Towards Attentive Robots

This paper introduces Atentive Robots: robots that attend to the parts of their sensory input that are currently of most potential interest. The concept of selecting the most promising parts is adopted from human perception where selective attention allocates the brain resources to the most interesting parts of the sensory input. We give an overview of current approaches to integrate computational attention into robotic systems, with a focus on biologically-inspired visual attention methods. Example applications range from localization with salient landmarks over object manipulation to the design of social robots. A brief outlook gives an impression of how future ways to obtain attentive robots might look like