Page 90 - Introduction to AI Robotics

3.2 What Are Animal Behaviors?
PATTERN 73
SENSOR BEHAVIOR OF MOTOR
INPUT
ACTION
Figure 3.3 Graphical definition of a behavior.

3.2 What Are Animal Behaviors?

Scientists believe the fundamental building block of natural intelligence is
BEHAVIOR abehavior. A behavior is a mapping of sensory inputs to a pattern of mo-
tor actions which then are used to achieve a task. For example, if a horse
sees a predator, it flattens its ears, lowers its head, and paws the ground. In
this case, the sensory input of a predator triggers a recognizable pattern of a
defense behavior. The defensive motions make up a pattern because the ac-
tions and sequence is always the same, regardless of details which vary each
episode (e.g., how many times the horse paws the ground). See Fig. 3.3.
Scientists who study animal behaviors are called ethologists.They often
spend years in the field studying a species to identify its behaviors. Often
the pattern of motor actions is easy to ascertain; the challenging part is to
discover the sensory inputs for the behavior and why the behavior furthers
the species survival.
REFLEXIVE BEHAVIOR Behaviors can be divided into three broad categories. 10 Reflexive behaviors
STIMULUS-RESPONSE are stimulus-response (S-R), such as when your knee is tapped, it jerks up-
ward. Essentially, reflexive behaviors are “hardwired”; neural circuits ensure
that the stimulus is directly connected to the response in order to produce the
REACTIVE BEHAVIOR fastest response time. Reactive behaviors are learned, and then consolidated to
where they can be executed without conscious thought. Any behavior that
involves what is referred to in sports as “muscle memory” is usually a re-
active behavior (e.g., riding a bike, skiing). Reactive behaviors can also be
changed by conscious thought; a bicyclist riding over a very narrow bridge
CONSCIOUS BEHAVIOR might “pay attention” to all the movements. Conscious behaviors are delibera-
tive (assembling a robot kit, stringing together previously developed behav-
iors, etc.).
The categorization is worthy of note for several reasons. First, the Reactive
Paradigm will make extensive use of reflexive behaviors, to the point that
some architectures only call a robot behavior a behavior if it is S-R. Second,
the categorization can help a designer determine what type of behavior to
use, leading to insights about the appropriate implementation. Third, the