12.12.2012 Views

Who Needs Emotions? The Brain Meets the Robot

Who Needs Emotions? The Brain Meets the Robot

Who Needs Emotions? The Brain Meets the Robot

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

motivation/emotion in behavior-based robots 261<br />

Figure 9.7. Various AIBO robots.<br />

(1974) motivational space and Blumberg’s (1994) action-selection mechanisms,<br />

particular behaviors are scheduled for execution by <strong>the</strong> robot that<br />

are consistent with <strong>the</strong> current set of environmental stimuli and <strong>the</strong> internal<br />

state of <strong>the</strong> robot itself. In <strong>the</strong> action-selection module, a behavior is selected<br />

based on inputs derived from external stimuli (releasing mechanisms, cf.<br />

Chapter 7, Ortony et al., and Chapter 8, Sloman et al.) and <strong>the</strong> robot’s current<br />

motivational state variables. A state-space diagram represents <strong>the</strong> mapping<br />

from <strong>the</strong>se inputs onto <strong>the</strong> appropriate behavior to be activated (Arkin,<br />

Fujita, Takagi, & Hasegawa, 2003).<br />

Fur<strong>the</strong>r extension of this research has resulted in <strong>the</strong> emotionally grounded<br />

(EGO) architecture (Fujita et al., 2001a,b), leading to potential applications<br />

in humanoid robots (Arkin, Fujita, Takagi, & Hasegawa, 2003). <strong>The</strong> generated<br />

motion patterns can be affected by <strong>the</strong> emotions <strong>the</strong>mselves. Fujita et al.<br />

(2001b) specifically addresses <strong>the</strong> symbol-grounding problem (Harnard,<br />

1990) in this architecture, allowing <strong>the</strong> robot to learn to associate behaviors<br />

with specific symbols through <strong>the</strong> use of an emotionally grounded symbol,<br />

where <strong>the</strong> physically grounded symbol is associated with <strong>the</strong> change of internal<br />

state that occurs when <strong>the</strong> robot applies a behavior in response to <strong>the</strong><br />

object. For example, when <strong>the</strong> robot hears <strong>the</strong> symbol’s name spoken, it<br />

knows which behavior(s) is associated with that symbol and can produce a<br />

change in its internal motivations. Thus, in a sense, <strong>the</strong> robot knows <strong>the</strong><br />

meaning of <strong>the</strong> symbol in <strong>the</strong> way in which it affects both its internal state<br />

and what behaviors are correct to use in <strong>the</strong> associated object’s presence.<br />

<strong>The</strong> symbols are grounded not only perceptually, by associating <strong>the</strong> correct<br />

perceptual stimuli with <strong>the</strong> spoken symbol, but also behaviorally, by producing<br />

<strong>the</strong> appropriate behavioral response in <strong>the</strong> presence of <strong>the</strong> stimuli<br />

that acts in a manner to produce a change in internal variables consistent<br />

with <strong>the</strong> I/E model. This use of symbols for emotional modeling diverges<br />

somewhat from strict ethology, especially when compared to <strong>the</strong> more faithful<br />

canine behavioral modeling employed (Arkin, Fujita, Takagi, & Hasegawa,

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!