Biomedical Engineering Reference
In-Depth Information
In the model proposed by Sevin and Thalmann,
motivations result from the evaluation of internal
variables according to a threshold system and gen-
erate actions when combined with environmental
information. There is one hierarchical decision
loop per motivation, which generates sequences of
locomotion actions (motivated behavior) that lead
to locations where the motivation can be satisfied.
All decision loops run in parallel and, at the end,
the most activated action is executed. Besides loco-
motion actions, there are motivated actions, which
can satisfy one or several motivations. Both types,
when executed, alter internal variables related to
their original motivations: locomotion actions
increase them, and motivated actions decrease
them. The threshold evaluation system represents
a non-linear model of motivation changes, and
it is specific to each motivation. Depending on
two thresholds T1 and T2 and according to the
value of its internal variable i, motivation M is
evaluated in three possible regions—Comfort
Zone, Tolerance Zone and Danger Zone—by the
following equations, respectively:
does not keep trying forever, and tends to express
frustration in some emotional way. But by the
model just described, when an agent reaches the
danger zone, it can not abandon the corresponding
motivation. The other motivations will continue
to increase, and will also not be satisfied, because
the agent is stuck in that former motivation. That
is not a plausible human attitude. When someone
becomes frustrated, some expression of this feel-
ing may be seen by an observer. Besides, eventu-
ally she/he moves on, and tries to satisfy other
motivations. That problem has to be tackled by
a believable model.
Motivational Graphs architecture (Chiva,
Devade, Donnart, and Maruéjouls, 2004) is an-
other interesting approach to character motivation
structure, developed in the context of computa-
tional games. It is based on activity propagation
in a graph structure where nodes represent rule
sets, instead of representing neurons like in
connectionist systems. Each set is part of a plan
decomposition, and propagates activity according
to its internal rules. The final nodes correspond to
actions that the character can perform. Environ-
ment input activates some nodes, and after energy
propagation, the node with greater energy repre-
sents the action to be performed by the character.
This architecture is meant to take advantage of
both symbolic and connectionist approaches by
hybridizing them in a single structure.
Another important aspect of mind structure
that Franklin (1995) pointed out is the relation
between Perception and Action. Since every in-
teraction with the environment has to be guided
by some knowledge, we may add its processing
in the mutual perceiving-acting relationship. But
instead of what may be suggested by the tradi-
tional sequence input- processing- output, these
three activities are not thought of as independent
processes. On the contrary, sensory processing
is strongly determined by the current action,
and motor activities greatly depend on sensory
feedback. Previous works extending this point
(Vilela, 1998 and 2000) argue that Perception, Ac-
2
(
i
−
T
)
M
=
=
T e
if i
<
T
2
1
1
M
i
if T
≤ ≤
i
T
1
2
i
M
=
if i
<
T
2
(1
−
i
)
2
If a Motivation lies in the Comfort Zone, it is
not taken into account. Otherwise, its chance to be
satisfied increases as its internal value increases.
However, when it reaches the Danger Zone, its
competitiveness is much more intense. The pur-
pose of the action selection mechanism is to lead
motivations to the Comfort Zone. A Motivation
M
is described as a “tendency to behave in particular
ways” (Sevin & Thalmann, 2005b), and then it
may be considered as a driving force of action
selection. But this quantitative model does not
account for an important human characteristic:
the possibility of frustration. In normal situations,
human motivations are not always satisfied. And
when a person can not satisfy a motivation, she/he
Search WWH ::
Custom Search