SCS Undergraduate Thesis Topics
|Stuart Anderson||Reid Simmons||Single Point of Contact Manipulation of Unknown Objects|
Working in the presence of uncertainty is a fundamental problem in robotics. We are developing a method for performing single point of contact manipulation on initially unknown rigid bodies. Our approach adapts methods which have been used in the context of other robotics problems to this particular domain.
We consider a robot that learns about the world through force feedback from its manipulator, fiducial tracking of the object's motion, and visual detection of the object's silhouette. Based on the information obtained from these sources we wish to generate a sequence of manipulation behaviors which minimizes the expected time required to move the object into a desired pose.
We are developing behavioral primitives which are robust in the presence of uncertainty, methods for representing and reducing that uncertainty during manipulation, and planning software that reasons about the relative values of the available behaviors based on current knowledge. Our experiments are performed with a simulated world and robot.