Out of ControlBy Edward Cone | Posted 2002-04-09 Email Print
Online exclusive: The agency wrote off $1.5 billion of its $2.6 billion investment to overhaul the nation's air traffic control computer systems. What went wrong? (Just about everything.)
Out of Control
Rather than admit the problem, IBM turned AAS into a research project on keeping distributed computer systems up and running on an almost nonstop basis, publishing papers instead of making concrete progress.
Deadlines vanished, and costs spiraled out of control. Only after the AAS project collapsed did the FAA set a more realistic requirement of no more than 5 minutes of downtime a year.
Another ex-IBMer who worked on AAS, Ellen Bass, says she quickly saw problems. "The science wasn't there, and the methodologies weren't there," says Bass, who spent the years following AAS' demise working as a human-factors specialist on submarine and aircraft control systems and now teaches at the University of Virginia. In addition to the unsolved problems of around-the-clock distributed computing systems, the AAS designers were venturing too far into uncharted territory with her specialty, the user interface.
Bass saw that a then cutting-edge plan to allow controllers to customize their workstations would make it virtually impossible to test whether the system would operate safely in every configuration. For example, by giving controllers the opportunity to customize the colors on their display, the design introduced the possibility that they might accidentally make planes appear invisible against the background. She suggested going back to the FAA with a proposal for a "simpler, cheaper, quicker" solution, but she was a junior member of the team whose opinions didn't count for much. Besides, her managers at IBM were "very concerned that we had promised FAA all this flexibility, and if we simplified we would be non-compliant," she says.
There were internal FAA politics at work, too. After the demise of PATCO (the Professional Air Traffic Controllers Organization), the FAA looked to computer automation as a means of allowing fewer controllers to handle more planes, and that was part of what AAS was supposed to accomplish. But before long there was a new union, NATCA (the National Air Traffic Controllers Association), which had every reason to frustrate that goal.
On paper, the AAS was also supposed to result in the consolidation of what today remain two separate functions performed by different facilities, the TRACONS that handle airport arrivals and departures and the En Route Centers that handle the long-haul portion of a flight.
The FAA wrote the consolidation plan into the project specifications without solving the political problems that would have to be overcome to make it happen. So the consolidation plan was eventually dropped, but the requirement that IBM produce a new workstation capable of handling both sorts of functions remained in the specs. "We were working toward a mission that was not even there," Britcher says. In other words, developers were creating a unified workstation for two functions that were now no longer going to be unified.
FAA officials say they have learned from the failure of the AAS and will not repeat the same mistakes. But one of the project's legacies is an enduring cynicism, from many fronts, about anything that FAA officials say.
"The FAA says it sees light at the end of the tunnel, but when you get to the end of that tunnel, there's another tunnel," says David Schaffer, the counsel to the U.S. House Aviation Subcommittee. Whenever he hears claims of progress, he can't help but think of what he kept hearing about the AAS until the very end.
"They would say there were a few problems, but they were being worked out. Everything seems to be going welluntil it collapses."