CADML – the future language of part feeding?

Assembly Automation

ISSN: 0144-5154

Article publication date: 1 December 2002




Goldberg, K. (2002), "CADML – the future language of part feeding?", Assembly Automation, Vol. 22 No. 4.



Emerald Group Publishing Limited

Copyright © 2002, MCB UP Limited

CADML – the future language of part feeding?

CADML – the future language of part feeding?

Ken Goldberg

Keywords: Parts feeding, Part presentation, CAD, HTML

“When it comes to technology, most people overestimate the impact in the short term and underestimate it in the long term” – Arthur C. Clarke

Although the economic impact of the Internet was overestimated in the short term, it has dramatically changed the way we exchange information. Now anyone with a net connection not only can read documents from a huge online library, but also write their own documents and easily add them to the shared library.

What does this have to do with part feeding, the subject of this issue of the Assembly Automation journal?

Part feeders continue to be a major bottleneck for automated assembly: a typical vibratory bowl feeder requires 200 h to create, with 190 of those hours spent handcrafting a custom helical track that will reject all but a single orientation of the part without getting jammed. Custom feeder bowls are a source of line failure and introduces substantial delays into the setup and changeover of automated assembly lines.

My hypothesis is that the key development that triggered the success of the Internet could also be applied to streamline the setup of part feeders. For the Internet, the key development was a new standard for documents called hyper text markup language (HTML). HTML documents use only a few dozen very simple parameters and hence can be easily developed and processed.

Just as web browsers accept any HTML document and properly present it on any display, we need part feeders that can accept any part and properly orient it for assembly. An analogous standard for industrial parts, call it CADML, would define parts using simple parameters in such a way that it would be fast and easy to build a modular feeder (or fixture) for parts defined in CADML.

A number of research and commercial efforts have attempted to create a Universal Turning Machine: a robotic device whose software can be adapted to feed a broad variety of parts. Although great progress has been made, existing systems are too costly or difficult to program for many applications.

Rather than trying to make flexible feeders that work for the widest possible class of parts, the idea is to develop flexible feeders for a small parameterized class of parts. We must define a part language, CADML, such that any part defined in CADML will be feedable.

What’s striking about the HTML standard is its simplicity: unlike other document standards like DOC or PS or PDF, HTML has far fewer parameters. But what it loses in flexibility, it gains in compatibility. The idea is to do the same for computer aided design of industrial parts.

We might start, as was done with HTML 1.0, with a very simple standard: CADML 1.0, and gradually expand it with new parameters to CADML 2.0, and so on. For example, CADML 1.0 might define a class of screws with parameters for shaft length, width, pitch, head type and radius. The equivalent of the HTML browser would be the CADML feeder. Along with the CADML language we develop modular hardware for a feeder that can be easily adapted to feed any screw defined in CADML.

Once CADML is demonstrated and made available, other groups would expand the language and hardware to define other classes of parts, say cylinders, then parallelpipeds, and incrementally build up the set of parameters to allow increasingly broad classes of parts. The expansion will require a group effort; it cannot be achieved by one lab or company alone.

HTML’s combination of simplicity, bilaterality, and incremental development worked extremely well for document exchange. Let’s not underestimate what it might do for automated assembly.

Related articles