Carnegie Mellon Robotics Institute
I am interested in the mechanics of both robotic locomotion and manipulation. Each of the four projects described below has led to a better understanding of the low-level mechanics of common human and robotic tasks.
Mobile manipulation: An ongoing project in the manipulation lab has been to answer the question, "How can a mobile robot manipulate the world with its wheels?" We have explored this problem using a mobile robot with four wheels, which it uses to manipulate paper on a desktop. Sometimes the problem can be viewed as purely a locomotion problem, in which the robot moves relative to the object to be manipulated, and sometimes the problem can be viewed as a manipulation task, in which the robot grasps the object and moves relative to the world. Deciding when to switch between the two modes (and when to combine them) makes for an interesting problem in planning and control.
Time optimal differential drive locomotion: What is the fastest way for a wheelchair to get from one location and orientation to another? We have found the time optimal trajectories for a simple model of a differential drive moving in an obstacle-free plane. This result has important implications for the problem of path planning for mobile robots. In fact, the time optimal trajectories are known only for three classes of mobile robot: two varieties of steered car, and (now) the diff drive. Although the methods used to find the optimal trajectories are somewhat involved, the result is simple: the time optimal trajectories are combinations of spins in place and straight lines, and sometimes include an intermediate ('via') point.
Rigid body contact mechanics: The problem of how two contacting rigid bodies will move under an applied force is well known and has been extensively studied. I have done some work in this area with Jeff Trinkle and Eric Gottlieb at Sandia National Labs. We designed an algorithm to find all externally applied wrenches consistent with any possible contact mode between the two bodies. Although algorithms to achieve similar results have been proposed (Reuleaux, Erdmann, Mason, and many others), ours has some advantages. The algorithm is very general; it can be applied to all of the contact modes usually considered, but can also be extended to 'augmented' contact modes described by linear constraints on body velocities or accelerations. The algorithm is easy to implement, since it takes the form of a numerical rather than geometric procedure. Perhaps most importantly, the technique can be extended to three dimensional problems in a principled way.
Manipulation of flexible objects: Although the majority of robotic manipulation tasks are designed to involve only rigid bodies, human beings manipulate flexible objects every day. Even before breakfast we must deal with clothing, hair, shoe-laces, newspaper, and the packaging on our food. My thesis project will begin to answer the question of how robots can manipulate the flexible world.
|Research Interest Keywords|
|manipulation, mobile robots|
|The Robotics Institute is part of the School of Computer Science, Carnegie Mellon University.|
Contact Us | Update Instructions