Grounding semantic categories in behavioral interactions: Experiments with 100 objects

摘要: Abstract From an early stage in their development, human infants show a profound drive to explore the objects around them. Research psychology has shown that this exploration is fundamental for learning names of and object categories. To address problem robotics, paper presents behavior-grounded approach enables robot recognize semantic labels using its own behavioral interaction with test method, our interacted 100 different grouped according 20 The performed 10 behaviors on them, while three sensory modalities (vision, proprioception audio) detect any perceptual changes. results was able use multiple sensorimotor contexts order large number Furthermore, category recognition model presented identify can be used specific Most importantly, robot’s reduce time by half dynamically selecting which exploratory behavior should applied next when classifying novel object.