Information Technology Reference
In-Depth Information
"#
"
&'
*
!
(
)(
$%
#$%
*
Fig. 2.
UML diagrams of IFormula and ITerm
constructs supported within the language in which operators implementing ITerm are
used to represent the arguments of predicate formulae and comparisons.
Agent Factory provides support mechanisms designed to enable the easy creation
and conversion of these constructs through two interfaces:
ILogicFactory
and
ILogicP-
resenter
respectively. Default implementations are provided for both interfaces which
can process any of the logical operators shown in Fig. 2, assuming the input consists of
well formed formulae only.
Finally, support for reasoning with logic is provided via multiple logic reasoning
engines allowing queries over sets of well formed formulae or the generation of be-
liefs based on these formulae and a set of inference rules. Both of these systems im-
plement the the
IQueryEngine
interface and where neither is appropriate, additional
reasoning engines can be built. The IQueryEngine interface is designed to support mul-
tiple sources, such as the belief and goal bases of an agent; when a query is run the
IQueryEngine determines the applicable sources to check based on the underlying type
of logic object. Sources are specified by implementing the
IQueryable
interface.
In summary, the logic framework of Agent Factory implements the standard func-
tionality one would expect to see in a basic logic system and provides clearly defined
extension points that allow the logic to be modified for a specific language. However,
in many cases, we expect that such modifications will focus on specific formulae that
are based on the standard AFAPL2 logic syntax and will not require modification of
the reasoning engines or the re-implementation of the default logic engine and logic
presenter. In cases where languages do not employ our syntax, we have preferred to
adapt the language syntax rather that undertake more time-consuming modifications to
the framework.
Environment Interface:
The interface between an agent and its environment is based
around two core components:
sensors
and
actions
. Generally speaking, sensors are the
components that are responsible for generating the agents model of its environment,
while actions are the components that cause some change to occur in the environment.
As such, the core focus of a sensor is belief generation and the core focus of an action
is to facilitate manipulation of the environment. The term action differs from other
systems where it is usually referred as actuator, this difference is due to the desire to
differentiate actions from perceptors, which are the equivalent concepts in the AFAPL2
logic system.
