Environmental Engineering Reference
In-Depth Information
the agent's environment or by other agents. When detected, the agent
responds immediately with a predetermined action.
Deliberative:
This type of behavior is perhaps the most dicult and inter-
esting. At the highest level of abstraction, this type of behavior involves
the establishing of a hierarchy of goals and subgoals, the development of
plans to achieve the subgoals, and the execution of the planned steps to
ultimately accomplish the goal that started the process of deliberation in
the first place.
4.3.2 Architecture Components
Components
A component in the agent architecture is a software module that performs
a defined task. Components when combined with other software components
can constitute a more robust piece of software that is easily maintained and
upgraded. Each component in the architecture can communicate information
to/from all other components as needed through various mechanisms includ-
ing a publish-and-subscribe communication mechanism, message passing, or
a request for immediate data.
Components may be implemented with a degree of intelligence through
the addition of reasoning and learning functions. Each component needs to
implement certain interfaces and contain certain properties. Components must
implement functionality to publish information, subscribe to information, and
accept queries for information from other components or external resources
being used by the component. Components need to keep track of their state
and to know what types of information they contain and what they need from
external components and objects.
The following describes the components in the ACT agent architecture.
Modeler
The modeling component was responsible for maintaining the domain model
of an agent, which included models of the environment, other agents in the
community, and the agent itself. The Modeler received data from the Percep-
tors and agent communication component. These data were used to update
state information in its model. If the data caused a change to a state vari-
able, the Modeler then published this information to other components in the
agent that subscribed to updates to that state variable. The Modeler was also
responsible for reasoning with the models to act proactively and reactively
with the environment and events that affected the model's state.
The modeler could also handle what-if questions. These questions would
primarily originate from the planning and scheduling component, but could
also come from other agents or from a person who wanted to know what the
agent would do in a given situation or how a change to its environment would
effect the values in its model.
