Human Action Understanding and Anticipation: From humanoid robots to brain function and cognitive systems

I will give an overview of recent developments and challenges in the area of Robotics and Artificial Intelligence, with an emphasis of a particular perspective we have followed in my research group at the Institute for Systems and Robotics, of the Instituto Superior Técnico. We use artificial vision and humanoid robotics to help understanding certain human brain functions. In our work, we model human behaviour in an interdisciplinary research involving computational neuroscience, developmental psychology, and engineering.One aspect is related to neurophysiology and the discovery of the mirror neurons which suggest that both action understanding and execution are performed by the same (motor) areas of the brain. These neurons facilitate social learning amongst con-specific individuals, and may constitute the basis of non-verbal communication, a fundamental mechanism for humans to interact with each other and anticipate the actions of others.I will present a computational model inspired by these findings and that outperforms (classic) approaches in gesture recognition from video. I will detail experiments with human subjects that we used to record data that unveils how humans communication through eye-head and body movements. The talk will focus on aspects of sensorimotor coordination, learning about affordances and social interaction.I will provide examples of the use of humanoid robots (with our first platform, Baltazar, the most-sophisticated, open source humanoid robot that we co-designed, the iCub, and our social robot Vizzy) as testbeds to study human cognition, learning and sensorimotor coordination, while offering engineers new approaches to build artificial systems.