Nonprehensile Robotic Manipulation: Controllability and Planning
K. M. Lynch


A good model of the mechanics of a task is a resource for a robot, just as actuators and sensors are resources. The effective use of frictional, gravitational, and dynamic forces can substitute for extra actuators; the expectation derived from a good model can minimize sensing requirements. Despite this, most robot systems attempt to dominate or nullify task mechanics, rather than exploit it. There has been little effort to understand the manipulation capabilities of even the simplest robots under more complete mechanics models.

This thesis addresses that knowledge deficit by studying graspless or nonprehensile manipulation. Nonprehensile manipulation exploits task mechanics to achieve a goal state without grasping, allowing simple mechanisms to accomplish complex tasks. With nonprehensile manipulation, a robot can manipulate objects too large or heavy to be grasped and lifted, and a low-degree-of-freedom robot can control more degrees-of-freedom of an object by allowing relative motion between the object and the manipulator.

Two key problems are determining controllability of and motion planning for nonprehensile manipulation. The first problem is to determine whether the goal state of the object is reachable by nonprehensile manipulation, and the second is to find a manipulator motion to bring the object to the goal state.

Part I studies these problems for quasistatic nonprehensile manipulation by pushing. I elucidate the controllability properties of objects pushed with point and line contact, and I describe a planner that finds stable pushing paths among obstacles. Pushing can also be used to simplify the hardware of a parts feeder; a one-degree-of-freedom robot, positioned over a conveyor, can position and orient any polygonal part on the conveyor by a series of pushes.

Part II of the thesis studies dynamic nonprehensile manipulation. By considering dynamics, I show that even a one-degree-of-freedom robot can take a planar object to a full six-dimensional subset of its state space. Then I describe a planner that finds manipulator trajectories to perform dynamic tasks such as snatching an object from a table, rolling an object on the surface of the arm, and throwing and catching.

To demonstrate the feasibility of nonprehensile manipulation, all planners have been implemented on actual robot systems.