Select your localized edition:

Close ×

More Ways to Connect

Discover one of our 28 local entrepreneurial communities »

Be the first to know as we launch in new countries and markets around the globe.

Interested in bringing MIT Technology Review to your local market?

MIT Technology ReviewMIT Technology Review - logo

 

Unsupported browser: Your browser does not meet modern web standards. See how it scores »

Office Autonomy

A deep-space probe obviously requires much more autonomy than, say, a photocopier. But heavily used office machines must meet a similar demand for reliability and efficiency-especially when the boss asks for 30 bound copies of the annual report for a board meeting that starts in three minutes. That’s why engineers at Xerox and its recently spun-off Palo Alto Research Center (PARC) have begun to build immobot intelligence into high-end copy machines.

Like a big-city airport with flights arriving and departing on shared runways, a copier’s biggest challenge is scheduling: it needs to launch the next sheet of blank paper into the printing system as soon as the previous one is out of the way. But as Xerox’s previous generation of copiers and printers became increasingly complex, says Daniel Bobrow, an artificial-intelligence researcher in PARC’s Systems and Practices Lab, the company noticed that the process was taking more and more time, especially when users selected options such as two-sided copying, sorting, and stapling.

What’s more, the machines’ heuristic scheduling software lacked full understanding of the interactions among stations, so it treated each of the processes as separate tasks, all of which had to be completed before the next sheet could enter the work path. “The only way to anticipate every possible request was to take any request that was different from the ones [programmers] had thought of and break it down into two or more jobs,” Bobrow explains. Even worse, the scheduling software had to be substantially rewritten each time Xerox wanted to add new features to its machines.

But the DocuColor 2045 and the DocuColor 2060, Xerox’s flagship printer-copiers, have taken a big step toward sentience. Both truck-size $90,000-to-$120,000 machines come with model-based scheduling programs, which PARC researchers designed to optimize moment-to-moment paper flow. “You want to have a number of sheets going through in succession, and when you have two-sided copies coming back around, you want them to be interspersed with things doing the first side,” says Bobrow. “So you have to actually look at the timing of these various things and build a model of what’s going on.”

Say you’re making 70 copies of a booklet. Even before you press the Start button, a machine running PARC’s model can predict that stapling will be the slowest part of the job and communicate this fact to other components of the machine, allowing them to run concurrently and without creating a big backup. Another advantage, Bobrow says, is that the DocuColor controller models are built from smaller models hard-wired into each component. When a new station such as a scanner or sorter is added, it transmits its internal model to the central controller, providing a painless software upgrade.

“Model-based programming is doing a tremendous job for us in terms of improving [copier] productivity,” says Bobrow. Owners of the machines can add and remove components as they like, and they can let the machines toil unattended for hours. What’s more, he says, Xerox can use its current models to simulate new equipment configurations and “evaluate how things would work” before investing a dime in physical prototypes of its next-generation machines.

“This distinction between telling a system how to do its job and telling the system the end result you want is very fundamental,” says Robert Morris, director of IBM’s Almaden Research Center in San Jose, CA. IBM is working to build what it calls “autonomic” characteristics-model-based features, as well as others that employ classic heuristic programming-into products such as Web servers and storage networks. These features will allow the products to reconfigure themselves for optimal performance, depending on what’s being asked of them.

One project, for example, involves building IBM’s DB/2 database software with models that allow databases to learn from past queries and suggest ways to rewrite new ones to retrieve data faster and more precisely. Simple efficiency is dictating this move, Morris says. “We spend so much of our time managing these systems and so much of our money paying people to run them. We’d better stop spending so much on the tedium and more on the new technologies.”

0 comments about this story. Start the discussion »

Tagged: Computing

Reprints and Permissions | Send feedback to the editor

From the Archives

Close

Introducing MIT Technology Review Insider.

Already a Magazine subscriber?

You're automatically an Insider. It's easy to activate or upgrade your account.

Activate Your Account

Become an Insider

It's the new way to subscribe. Get even more of the tech news, research, and discoveries you crave.

Sign Up

Learn More

Find out why MIT Technology Review Insider is for you and explore your options.

Show Me
×

A Place of Inspiration

Understand the technologies that are changing business and driving the new global economy.

September 23-25, 2014
Register »