Michael J. Northrop

Parts Feeding with a Throwing Robot

Abstract

This thesis discusses how a simple one-joint robot can orient polygonal
objects by exploiting the mechanics of throwing and catching. Ordinarily,
automated parts feeding is done using a robot arm with enough joints to
grip and fully constrain the motion of the object. By incorporating the
knowledge of throwing dynamics we can replace hardware with intelligent
software, thereby making the process cheaper, more reliable, and more
flexible for handling different tasks.

We implemented this parts feeding technique using a one-joint robot,
consisting of a motor and a flat rectangular arm, and a tilting air
hockey table to simulate a slow time scale for throwing and catching
floating parts. We have developed planners for the arm which calculate
a sequence of one or more throws to take an object from an unknown
initial state (edge and position) to a known final state.

Source: Masters Thesis, Northwestern University, 1999.