Date of Award

6-2012

Document Type

Open Access

Degree Name

Bachelor of Science

Department

Computer Science

First Advisor

John Rieffel

Language

English

Keywords

evolution, design, fabrication, printing

Abstract

Evolutionary algorithms have had success in designing complex objects, ranging from antennae used in NASA's Space Technology 5 mission to astronomical telescope lenses. However, evolutionary design is limited by the ability of a simulation to accurately represent the physical world. Addition-ally, evolved designs may be well described, but they carry no set of speci˝c instructions describing how to physically create such a design. Evolutionary Fabrication (EvoFab) recti˝es this: EvoFab is a machine built upon a process that can, in principle, automatically invent and build anything, from soft robots to new toys, by evolving the process, not the product. We have designed EvoFab, which consists of four components: A) a genotype for printing objects, consisting of a linear set of instructions sent to a Fab@Home, an open-source 3D printer; B) a way to evaluate printed objects using custom machine vision algorithms; C) a way to automate printing by implementing a cus-tom conveyor belt; D) a way of elaborating upon designs by implementing a genetic algorithm. In the near term, we aim to produce an evolved arch. Current results indicate increased ˝tness over time. Future improvements are possible through restrictions in extrusion along the Y-axis as well as re˝ning ˝tness evaluation to be less exploitable.

Download

Share

COinS