Biomedical Engineering Reference
In-Depth Information
the higher intensity) are going to trigger energy
propagation in the cognitive structure as in the
motivational graph discussed previously. This
procedure has complexity O(n) on the number
n of Motivations Agents, in the worst case. So,
up to this point of the action selection process,
reactive agents have not executed too complex
processing, which is satisfactory from the point
of view of computational requirements. The next
step is cognitive processing, which usually is the
more complex of any intelligent action selection
mechanism. However, this previous competition
is able to prune significantly this processing. Next,
Cognition Agents and the cognitive negotiation
performed by them are described.
may be performed to achieve the intended goal.
Such knowledge can be either encapsulated in
a single Cognition Agent, or distributed among
many of them. This cognitive structure is similar
to Minsky's society of agents, and to the motiva-
tional graph, both already previously discussed
(see above). A hierarchical tree of Cognition
agents is built, its root being the agent activated
by the motivation in question, and its leafs being
Cognition Agents that trigger specific Execution
Agents. Different trees may share Cognition
agents at any level, except the root. When the
Motivation Agent activates the root, the intensity
value i is passed to the Cognition Agent intensity.
This intensity propagation continues in the cogni-
tive tree, in a way similar to energy propagation
in motivational graphs, allowing the negotiation
mentioned previously. When intensities coming
from more than one root converge in a shared
node, the chance of selecting a leaf satisfying more
than one motivation increases. As a consequence,
a higher amount of discharging probably occurs
in the Inner Motivational Environment.
Active leafs pass their intensity to specific
Execution Agents and to the roots that have con-
tributed to their activation. These roots, in turn,
send this value ( d ) to their respective Motivation
Agent.
At this point, if an active Motivation Agent
in discharge zone does not pass at least to toler-
ance zone, its root Cognition Agent triggers the
Discharge Procedure according to the context.
Next, a simple example is presented for the
reader to feel a flavor of the architecture possible
applications.
Cognition Agents
A Cognition Agent integrates perception and
decision-making in a same set of rules. Percep-
tion searches and analyses only environment
information that is relevant to the agent task. Once
the needed information is obtained, a decision-
making process takes place and selects which
agent to pass the intensity value. If it is possible
that relevant environment elements could not
be immediately available, additional knowledge
and a decision-making process are necessary to
choose where to look for them.
Alert Cognition Agents are special agents re-
sponsible only for tracking specific stimuli, objects
or events in the environment, and evaluating them
in terms of distance and/or intensity. Once this
evaluation reaches predefined thresholds, its value
is sent to a specific Motivation Agent, increasing
its intensity value, as already described earlier.
The main purpose of the whole action selection
mechanism is to reduce the total intensity of the
Inner Motivational Environment by acting in the
Outer Environment. When each Motivation Agent
is designed, its corresponding way of satisfaction
has also to be defined. This involves determining
what environment elements are possibly relevant
and must be searched, and what alternative actions
A Simple Example: Threatening
Situations and Need for Safety
Threatening conditions are frequently included
in virtual storytelling for creating thrill, but also
in some simulations where human reactions to
danger are to be evaluated. That is the case of
some building analyses for safety conditions in
Search WWH ::




Custom Search