Teaching STAIR to Identify and Manipulate Tools

One of the current areas of research in robotics is in the development of robots which can interact with humans in performing a variety of tasks. In particular, there is a desire to enable robots to use objects as tools to perform tasks. For example, assembly tasks may require the robot to use a screwdriver to tighten screws or a hammer for nailing. In this project, I am concerned with the problem of robotic tool manipulation for the purpose of carrying out higher level tasks. This involves developing a learning algorithm to detect the tip of a tool (the control point), as well as algorithms for incorporating the tool as an extension of the robot system and controlling its subsequent movement. I performed this research on the STAIR robot in the CS Department. Ashutosh Saxena and others have recently completed work in identifying grasping locations on generic objects . Their algorithm enables STAIR to pick up most objects (cups, dishes, pens, tools, etc.) within its field of vision. The goal of my project is to extend this work to allow STAIR to manipulate tools once it has picked them up with the grasping algorithm. One long-term application being considered is to enable STAIR to construct a bookcase from IKEA, given a set of simplified tasks.