A NICE GESTURE
Actroid-SIT — a lifelike robot from Japanese firm Kokoro —
hasn't received as much attention as her cousin, Geminoid F, which
happens to be a copy of a real woman. While Geminoid F is a
teleoperated robot, Actroid-SIT can function autonomously —
talking and gesturing while interacting with people. In fact,
researchers have recently demonstrated how improvements to
Actroid's behavior can make it look smarter and more expressive
than your average android.
Actroid makes eye contact and gestures in the direction of a
person trying to speak to her, allowing it to adeptly handle crowds
of people. To develop the new behavior, researchers from Nara
Institute of Science and Technology studied how individuals and
groups interacted with the robot. Based on their observations, they focused on two new features which they call
"interruptibility" and "motion parameterization," hoping to improve human-robot interaction.
The first feature lets Actroid-SIT handle interruptions in a graceful way. In their human-robot interaction experiments, the
researchers noted that interruptions occurred about 26 percent of the time. For example, the human switched topics or
handed off the speaking role to someone else. The problem is that despite these interruptions, the robot would obliviously
carry on until it finished its spiel. Totally not socially elegant!
With the new interruptibility feature, the robot can immediately end its current topic and elegantly transition to the new
response. Those few seconds count — people interacted significantly longer with the robot when it was interruptible.
Actroid-SIT’s motion parameterization system gives her 18 gestures like pointing or waving to let her adapt to the location
of the speaker, making the person feel like the robot is really paying attention to him or her.
So, even though talking with the Actroid is
still far from a natural conversation, the
researchers say this improvement makes a big
difference in how people perceive the robot.
Participants called the android more friendly,
sensitive, sophisticated, and warm when the new
gesturing system was used, compared to a
normal gesturing approach.
ALICE IN WANDERLAND
Insects are masters of the swarm. Bugs like bees, termites, and ants manage to do
all sorts of complicated and productive things, despite the fact that on an individual level,
each insect is really not that smart. So, creating complex behaviors from simple systems
is appealing to roboticists. However, researchers from the New Jersey Institute of
Technology in Newark, and at the Research Centre on Animal Cognition in Toulouse,
France, are using swarms of ant-like robots to efficiently navigate networks without any
sort of cleverness at all.
These robots are called Alice — either collectively or as individuals. Their behavior
is based on Argentine ants and, as such, they have very, very primitive sensing systems —
little more than a couple of light sensors. The little robots were released into a simple
network maze where they wandered around a bit looking for an objective while trying
to take the least number of turns possible. Wherever they went, they left a trail of
"pheromones" as lights turned on above them. All of these behaviors are what ants do,
and it turned into a very effective way of autonomously discovering an efficient path.
It appears this research wasn't intended to be specifically about robots, but rather
to use robots to try and figure out how ants manage to do ant-y things despite having
tiny brains and lousy eyesight.
"This research suggests that efficient navigation and foraging can be achieved with
minimal cognitive abilities in ants," commented lead author Simon Garnier."It also shows that the
geometry of transport networks plays a critical role in the flow of information and material in ant, as well
as in human societies."
SERVO 06.2013 25