Current Research Projects:

• Autonomous Robot Tool Use
  Collaborators: Alexander Stoytchev
  
  Tool use is one of the hallmarks of intelligence and is fundamental to human life. Many animals have also been observed to use tools, indicating that such an ability is a general adaptation mechanism that helps overcome physical limitations imposed by an organism’s anatomy.
  
  For a robot to adapt to human environments, it needs to be able to recognize, reason, and learn about the functional properties of different tools it encounters. In this work, I address several important questions: how can a robot represent the functional properties of tools? How can a robot learn the affordances of a tool through active interaction? How can a robot detect and generalize functional similarities between multiple objects?
  
  Developmental psychology, cognitive science and neuroscience offer useful insights when addressing those questions. My work in this area has resulted in several publications in the proceedings of the IEEE International Conference of Development and Learning, as well as the AAAI student program.
  


• Autonomous Learning of the Acoustic Properties of Objects
  Collaborators: Mark Wiemer, Alexander Stoytchev
  
  While our sense of vision is always constrained to a particular viewing direction, our auditory sense allows us to infer events in the world that fall outside the reach or range of other sensory modalities. Studies have shown that humans can perceive the physical properties of objects from the sounds that they produce. The importance of everyday natural sounds is perhaps best summarized by Don Norman in his book, “The Design of Everyday Things”:
  
  “[…] natural sound is as essential as visual information because sound tells us about things that we can't see, and it does so while our eyes are occupied elsewhere. Natural sounds reflect the complex interaction of natural objects: the way one part moves against another; the material of which the parts are made -- hollow or solid, metal or wood, soft or hard, rough or smooth."
  
  Most robots today, however, do not use environmental sounds as a source of information about events in their surroundings. Nevertheless, there are many situations in which such an ability would help a robot detect and reason about events in a human-inhabited environment. For example, if a robot accidentally knocks over an object that it cannot see, the sound generated by the interaction will be the primary source of information about the nature of the object. Similarly, if a human interacts with an object outside the robot's field of view, the robot will only be able to recognize the type of object and the type of interaction using the detected sound.
  
  In this line of work, I investigate the problem of how a robot can learn to recognize objects and behavioral interactions through active interaction and using only auditory information. Our work has resulted in a paper published and presented at the Robot Manipulation Workshop at the 2008 Robotics Science and Systems conference.
  
  


• Tool Tracking for Complext Manufacturing Tasks
  Collaborators: Matt Miller, Peter Wong, Alexander Stoytchev
  
  Grant Support:Deere & Company (a.k.a John Deere)
  
  This project is aimed at developing intelligent software which monitors and assists human workers at a given manufacturing task (e.g. welding, using an air gun, etc). The goal of the system is to detect errors in real time, as well as alert the workers in a user-friendly manner ragarding the location of the mistake made. The highlights of this work involve tracking human actions and movements, as well as making intelligent inferences regarding whether the manufacting task is being executed correctly. This project involves the development and deployment of a system capable of operating robustly in a real factory environment, as well as a virtual reality cave, such as the ones operated by the department of Human-Computer Interaction at Iowa State University.