This is a map of some of the areas that the two
Lingodroid robots have seen and shared names of.
The red scribbling represents a path around objects
that the robots can take or have taken to reach each
named area or destination.
communication,” explains Dr. Schulz. The beeps are
used in language games.
The robots learn what the words mean through
these games. Each language game establishes
certain concepts or topics about the word at hand.
“The Lingodroids learn an instance of a word in a
single experience in the same way that a child does.
They refine what they understand about the word in
later interactions in which they use the word. New
words compete with old words to determine which is
the best word for a given situation,” says Dr. Schulz.
Locations are a great place for robots to start
creating their own language because the meanings
of the words are concrete and specific, and the robots can
easily associate the words with things they understand.
experienced by the robots but rather have been indirectly
referred to through language,” says Dr. Schulz.
The Games Robots Play
The two Lingodroids start by playing “Where are we?”
games. In these interactions, if the robots cross into an area
that they have not named yet, one or the other of them
will invent a word for it. “The robot then communicates the
word to the other robot upon meeting, defining the name
of the location,” says Dr. Schulz. These words are known as
toponyms. “Topo means place and nym means name,” Dr.
The next games the robots play are “How far?” and
“Which direction?” These communications
enable the robots to develop relationship
words that are much like our human English
prepositions. “The resulting language makes
use of location, distance, and direction
words which enable the robots to refer to
new places based on their relationship to
known locations,” Dr. Schulz continues.
These are powerful and generative
languages because they enable the robots
to refer to new places they haven’t been to
yet or even places they imagine beyond the
edges of their known environment.
The robots are able to associate the
new or unknown places with positions
within the global framework of the world
they know and have currently mapped out.
Each unvisited but named new or unknown
location or place is looked at as a new
experience waiting for the robot that it can
find based on its known map. “We call
these new experiences pseudo-experiences.
We use the term pseudo as the places are
The Lingodroids each have numerous components that
add to their linguistic and navigational capabilities. The
mapping is performed by the SLAM system referred to
earlier. This system constructs the maps of the robots’
experiences based on the paths they took to discover each
new location in their world.
In this photo, both Lingodroids are at
one of the areas in the maze. This
particular area has been named Kuzo.
This also shows you the front and back
of the Pioneer3 DX robot platform.
SERVO 11.2011 11