Affective Visuomotor Interaction: A Functional Model for Socially Competent Robot Grasping

In the context of human-robot social interactions, the ability of interpreting the emotional value of objects and actions is critical for robots to achieve truly meaningful interchanges with human partners. We review here the most significant findings related to reward management and value assignment in the primate brain, with particular regard to the prefrontal cortex. Based on such findings, we propose a novel model of vision-based grasping in which the context-dependent emotional value of available options (e.g. damageable or dangerous items) is taken into account when interacting with objects in the real world. The model is both biologically plausible and suitable for being applied to a robotic setup. We provide a testing framework along with implementation guidelines.