The Goal of Robotics Research in ARRG (GRR, ARRG)

The goal of the Arizona Robotics Research Group (ARRG) is to further our understanding of learning and development by creating artificial agents– both softbots and robots– that have to solve many of the same problems that humans have to solve.  Robots in the ARRG are essentially “baby robots”, which are created with minimal skill sets and uninterpreted sensors and go on to develop higher-level motor and perceptual skills from prolonged interaction with the world.  Much of this work focuses on the idea of “curiosity based learning” – learning driven by an intrinsic desire to better explain and control the world rather than the desire to maximize some external (human defined) reward.  Other robots are designed primarily to learn about human-robot interaction, particularly human-robot teaching.

Developmental Robots

One example of a developmental robot is one which uses curiosity based learning to discover the presence or absence of something which produces sounds contingent on the robot’s vocalizations, and uses that to train its visual system to localize in space those contingent things.  When we put this on a robot baby, and asked participants to “make the baby excited”, it was able to discover a category of thing which (a) responds contingently in a certain way, and (b) looks like a human face or torso.

Affective Computing

Human interaction relies on recognizing subtle cues in the face and voice.  We have developed software for recognizing emotions in speech, and emotions in the face, as well as other facial movements such as blinks, eyebrow raises, and a variety of other facial actions.

Ongoing Human Robot Interaction (HRI) studies

Much of our focus is on allowing robots to learn from natural human instruction.  We are currently recruiting participants for a study about robots that can learn from human teachers.  Please email ianfasel at cs dot arizona dot edu if you are interested in becoming a participant.

Affiliated Labs at the University of Arizona

Robotics and Neural Systems Lab, PI: Tony Lewis

Computational and Experimental Neuroscience Laboratory, PI: Jean-Marc Fellous