Information Technology Reference
In-Depth Information
«event»
Update desires
«event»
Start
«effector»
Update desires
«event»
Update
intentions
«effector»
Update intentions
«ev
ent»
E_1
«effector»
e_1
«ev
ent»
E_k
«effector»
e_k
«attribute»
knowledge base
«event»
E_s1
«sensor»
s_1
«event»
Update beliefs
«effector»
Update beliefs
«event»
Update beliefs
«event»
E_sn
«sensor»
s_n
Figure 6.11:
Mapping of a deliberative (BDI-inspired) agent architecture
onto the GRAMS reference model.
The deliberation process of a BDI-agent usually consists of multiple
steps [141, p. 58]: The belief revision function takes the current
sensor input as well as beliefs and updates the agents' set of beliefs.
The option generation function determines available options (desires),
based on beliefs and desires. The set of current intentions is updated
and finally an action is selected which is performed next.
Modeling deliberative agents within the frame of the GRAMS ref-
erence model is not unlike representing reactive agents as described
previously. Following the
Sense-Reason-Act
-cycle, Figure 6.11 out-
lines the basic mapping of a deliberative agent architecture onto the
GRAMS reference model.
The main difference to reactive agents is the explicit modeling of an
agents internal knowledge. The
knowledge base
depicted in Figure 6.11
simply refers to a set of attributes
b
i
which are associated with this
kind of agent. With respect to Section 6.6.1, the knowledge base is
defined as follows:
knowledge base
=(
b
1
,...,b
n
)
⊆
ATT
(6.11)
