavior
and cannot be imposed its action by other agents or by the environment.
The first type of agent is reactive agent, with limited intelligence. It acts in
reaction to some stimulation from its environment. The intelligent agent
model was then introduced. This kind of agent can remember the action
history and compare which combination of actions is better. The intelligent
agent model includes perception, control algorithm, state and actions parts
(Schumacher 2001).
Schumacher (2001) gives a model of intelligent agent in
figure 2
. Percep-
tion part gives the surrounding environment conditions to agent. State part
stores the history of actions to evaluate the situation of agent. Action part
is a list of actions including all acceptable behaviors. The intelligent agent
can make action plans and evaluates the best one for action.
Figure 2:
Architecture of intelligent agent (Schumacher 2001)
Agents were introduced in map generalization by Baeijs et al. (1996) with
the system SIGMA (SemI automated Generalization using Multi-Agent
System). It designs a group of scopes to control agent actions. It used reactive
agent. The agents react when the conditions are triggered. It does not
contain a state model to evaluate the result of its action. In a problem such
as line generalization, line agents would act when some constraints are
