Distributed manipulation using naturally existing force fields.

Abstract

Many distributed manipulation systems are capable of generating planar force fields that act over the entire surface of an object to manipulate it to a stable equilibrium within the field. Passive air flow and other physical phenomena naturally generate force fields via linear superposition of logarithmically varying radial potential fields. The main advantage of these fields is that they are realizable through very simple actuation. However, they do not lend themselves to analytical prediction of net forces or equilibria. In this thesis, I present a distributed manipulation system with naturally existing passive force fields. First, I develop theoretical backgrounds to understand properties of naturally existing passive force fields such as the logarithmic force field of air flow. I present an efficient means of numerically computing the net force and moment exerted by such fields on objects composed of multiple simple shapes, as well as efficient means of finding equilibrium points on these fields. The second part concerns manipulation issues. Based on the analysis of the first part, I propose two possible manipulation concepts. One concept is a direct application of potential field combinations. With proper combinations of flow sinks, one can predict the equilibrium point efficiently so as to place an object at a certain position with known orientation. I propose a squeeze-like sequential manipulation using relatively simple flow fields to bring an object with a unique pivot point to a unique pose. A numerical analysis is provided to check the uniqueness of an object pivot point. The other proposed manipulation concept is to manipulate an object against a stationary barrier. Frictional contact mechanics in a logarithmic potential force field is studied and an algebraic representation is employed to study orientation controllability of the manipulation system. Experimental implementations demonstrate the feasibility of manipulation using logarithmic potential force fields.