Doman’s Inclined Floor Method for Early Motor Organization Simulated with a Four Neurons Robot - Foundations on Natural and Artificial Computation - page 117

Table 1. Number of iterations that takes for the robot to learn. Each iteration corre-
sponds to one movement of the robot.

Mean Standard Deviation

Stripes 32.00 11.15

Sound 21.69 5.63

Acceleration 27.53 4.15

Stripes+Sound 19.23 3.57

Sound+Acceleration 21.15 3.44

Stripes+Acceleration 24.07 6.72

Stripes+Sound+Acceleration 19.38 2.99

In the case of having three stimuli, the number of iterations that it takes for
the robot to learn is less than when having two or one stimuli. This result is
consistent with Doman's methods in which parents are suggested to combine all
these stimuli for improving the methods' ecacy.
It can be also noticed that the presence of sound (the noise of a maraca when
the robot performs correct movements) yields the lower number of iterations,
either when the sound is alone or combined with different kinds of stimuli. This
happens because the intensity of sound is always the same, regardless of how
much the robot slides on the floor (which depends on the circumstances like the
adjustment of the robot brakes and the slipperiness and inclination of the ramp).
According to our studies, sound seems to be an important stimulus for stimu-
lating crawling in the robot and, hypothetically in children. Clapping, shouting
or even playing a musical instrument would help children to consolidate the suc-
cessful motor commands that make them crawl downwards on the inclined floor.
All these suggestions should be empirically tested with children.

5Con lu on

In this work we have constructed a very simple robotic structure with three type
of sensors and two independent wheel sets. A fully interconnected four-neurons
neural network receive sensory inputs and controls the wheel sets. Each neuron
performs a single movement of a specific wheel set. For example, when neuron 1
fires, it only produces clockwise movements in the frontal wheel set. This robotic
structure was placed in an inclined floor to mimic Doman's method for stimulat-
ing crawling abilities in children. Other stimuli like sound or alternating black
and white stripes painted on the ramp are also added to the robotic environment
for the purpose of mimicking the arrangement of the environment in Doman's
method.
Cybernetic principles envisioned by Doman's team are present in the way
our robotic neural network works. For example, the presynaptic rule (used for
synaptic weight adjustment) and our rule for modeling intrinsic plasticity are
consistent with cybernetic and biological principles.