Interactive Identification of Writing Instruments and Writable Surfaces by a Robot

This study asks the question: Can a robot learn to identify object-surface pairs that are useful for trace-making and, thus, facilitate writing? Writing is an important skill that humans use everyday and future household robots may have to master in order to be useful. Specifically, our robot performed scribbling behaviors with different objects on different surfaces and monitored which objects leave marks on which surfaces. The frequency with which marks were detected was used by the robot to group the writing instruments and the writable surfaces into meaningful categories, which capture their functional properties from the robot’s point of view.