Open Object Recognition for Humanoid Robots

摘要

Robots must be able to adapt gracefully to frequent and dramatic changes in their workspace if they are to operate successfully in human- centered environments, as opposed to controlled industrial settings. At the MIT Humanoid Robotics Group, investigators are developing methods that permit their robots to deduce the structure of novel activities, adopt the vocabulary appropriate for communication about the task at hand, and learn about the appearance and behavior of unfamiliar objects. The latter ability is discussed in this paper. The humanoid robot, 'Cog,' uses active exploration to resolve visual ambiguity in its workspace. As Cog accumulates experience, it clusters episodes of object interaction to learn the appearance and properties of novel, unfamiliar objects. This process is called 'open object recognition.' An operator can then introduce names for objects to facilitate further task- related communication.